Special:Special: Applications Applications and Service and ... · for the Mobile Userfor the Mobile User. Contents JOINT ERCIM ACTIONS 4 8th Workshop on Formal Methods for Industrial

European Research Consortium for Informatics and Mathematicswww.ercim.org

Number 54, July 2003

Special:Special:

Applications Applications and Service and Service Platforms Platforms for the Mobile Userfor the Mobile User

Contents

JOINT ERCIM ACTIONS

4 8th Workshop on Formal Methods for Industrial Critical Systems by Thomas Arts, IT University ofGöteborg, Sweden and Wan Fokkink,CWI

4 ERCIM PhD Fellowship Programme

5 GRIDS Expert Group

5 Euro-Legalby Heather Weaver, CCLRC

6 News from W3C

SPECIAL THEME

8 Applications and Service Platforms for the Mobile User - Introductionby John Krogstie, NTNUInvited article:

10 In Search of Knowledge about Mobile Usersby Ee-Peng Lim, Yida Wang, Kok-Leong Ong, NanyangTechnological University, Singapore and San-Yih Hwang,National Sun Yat-Sen University, Taiwan

11 The Missing Link – User-Experience and IncrementalRevenue Generation on the Mobile Internet by Barry Smith, ChangingWorlds, Dublin and UniversityCollege Dublin

12 User-Centric Peer-to-Peer Service Environment for Interaction with Networked Appliancesby Daniel Pakkala, Pekka Välitalo, Pekka Pääkkönen and Juhani Latvakoski, VTT

14 Kontti — Context-aware Mobile Portalby Juha Kolari, VTT

15 Contextual Mediation enables Appropriate Data Selectionby Dan Chalmers, Morris Sloman and Naranker Dulay,Imperial College London, UK

16 mBlog: a Mobile Information Service for Allby Emmanuel Frécon, SICS

18 A Mobile Schema Browser for Integration Specialistsby Gerald O’Connor and Mark Roantree, Dublin City University, Ireland

19 A Peer-to-Peer Middleware for Mobile TeamWorkby Carlo Ghezzi, Gianpaolo Cugola, and Gian Pietro Picco,Politecnico di Milano, Italy

21 Fluid Computingby Marcel Graf, IBM Zurich Research Lab, Switzerland/SARIT

23 Designing Usable Multi-Platform Interactive Applicationsby Fabio Paternò and Carmen Santoro, ISTI-CNR

24 Interoperating with Heterogeneous Mobile Servicesby Paul Grace and Gordon S. Blair, Lancaster University, UK

26 NAC: An Architecture for Multimedia Content Adaptationfor Mobile Devicesby Nabil Layaïda and Tayeb Lemlouma, INRIA

27 Integrating GSM Networks and Internet: New Unstructured Servicesby Rita Arrighi, Maurizio A. Bonuccelli, Francesca Lonetti and Francesca Martelli, ISTI-CNR

28 From Ad Hoc Networks to Ad Hoc Applicationsby Benoît Garbinato and Philippe Rupp, University ofLausanne, Switzerland

30 Light Support for Dynamic and Pervasive Services on P2P Networksby Stéphane Frénot, Anis Krichen and Stéphane Ubéda,INRIA and INSA-Lyon Scientific and Technical University

Next issue:October 2003

Special theme:Machine Perception

Cover photocourtesy of VTT

3

31 Towards a Virtual Home Environment Application Serverby Ali Hafezi and Klaus-Peter Eckert, Fraunhofer FOKUS

32 M-commerce API Package for Mobile Phonesby Ioannis G. Askoxylakis, Diomedes D. Kastanis and Apostolos P. Traganitis, ICS-FORTH

34 Wireless Ad Hoc Network for Dublin (WAND): A Large-Scale Ad Hoc Network Test-Bedby Stefan Weber, Vinny Cahill, Siobhan Clarke and Mads Haahr, Trinity College Dublin

35 Lightning Active Node Engine — An Active Network User Service Platformby Sami Lehtonen, VTT

36 Middleware Support for Ubiquitous Computingby Luc Hogie and Christian Hutter, Luxembourg University of Applied Sciences

38 Sentient Objects: Towards Middleware for Mobile Context-Aware Applicationsby Gregory Biegel and Vinny Cahill, Trinity College Dublin

39 Database Components on Chipby Nicolas Anciaux, Luc Bouganim and Philippe Pucheral, INRIA

41 Object-driven Application Design: From Mobility towards Ubiquitousnessby Davide Carboni, Gavino Paddeu and Stefano Sanna,CRS4, Italy

42 Development Tools for Dynamic Environment-AdaptedMobile Interfacesby Anthony Savidis and Constantine Stephanidis, ICS-FORTH

44 Introducing New Mobile Services Fasterby Timber Haaker and Oscar Rietkerk, TNO Telecom, The Netherlands

46 Uluru: Mobile Interactive Multimedia Experimental ServiceEnvironmentby Johan de Heer, Andrew Tokmakoff, Henk Eertink and John Anijs, Telematica Instituut, The Netherlands

48 Next Generation Public Travel Information Systemsby Karl Rehrl, Harald Rieser and Siegfried Reich, SalzburgResearch Forschungsgesellschaft/AARIT

49 Mobile Communication in Newspaper Distribution and Transportby Olli Kuusisto and Janne Pajukanta, VTTInvited article:

50 Values of Mobile Applications to End-UsersUniversity of Nebraska-Lincoln, USA

R&D AND TECHNOLOGY TRANSFER

53 Algorithmic Clustering of Musicby Rudi Cilibrasi, Paul Vitányi and Ronald de Wolf, CWI

54 Extension of Electronic Documentsto enable Server-IndependentArchitecturesby Paola Carrara, Giuseppe Fresta,Daniela Poggioli, CNR

55 COSMOS: Safe and OptimalOperation in Complex IndustrialSystemsby Péter Inzelt, SZTAKI

57 Leading-Edge KnowledgeManagement for Public Employeesby Simon Lambert, CCLRC

58 Radiosity for Realistic Real-TimeRenderingby François Cuny, VSP Technology,France

59 Ultra-High-Speed Image Processing: Over 10 000 Frames Per Secondby Ákos Zarándy and Róbert Fekete,SZTAKI

60 Internet-Based Learning and Collaboration for Dispersed Co-Workers by Dian Tan and Peter Mambrey

62 Internet and Isolated Villages: An Opportunity for Developmentby Diego Puppin, ISTI-CNR

EVENTS

63 SCILAB Consortium launchedby Claude Gomez, INRIA

65 World Wide Web Conference in Hungaryby László Kovács, SZTAKI

66 Announcements

55 IN BRIEF

4 ERCIM News No. 54, July 2003

JOINT ERCIM ACTIONS

The fourteen regular contributions to theworkshop, which were selected from 25submissions, dealt in large part with theapplication of formal methods in indus-trial case studies. Some notable applica-tion areas were smart cards, Java, MPEGand avionic systems. Two invitedspeakers were present: Reiner Hähnlefrom Chalmers University ofTechnology presented a paper entitled'Integration of Informal and FormalDevelopment of Object-Oriented Safety-Critical Software: A Case Study with theKeY System', and Werner Damm fromthe University of Oldenburg talked about'Recent Trends in the Development ofSafety-Critical Systems'.

The FMICS workshop series is organ-ised by the ERCIM Working Group onFormal Methods for Industrial CriticalSystems. On the occasion of the co-loca-tion with the ERCIM spring meetings,ten representatives from ERCIM insti-tutes within this Working Group were

invited to present an overview of currentresearch on formal methods in their ownresearch groups.

As a gesture of involvement, theEuropean Association for the Study ofScience and Technology handed out twobest paper awards, one being given toPaul Ziemann and Martin Gogolla fromthe University of Bremen, and one toStefan Blom and Simona Orzan fromCWI.

The FMICS Working Group hasmanaged to achieve a broad public visi-bility. This good interaction with thewider scientific community was recog-nised by the ERCIM board of directors,and as a result FMICS was, during theworkshop dinner, rewarded with theERCIM award for the most successfulWorking Group of 2002.

Link:FMICS Working Group:http://www.inrialpes.fr/vasy/fmics/index.html

8th Workshop on Formal Methods for Industrial Critical Systems by Thomas Arts and Wan Fokkink

The 8th Workshop on Formal Methods for Industrial Critical Systems (FMICS'03)was co-located with the ERCIM meeting in Røros in Norway, June 5-7, 2003. TheFMICS workshops aim to strengthen research on formal methods, and topromote the use of formal methods within industry. This year the workshopattracted over thirty participants.

ERCIM PhD FellowshipProgrammeERCIM offers 18-month fellowships inleading European information tech-nology research centres. Fellowshipsare available for PhD-holders from allover the world. Next Deadline for appli-cation: 30 September 2003

Fellowships are of 18 months duration,generally spent in two ERCIM memberinstitutes. The fellow will receive amonthly allowance which may varydepending on the country. In order toencourage the mobility, a member insti-tute will not be eligible to host a candi-date of the same nationality.

A preliminary list of topics includes butis not restricted to the scientific fieldscovered by the ERCIM WorkingGroups: • Applications of Numerical

Mathematics in Science• Constraints Technology and

Applications• Control and System Theory• Dependable Software-Intensive

Systems• Digital Libraries• E-Learning• Environmental Modelling• Formal Methods• Health and Information Technology• Image and Video Understanding• Matrix Computations and Statistics• Soft Computing• User Interfaces for All.

Conditions Applicants must: • have a PhD degree or be in the last

year of the thesis work• be fluent in English • be discharged or get deferment from

military service.• start the grant before October 2004 • have completed their PhD before

starting the grant.

How to applyApplications must be submitted online.A detailed description and the application form are available at:http://www.ercim.org/fellowship/

Simona Orzan proudly shows her best

paper award.

ERCIM President Gerard van

Oortmerssen presents the working group

award 2002 to Stefania Gnesi, coordinator

of the Working Group on Formal Methods

for Industrial Critical Systems.

5

JOINT ERCIM ACTIONS

Euro-Legal

News about legal information relatingto Information Technology fromEuropean directives, and pan-European legal requirements andregulations.

Unsolicited MailMost of us have received at some time orother unsolicited emails advertisingproducts and services. A number ofPreference Services now operate withinthe Communication Sector, which enable

consumers towithdraw theirnames andaddresses frommarketing circulationlists. Membero r g a n i s a t i o n sundertake to removethe addresses ofpeople from their

mailing lists who want this done.However, unscrupulous companies tendto trade names and addresses with otherscam promoters, and those mailing listsmay never be updated. The Stop NowOrders (EC Directive) Regulations 2001,which is a Code of Practice enforced bythe Advertising Standards Authority andthe Independent Committee for theSupervision of Standards of TelephoneInformation Services (ICSTIS), enablesthe Office of Fair Trading to bring aninjunction against traders breachingregulations on misleading advertising.Action can now be taken across EUborders so scams operating acrossEurope can be pursued.

But even worse than the marketing scamsthemselves is the sheer volume ofunsolicited commercial email (UCE), morecommonly known as "spam", that canshift the cost of advertising from theadvertiser to the recipient when with oneemail account a junk emailer can send amessage to millions of recipients, turningthat message into hundreds or possiblythousands of megabytes of data.

Vint Cerf, Senior Vice President, MCI isquoted as saying "Spamming is thescourge of electronic-mail andnewsgroups on the Internet. It canseriously interfere with the operation ofpublic services, to say nothing of the

effect it may have on any individual's e-mail system…..Spammers are, in effect,taking resources away from users andservice suppliers without compensationand without authorization." Whilst wewould all agree that spamming isannoying, it is true to say that it alsoconstitutes a theft of service.

One way of dealing with spam is to blockthe sites from which these UBE's(Unsolicited Bulk Emails) come. Pressurecan be put on spammers by blockingaccess to websites that persistentlyadvertise by UBE. However, spam isalmost always forged. The address fromwhich the message appears to havecome is made to look like a real orinnocent person.

Organisations such as EuroCAUCE (TheEuropean Coalition Against UnsolicitedCommercial Email) have been set up topromote legislation which would outlawUCE. European Directive 2002/58/EC onPrivacy and Electronic Communication,first published in July 2002, will beimplemented in Members States by 31October 2003. The Directive states "Forsuch forms of unsolicited communicationsfor direct marketing, it is justified torequire that prior explicit consent of therecipients is obtained before suchcommunications are addressed to them"."When contact details are obtained, thecustomer should be informed about theirfurther use for direct marketing in a clearand distinct manner, and be given theopportunity to refuse such usage"."Certain electronic mail systems allowsubscribers to view the sender andsubject line of an electronic mail, and alsoto delete the message without having todownload the rest of the electronic mail'scontent or any attachments, therebyreducing costs which could arise fromdownloading unsolicited electronic mailsor attachments. These modalities maycontinue to be useful in certain cases asan additional tool to the generalobligations established in this Directive."For more details on the Directive see:http://www.euro.cauce.org/en/amendments1a.html.

by Heather Weaver, CCLRCHeather Weaver regrets that she is unableto reply personally to emails or telephonecalls seeking legal advice.

ERCIM News No. 54, July 2003

GRIDs Expert GroupThe European Commission'sInformation Society Directorate(INFSO), headed by Commissioner ErkiiLiikanen and with Fabio Colosanti asDirector-General, has a new Unit (F2)led by Wolfgang Boch concerned withGrids for Complex Problem Solving.

The F2 unit arose from the work done onGRIDs within Wolfgang Boch’sprevious Unit concerned with IST andthe environment,including natural andman-made hazards andlong-term environmentalmonitoring.

A workshop was calledin January 2003 withsome key speakers. Itattracted more than 200participants. Among the key speakers,three formed a 'core group of experts' tomoderate discussions and to assist theCommission (in particular Roman Tirlerwho is the officer in charge of thisactivity) in pushing forward the researchpriorities and work programme. Thesethree are ERCIM members: Thiery Priol(INRIA), Domenico Laforenza (CNR)and Keith Jeffery (CCLRC).Subsequently names were suggested togrow the group to the current 14 experts.This larger group includes Seif Haridi(SICS) and Ludek Matyska (CRCIM).

The group is working very productively.It has convened in two meetings toprovide an expert's view on the five toseven year research priorities for theEuropean context. The final report enti-tled 'Next Generation Grid(s) –European Grid Research 2005 - 2010' isnow available on the CORDIS websiteat:http://www.cordis.lu/ist/grids/index.htmunder the heading 'GRIDs Highlights'.

Expert Group report for downloading: ftp://ftp.cordis.lu/pub/ist/docs/ngg_eg_final.pdf


W3C Approves Patent PolicyThe World Wide Web Consortium has approved the W3CPatent Policy based on review by the W3C AdvisoryCommittee and the public on 21 May 2003. Written by thePatent Policy Working Group, the policy received moresupport from the W3C members than any Recommendationin recent history. The W3C Patent Policy encourages royalty-free Web standards and aims to solve a specific problem – toreduce the threat of blocking patents on key components ofWeb infrastructure.

"On April 30, 1993, CERN decided to make the Web workavailable to any users, without charging a licensing fee. Tenyears later, W3C's members have taken the lead and affirmedthat, even after revolutionary changes in the world of informa-tion technology, the path to new markets is more smoothlypaved with open standards," declared Tim Berners-Lee, W3CDirector. "W3C now sets a benchmark for the pragmatic wayto successfully develop royalty-free Web standards in thecurrent patent environment."

W3C's Membership takes Lead in Supporting Royalty-Free Standards for the Web's Second DecadeBased on overwhelming support of the W3C Membership,consensus in the Patent Policy Working Group and supportfrom interested members of the public, Tim Berners Lee hasdetermined that the proposed Royalty-Free Patent Policyshould become the Patent Policy for W3C. "The policy affirmsand strengthens the basic business model that has driven inno-vation on the Web from its inception. The availability of aninteroperable, unencumbered Web infrastructure provides anexpanding foundation for innovative applications, profitablecommerce, and the free flow of information and ideas on acommercial and non-commercial basis," he declared.

Beyond establishing a commitment to royalty-free standards,the policy provides W3C with:• a stable, practical patent policy• a clear licensing framework• consistent disclosure obligations• an exception handling process when problems arise.

After Three Years of Work, Diverse Parties Create a Common PathThe W3C Patent Policy Working Group was launched inOctober 1999, after a patent claim against P3P derailed thedevelopment of that technology. Based on a legal analysis ofthe claim, the threat was removed, and work successfullyresumed. This and other experiences raised awareness ofpatent issues in the W3C membership. The Working Groupwas created in part to make more concrete how W3C WorkingGroups could successfully work on Web standards in theevolving patent climate.

"The Patent Policy represents what may be the most thorougheffort to date in defining a basic patent policy for standard-setting," continued Berners-Lee.

The primary goal of the W3C Patent Policy is to enable W3CRecommendations to be implemented on a royalty-free basis.The policy also requires patent disclosure by W3C memberswhen they are aware of patents that may be essential to theimplementation of W3C Recommendations.

In simple terms, the Patent Policy provides that:• All who participate in the development of a W3C

Recommendation must agree to license essential claims (thatis, patents that block interoperability) on a royalty-free basis.

• Under certain circumstances, Working Group participantsmay exclude specifically identified patent claims from theroyalty-free commitment. These exclusions are requiredshortly after publication of the first public Working Draft,reducing the likelihood that surprise patents will jeopardizecollective Working Group efforts.

• Patent disclosures are required from W3C Members andrequested of anyone else who sees the technical drafts andhas actual knowledge of patents that may be essential.

• Patent claims not available with terms consistent with theW3C Patent Policy will be addressed by a exceptionhandling process.

Links: W3C Patent Policy:http://www.w3.org/Consortium/Patent-Policy-20030520.html

Director's Decision, W3C Patent Policy:http://www.w3.org/2003/05/12-director-patent-decision-public.html

The W3C Semantic TourThe World Wide Web Consortium (W3C) held a series of fiveone-day events across Europe. The tour, supported by theEuropean Commission, through the Question-How project,started on 10 June in Rome, 12 June in London, 17 June inMunich, 19 June in Athens, and ended on 24 June in Brussels.The goal of this W3C Semantic Tour was to promote W3Ctechnologies bringing to the Web the idea of having datadefined and linked in a way that it can be used for more effec-tive discovery, automation, integration, and reuse acrossvarious applications. The events have been organized byERCIM, the European Host of W3C, and five of the EuropeanW3C Offices (Benelux, Germany and Austria, Greece, Italy,and United Kingdom and Ireland).

For the Web to reach its full potential, it must evolve into aSemantic Web, providing a universally accessible platformthat allows data to be shared and processed by automated toolsas well as by people. The Semantic Web is an initiative of theWorld Wide Web Consortium, and has been established toserve a leadership role, in both the design of specificationsand the open, collaborative development of technology.ERCIM News has dedicated a special issue to the SemanticWeb in October 2002 (No. 51).

Link:http://www.w3.org/2003/03/semantic-tour.html

News from W3C

ERCIM News No. 54, July 2003 7

Continued Focus on the Mobile WebW3C continues to focus on the mobile web. Mobile versionsof the vector grahpics format SVG (Basic and Tiny) wereissued as W3C Recommendations early this year. Moreover,the W3C Device Independence Working Group has justreleased the first public Working Draft of Core PresentationCharacteristics. The group is building a basis for adaptingcontent to device presentation capabilities. This draft providesa common set of property or attribute definitions that may bereused in future vocabularies. Finally, the MultimodalInteraction Working Group released an update to the W3CMultimodal Interaction Framework W3C Note. The frame-work identifies the major components for multimodalsystems. The group is writing specifications to extend theWeb user interface to offer input and output choices'anywhere, on any device, anytime.'

Link: http://www.w3.org/2001/di/Activity

Web Services at W3CThe World Wide Web is more and more used for application-to-application communication. The programmatic interfacesmade available are referred to as 'Web services'. W3C's goal isto build an open, extensible technical foundation for the Web,and Web Services is one part of this foundation. W3C has fiveWorking Groups/Task Forces focusing on Web services tech-nologies, and at least another five which have produced thefoundation for all other Web services work. With the recentcompletion of the SOAP 1.2 W3C Recommendation, over 19technical specifications are under development at W3C withinthe Web Services Working Groups and Task Forces.

The first buildingblock for Webservices, impor-tant to all organi-zations interestedin Web Services,is being producedby W3C's XML Protocol Working Group – SOAP Version1.2. SOAP is a central piece of the Web services architecture,enabling complex application-to-application communication.SOAP Version 1.2, advanced to W3C Recommendationstatus on 24 June 2003, is a robust and extensible XML-basedmessaging framework based on SOAP/1.1, with full supportfor W3C Recommendations, support for multiple protocols(HTTP/1.1 as well as others).

The Web Services Activity is composed of three otherWorking Groups, addressing the overall architecture of Webservices, their description and their composition. The work inWeb Services Activity rests in turn on other work done inW3C. The XML Activity is responsible for the XML specifi-cation, XML Schema, XPath, XQuery, and other relevant

specifications. Work in our Technology and Society Domainhas defined specifications for encrypting and digitally signingXML documents or parts of XML documents, which answerimportant security needs in the Web services area. Othergroups at W3C are connected to these groups, either throughrequisite coordination, or through task forces. For example,the W3C Internationalization Web Services Task Force,which looks at internationalization issues and how theyimpact Web services applications and infrastructure. Thesetechnologies serve as the basis for many other specifications,such as the WS-Security Specification first produced by IBMand Microsoft and now being worked on further in OASIS.

Links:Homepage for W3C's WS Activity:http://www.w3.org/2002/ws/

Web Services published Technical Reports:http://www.w3.org/TR/tr-activity.html#WebServicesActivity

SOAP 1.2 FAQhttp://www.w3.org/2003/06/soap12faq.html

From SOAP/1.1 to SOAP Version 1.2 in 9 pointshttp://www.w3.org/2003/06/soap11-soap12.html

The Working Groups and Task Forces at a glance:XML Protocol Working Grouphttp://www.w3.org/2000/xp/Group/All work on SOAP 1.2 happens here.

Web Services Architecture Working Grouphttp://www.w3.org/2002/ws/arch/This group identifies the overall architecture, identifies missing pieces, and determines whether new work is needed or existing work (inside or outside of W3C) meets needs.

Web Services Description Working Group.http://www.w3.org/2002/ws/desc/This group is at work on WSDL 1.2, the Web Services DescriptionLanguage, based on WSDL 1.1..

Web Services Choreography Working Group.http://www.w3.org/2002/ws/chor/This group is developing a language describing the composition ofWeb services.

Internationalization Web Services Task Forcehttp://www.w3.org/International/ws/Internationalization of Web services allows applications to functionacross national, linguistic, or cultural boundaries.

Last W3C Recommendations• XPointer element() Scheme• XPointer Framework• XPointer xmlns() Scheme• SOAP Version 1.2 Part 0: Primer• SOAP Version 1.2 Part 1: Messaging Framework• SOAP Version 1.2 Part 2: Adjuncts• SOAP Version 1.2 Specification Assertions and Test

Collection

An exhaustive list of all W3C Technical Reports:http://www.w3.org/TR/

ERCIM is the European host of W3C.


by John Krogstie

The most recent wave in informationsystems development is the introductionof networked palmtop computers, madepossible by the continued miniaturisa-tion of electronic circuits and the enor-mous resources spent by industrialisedcountries on deploying a ubiquitouslyavailable broadband wireless communi-cation infrastructure.

Because of the mobility of users of thesedevices and the characteristics of wire-less communication, the operating envi-ronment for any accompanying serviceswill be much more dynamic than istypical for traditional distributedsystems. In order to retain system qualityand usability under such circumstances,R&D problems within a number of areasmust be solved.

User-orientation and Personalisation Mobile information systems oftenaddress a wide user group, which meansthat user interfaces should feature promi-nently and early in the design processand often need to be very simple. Theresulting user interfaces cannot assume aprior acquaintance with computers, andinput and output facilities may beseverely restricted (no keyboard, smallscreen-size etc) or based on new modali-ties (speech recognition and synthesisetc). This means that individualisation ofmobile information systems becomesincreasingly important, both at the indi-vidual level, where user-interface detailssuch as commands and screen layout aretailored to personal preferences andhardware, and at the work level, wherefunctions are tailored to fit the user’spreferred work processes. The latter casecan also include the support of novelworkplace designs and methods of work

organisation that enable collaboration ofmulti-location and mobile workers.

Individualisation means informationsystems that are able both to automati-cally adapt themselves to the preferencesof the user and to be explicitly tailoredby users through a specific user inter-face. The main goal is to achieveusability of the applications on allpossible interfaces, based on adaptationto the different physical devices. Thiscalls for intelligent, adaptive and self-configuring services that enable auto-matic context-sensitivity, user profilingand personalisation in a trusted andsecure environment, as well as multi-lingual and multi-cultural presentation,and multiple modes of interaction.

Technological Aspects including Convergence and Multi-Channel SupportMobile devices have severely limitedprocessing, memory and communicationcapacities compared to other kinds ofcomputers. Performance considerationstherefore become increasingly importantin the first design steps. Analyticalpredictive methods are necessary inorder to assess a large number of alterna-tives during the design of mobile infor-mation systems. Mobile informationsystems also pose new challenges toachieving information systems depend-ability. The drive toward mobile deviceshas led to the integration and conver-gence of various technologies into awide range of innovative mobile andmulti-modal applications. Mobile andother new technologies provide manydifferent ways to offer the same orsimilar services to customers. Novelapproaches are therefore needed for the

Applications and Service Platforms for the Mobile User — Introduction

SPECIAL THEME: Applications and Service Platforms for the Mobile User

ERCIM News No. 54 July 2003 9

8 Applications and Service Platformsfor the Mobile User - Introductionby John Krogstie, NTNUInvited article:

10 In Search of Knowledge aboutMobile Usersby Ee-Peng Lim, Yida Wang, Kok-Leong Ong, Nanyang TechnologicalUniversity, Singapore and San-YihHwang, National Sun Yat-SenUniversity, Taiwan

11 The Missing Link – User-Experienceand Incremental RevenueGeneration on the Mobile Internet by Barry Smith, ChangingWorlds,Dublin and University College Dublin

12 User-Centric Peer-to-Peer ServiceEnvironment for Interaction withNetworked Appliancesby Daniel Pakkala, Pekka Välitalo,Pekka Pääkkönen and JuhaniLatvakoski, VTT

14 Kontti — Context-aware MobilePortalby Juha Kolari, VTT

15 Contextual Mediation enablesAppropriate Data Selectionby Dan Chalmers, Morris Sloman andNaranker Dulay, Imperial CollegeLondon, UK

16 mBlog: a Mobile Information Servicefor Allby Emmanuel Frécon, SICS

18 A Mobile Schema Browser forIntegration Specialistsby Gerald O’Connor and MarkRoantree, Dublin City University,Ireland

19 A Peer-to-Peer Middleware forMobile TeamWorkby Carlo Ghezzi, Gianpaolo Cugola,and Gian Pietro Picco, Politecnico diMilano, Italy

21 Fluid Computingby Marcel Graf, IBM Zurich ResearchLab, Switzerland/SARIT

23 Designing Usable Multi-PlatformInteractive Applicationsby Fabio Paternò and Carmen Santoro,ISTI-CNR

24 Interoperating with HeterogeneousMobile Servicesby Paul Grace and Gordon S. Blair,Lancaster University, UK

26 NAC: An Architecture for MultimediaContent Adaptation for MobileDevicesby Nabil Layaïda and TayebLemlouma, INRIA

27 Integrating GSM Networks andInternet: New Unstructured Servicesby Rita Arrighi, Maurizio A. Bonuccelli,Francesca Lonetti and FrancescaMartelli, ISTI-CNR

28 From Ad Hoc Networks to Ad HocApplicationsby Benoît Garbinato and Philippe Rupp,University of Lausanne, Switzerland

30 Light Support for Dynamic andPervasive Services on P2P Networksby Stéphane Frénot, Anis Krichen andStéphane Ubéda, INRIA and INSA-Lyon Scientific and Technical University

31 Towards a Virtual HomeEnvironment Application Serverby Ali Hafezi and Klaus-Peter Eckert,Fraunhofer FOKUS

32 M-commerce API Package for MobilePhonesby Ioannis G. Askoxylakis, DiomedesD. Kastanis and Apostolos P.Traganitis, ICS-FORTH

34 Wireless Ad Hoc Network for Dublin(WAND): A Large-Scale Ad HocNetwork Test-Bedby Stefan Weber, Vinny Cahill, SiobhanClarke and Mads Haahr, Trinity CollegeDublin

35 Lightning Active Node Engine — AnActive Network User ServicePlatformby Sami Lehtonen, VTT

36 Middleware Support for UbiquitousComputingby Luc Hogie and Christian Hutter,Luxembourg University of AppliedSciences

38 Sentient Objects: TowardsMiddleware for Mobile Context-Aware Applicationsby Gregory Biegel and Vinny Cahill,Trinity College Dublin

39 Database Components on Chipby Nicolas Anciaux, Luc Bouganim andPhilippe Pucheral, INRIA

41 Object-driven Application Design:from Mobility towardsUbiquitousnessby Davide Carboni, Gavino Paddeuand Stefano Sanna, CRS4, Italy

42 Development Tools for DynamicEnvironment-Adapted MobileInterfacesby Anthony Savidis and ConstantineStephanidis, ICS-FORTH

44 Introducing New Mobile ServicesFasterby Timber Haaker and Oscar Rietkerk,TNO Telecom, The Netherlands

46 Uluru: Mobile Interactive MultimediaExperimental Service Environmentby Johan de Heer, Andrew Tokmakoff,Henk Eertink and John Anijs,Telematica Instituut, The Netherlands

48 Next Generation Public TravelInformation Systemsby Karl Rehrl, Harald Rieser andSiegfried Reich, Salzburg ResearchForschungsgesellschaft/AARIT

49 Mobile Communication inNewspaper Distribution andTransportby Olli Kuusisto and Janne Pajukanta,VTTInvited article:

50 Values of Mobile Applications toEnd-Usersby University of Nebraska-Lincoln, USA

ARTICLES IN THIS SECTION

development and evolution of applica-tions on and across different mobile andtraditional platforms.

Methodology for Development to ensure Organisational ReturnMobile information systems are radical,and will therefore reward an increasedfocus on idea generation early in thedesign process. Understanding andmodeling mobile users' requirements fornew services is consequently of greatimportance. One needs both to be able todevelop these systems and to address themajor hurdles for the deployment ofapplications and services for the mobileuser. Another effect of the introductionof mobile information systems is thespawning of further initiatives forchanging other information systems.Overall, it is important to focus on theinteroperability of services and theability to roam across heterogeneousnetworks and service environmentsincluding, for example, working, billing,payment, ticketing and accountingservices, as well as having seamlessaccess to corporate and governmentresources.

This special issue highlights some of theexisting European R&D in Applicationsand Service Platforms for the MobileUser. Although far from being acomplete collection, it gives a goodimpression of where the Europeanresearch community is putting its efforttoday, and where one can expect, or atleast hope for, more results in the future.

Please contact:John Krogstie, NTNU and SINTEFTel +47 22067425E-mail: [email protected]://www.idi.ntnu.no/~krogstie



Mobile phones and other mobile devicesare fast becoming indispensible in ourmodern society. According to a recentsurvey by Frank N. Magid Associatesand Upoc.com, 59 percent of Americansage 12 and over (about 140 millions ofthem) own mobile phones, and thatalmost a quarter of non-owners plan tobuy a mobile phone in the near future.The sales of mobile phones worldwidewas 385 millions in 2001 and it has beenpredicted to reach 675 million in 2006.In tandem to this growth trend, we alsowitness the emergence of many newapplications and businesses that exploitmobile phone technologies in differentways. Before other wearable computinggadgets become more feasible andpopular, mobile phones are likely toremain as the more dominant wearabledevices in the coming years.

Mobile phones, unlike computersconnected to wired networks, are highlypersonalizable. While it is common for auser to own a few mobile phones, it isvery unlikely for different users to sharea mobile phone. Also unlike otherpersonalized accessories such aswatches, walkmans, etc., many of themobile phones are trackable. They aretrackable because they have to maintainregular contacts with the mobiletelecommunication networks in order toreceive and make calls. With these track-ability and personalization features, onecan conceive many unique and inter-esting mobile applications for end users.

At the Center for Advanced InformationSystems, we conduct extensive researchon using knowledge discovery tech-niques such as frequent pattern mining,classification, and clustering to createnew mobile applications or to enhanceexisting mobile applications. Examplesof such mobile applications include e-commerce systems, databases, emailsystems, search engines and web

browsers. We investigate both the func-tionalities and operational efficiencies ofmobile applications, and determine thekinds of knowledge required forenhancing them. We also develop algo-rithms for discovering knowledge frommobile user data and evaluate theirperformance. We further study how thediscovered knowledge and mining algo-rithms can be integrated with the opera-tional systems turning the knowledgeinto actions.

In the e-commerce application domain,we envisage the importance of usingmobile user movement data to deriveknowledge about mobile users. As users’purchase behaviors are often highly corre-lated to their group affiliations, knowingthe latter well will allow e-commerce

vendors to develop group-specific pricingmodels and marketing strategies to bettermeet the buying needs of the user groups.While there have been several existingmethods developed to address theproblem of discovering user groups, theyare usually based on transaction historiesand user profiles, and the accuracies maynot be satisfactory.

In our research, we introduce a newapproach to mine user groups, known asgroup pattern mining. In group patternmining, we determine the customergrouping information based on thespatio-temporal distances among thecustomers. We assume that the usermovement data are first collected bylogging location data emitted from themobile phones and similar devices.Mobile users that are always close to oneanother for significant amount of time aremodeled as a group pattern representedby 4-tuple <G, max_dist, min_dur>where G denotes the set of mobile users,max_distdenotes the maximum distancebetween any pair of users for them to beconsidered close, and min_durdenotesthe minimum duration in which the usersin G are close to one another. Forexample, for {Tom, Mary} to be a validpattern, Tom and Mary must be less thanmax_dist apart for at least min_durcontinuous time period.

Compared to traditional frequent patternmining, group pattern mining incursmuch more computation and storage. Todiscover group patterns from a set ofmobile users, we have proposed a fewefficient algorithms that are based onfrequent pattern mining. As part of thisresearch, we also investigate new datasummarization techniques and data struc-tures to reduce the computational andstorage overheads. In brief, our researchframework for mobile data mining isdivided into three core areas, which we

At the Center for Advanced Information Systems at Nanyang TechnologicalUniversity, Singapore, research is carried out on using knowledge discoverytechniques such as frequent pattern mining, classification, and clustering tocreate new mobile applications or to enhance existing mobile applications.

In Search of Knowledge about Mobile Usersby Ee-Peng Lim, Yida Wang, Kok-Leong Ong and San-Yih Hwang

Further Reading• Upoc.com. “More Mobile Owners

Turning to Text Messaging”,http://www.upoc.com/corp/news/news-emarketer.html, Feb. 2003.

• Reed Electronics Research, “RER-The Mobile Phone Industry - AStrategic Overview,” Oct. 2002.

• U. Varshney, R. Vetter and R.Kalakota, “Mobile commerce: A NewFrontier,” IEEE Computer: SpecialIssue on E-commerce, Oct. 2000.

• Ee-Peng Lim, Keng Siau, Advancesin Mobile Commerce Technologies,Idea Group Publishing, Jan. 2003.

• MIT Media Lab. The Context AwareCell Phone Project. (under theMIThril wearable computinginitiative).http://www.media.mit.edu/wearables/mithril/phone.html

• Yida Wang, Ee-Peng Lim, San-YihHwang, “On Mining Group Patternsof Mobile Users,” 14th InternationalConference on Database and ExpertSystems Applications (DEXA2003),Prague, Czech Republic, September,2003.



shall highlight some of the issues thatshould be further investigated. They are:• Infrastructure for mobile data mining –

we explore the design issues involvinga warehouse for mobile data. This mayinclude the algorithms for efficientaggregation and transformation ofmobile data. The challenge in this caseis to deal with the heterogeneous dataformat from different mobile devicesat different locations with differentbandwidth and computing resources.

• Algorithms for mobile data mining –once data is available, the next chal-lenge is to make sense of the data.Hence, algorithms are needed to findknowledge to improve the efficiencyof mobile applications/queries, and to

enhance the user experiences of thephone.

• Incorporating mobile mining resultsinto operational systems – knowledgeobtained from data mining must be inte-grated with the operational systems. Thechallenge is to develop algorithms thatevaluate which are the 'actionable' datamining results, and then apply them in atimely fashion to ensure the effective-ness of the mobile data mining system.

Mobile data mining, as a very new areaof research, has created a wide range ofopportunities for researchers, engineersand developers to create new interestingapplications for both the end users andbusinesses. We believe that our group

pattern mining work is only one of thesemany examples. As many more know-ledge about mobile users can be mined inthe near future, we therefore expect theupcoming applications and systems to beable to adapt more seamlessly into ourdaily lives.

Please contact: Ee-Peng Lim, Yida Wang and Kok-Leong OngNanyang Technological University,SingaporeE-mail: [email protected],[email protected],[email protected]

San-Yih HwangNational Sun Yat-Sen University, Taiwan E-mail: [email protected]

Paradoxically, the mobile Internet repre-sents both a dramatic step forward and asignificant step backward from an infor-mation access standpoint. While it offersusers greater access to valuable informa-tion and services 'on the move', mobilehandsets are hardly the ideal accessdevice in terms of their screen-size andinput capabilities. As a result, MobileInternet users are often frustrated by howdifficult it is to quickly access the rightinformation at the right time. In ourresearch we have pioneered the use ofartificial intelligence and personalizationtechniques to provide practical solutionsto these access problems and our tech-nology is now deployed with leadingEuropean operators. Specifically, theClixSmart NavigatorTM product-suite,developed by ChangingWorlds Ltd.(www.changingworlds.com), providesmobile operators with a complete multi-access, portal platform that includes

patented intelligent navigation servicesthat automatically learn about user prefer-ences in order to automatically adapt thestructure of a portal for the individualuser. ClixSmart Navigator can reducecontent access times by 50%, signifi-cantly improving usability, and so leadingto direct increases in mobile usage.

The Mobile Navigation ProblemThe core usability problem with MobileInternet services (such as WAP portals) isthat users spend a significant time navi-gating to content through a series ofmenus (see Figure). Recent research high-lights the scale of this problem and themismatch between user expectations andrealities. One study claims that while theaverage user expects to be able to accesscontent within 30 seconds, the reality iscloser to 150 seconds. WAP navigationeffort can be usefully modelled as click-distance – the number of menu selections

and scrolls needed to locate a contentitem. A recent analysis of 20 Europeanmobile portals reports an average click-distance in excess of 16. In other words, atypical European mobile user can expectto have to make 16 clicks (scrolls andselects) to navigate from their portal homepage to a typical content target. Moreover,on average, European portals are organ-ised such that less than 30% of contentsites are within 10-12 clicks of the portalhome page; 10-12 clicks corresponds to anavigation time of about 30 seconds,which is the tolerable limit for most users.In other words, more than 70% of mobileportal content is essentially invisible tousers because of its positioning within itsparent portal.

The Personalization SolutionLarge click-distances are a fundamentalfeature of a 'one size fits all' approach toportal design and even optimising a

by Barry Smyth

Intelligent navigation and personalization technology developed byChangingWorlds and University College Dublin’s SmartMedia Institute has provento significantly enhance the mobile user experience, drive usage and take-up,and so increase data-services revenues for mobile operators.

The Missing Link – User-Experience and Incremental Revenue Generation on the Mobile Internet



menu structure for the needs of someimaginary 'average user' is unlikely tobenefit individuals. However, instead ofpresenting the same portal to each andevery user it is possible to use personal-ization techniques to learn about thepreferences of individual users and tostrategically adapt the structure of theportal on a user-by-user basis.

ClixSmart Navigator adopts a collabora-tive, probabilistic approach to personal-ization. Individual user accesses aretracked and used to generate comprehen-sive user profiles that reflect the user’sportal access patterns (the services theyaccess, when, and how often), theiraccess-device characteristics, andvarious temporal data. These profiles areused to develop a probabilistic accessmodel for each user that is capable ofestimating the probability that a user willaccess an option, o (which could beanother portal menu or content page),given that they are currently viewingportal menu, m. This in turn allowsClixSmart Navigator to personalizestatic portal menus by reordering theirdefault links based on the target user’saccess probabilities and by promotingnew links that are likely destinationtargets for this user.

Over time the portal structure graduallyadapts to the access patterns of an indi-vidual user. For example, a regularmovie goer may find that, at theweekend, their local cinema has beenpromoted to their portal home page. Ofcourse, it is vitally important that anyrestructuring takes place in a fashion thatmakes sense to the user. Rapid changesin portal structure can confuse the userrather than assist them, and as suchClixSmart Navigator contains a range offeatures to control the speed and scope of

perosnalization in order to maximise thebenefits to the end-user.

The BenefitsClixSmart Navigator is currentlydeployed on the Vodafone (Ireland) andO2 (Germany) networks and is underevaluation by a range of Europeanmobile operators. To date the benefits tothe end users, mobile operators, andcontent providers are clear andcompelling. For example, extensiveevaluations have proven that ClixSmartNavigator’s intelligent navigation canreduce navigation effort by up to 50%within 4 weeks. Moreover, for every onesecond of navigation time that is saved,the average user engages in additional 3seconds of content time. This translatesinto significant increases in overallairtime (20%+), page impressions(30%+) and data download volumes(30%+) when compared to control-group usage.

Link:http://www.changingworlds.com/

Please contact:Barry SmythChangingWorlds, Dublin and University College DublinE-mail: [email protected]

The mobile Internet is now being realisedand launched on the market. It enablesdifferent kinds of mobile terminal to beconnected to the Internet anywhere and atanytime. However, due to the massiveamount of information content and thenumber of devices connected to theInternet, users’ lives are becoming moreand more complicated. Consequently, inthe future, user centricity requirements willbecome increasingly important, makingthis an essential research challenge.

User centricity and the flexibility to dealwith multiple use scenarios have beenstarting points for the development of theservice platform. We have chosen thefuture residential home environment forthe validation of solutions because it is agood example of a future intelligentenvironment using many networkedappliances with different capabilities. Ina residential home environment, somenetworked appliances can be very simpleand it would not be cost effective to

implement the control channels of thesedevices based on Internet Protocol (IP).Instead, these simple appliances - likeX.10 light switches, for example - can becontrolled via a service gateway from theIP networks. For devices like alarmsystems or remotely controllable deviceswith more processing capabilities andfunctionalities, it is reasonable to imple-ment the control channels based on IP.This variety of appliances creates theneed for flexibility and transport layer

by Daniel Pakkala, Pekka Välitalo, Pekka Pääkkönen and Juhani Latvakoski

The focus of research at VTT has been to design a service platform for the futurewireless world. Special attention has been paid to user centricity in the designof future service platforms.

User-Centric Peer-to-Peer Service Environmentfor Interaction with Networked Appliances

A sample navigation sequence on a WAP

portal: accessing the Ormonde Cinema

site takes 16 clicks (scrolls and selects)

from the portal home page.



protocol independence for the serviceplatform.

The problem to be solved by usercentricity in service platform design canbe compared to the situation some of usalready have in our living rooms. Forexample, owning a TV, VCR, hi-fisystem and DVD means there are at leastfour different remote controls to manage.Even with this relatively small numberof devices, the separate remote controlspresent an annoying configuration fromthe user perspective. In future homesthere could be many more appliancesthat need to be controlled and managedby the user. The purpose of usercentricity in the service platform is to

make the control and management offuture networked appliances (NAs) aseasy as possible for the end user. Forexample, a single button press on amobile terminal (MT) can display a floorplan of the home that shows all the appli-ances that the user is able to control.When a user selects, for example, aliving room lamp, the user interface (UI)of the living room lamp appears on thescreen and the user can then remotelycontrol it.

In Figure 1, two different approaches forcontrolling networked appliances arepresented. On the left is the legacy inde-pendent UI approach, and on the right,the applied user centric peer-to-peer

approach that we have used in the devel-oped service platform. In the legacyapproach, the user interfaces of the IP-capable appliances are located at theMT, whilst those of the non IP-capableappliances are located at the servicegateway. This kind of approach tocontrolling system design will lead to anon user-friendly situation in the future,when the number of NAs will grow fast.In the applied user centric peer-to-peerapproach, all the devices should berepresented in the service gateway bytheir agents or control software, and theUI components of all the NAs present inthe system should be available at theservice gateway. The final UI that isuploaded to the user’s mobile terminal isgenerated from these UI components.The UI components contain the addressinformation of the NAs providing theservice in question. After the final UI isactivated at the mobile terminal, it auto-matically establishes the control connec-tions needed and becomes visible to theuser.

In our latest demonstration we intro-duced the concept of the User CentricPeer-to-Peer Service Environment (seeFigure 2). In the demonstration wecontrolled two X.10 devices and aremotely controllable car with a videocamera through the service platform.The X.10 devices represented simplenon IP-capable devices and the car repre-sented a more complicated device withIP capabilities. The configuration of thedemonstration, final data channels anduse cases are presented in Figure 2.

The work described in this article hasbeen carried out in a strategic projectOmniPresent rEference seRviceArchitecture - OPERA at VTTElectronics during year 2002.

Link:http://opera.vtt.fi/

Please contact:Daniel Pakkala, VTT ElectronicsE-mail: [email protected]

Juhani Latvakoski, VTT ElectronicsE-mail: [email protected]

Pekka Välitalo, VTT ElectronicsE-mail: [email protected]

Pekka Pääkkönen, VTT ElectronicsE-mail: [email protected]

Figure 1: Approaches for controlling networked appliances.

Figure 2: Demonstration system.



In order to study the technology andconcept of context-awareness, VTT hasdeveloped a context-aware mobileportal. The contents of the service areadapted to the user’s current context anddevice. The contexts can be identifiedautomatically according to time or loca-tion. They can also be manually definedand activated by the user. The contentsof the portal can be existing mobileservices and Web pages as well aspersonal notes and files.

There will be two field trials to evaluatethe prototype and the concept. Over 70users will be involved as the prototype istested in the field. Conceptually, theevaluations approach the subject fromthree different angles, namely, everydaycontexts, context-aware services andsocial surroundings.

Everyday ContextsEveryday use plays a visible role in theevaluation. The goal is a natural integra-tion of the service with what the usersactually do. As 'context' itself is a highlypersonal concept, so context-awareservices need to be flexible enough toconform to the varied situations in users’lives. The system allows users to createtheir own personal contexts, such as 'athome', 'at work', 'out partying' and'feeling blue'. The users can link personalnotes and existing services to thecontexts relevant to them. They can alsoinform others of their current personalcontext. The challenge is in achievingthis high personal conformity withoutrequiring too much effort from the users.

Context-Aware ServicesReady-made context-aware servicesmay not adapt to the user’s everyday lifebut can provide content with little needfor further personalisation. The portalallows the delivery of a context-awareservice directly to the user. Visitors to a

festival could, for instance, subscribe tothe festival schedule. The schedule andupcoming events can be presentedaccording to time and the user’s location.We will study the service from this angleat the week-long Tampere TheatreFestival in August 2003. Another casewill be a location-based historical tour inthe city of Tampere, which will be evalu-ated in the summer of 2003.

Social SurroundingsOne motivation for creating and main-taining everyday contexts is to conveythe user’s own status to others; egwhether you are available for contact orwhat your plans for the evening are. Theuser can specify which other users areallowed to view his/her current status.Even then, the user can define how thestatus will appear to others, and the loca-tion or context itself is not necessarilyconveyed in this description.

User-generated material naturally meansthe inclusion of context-based messages.The system allows users to send messageswhich are delivered only when the chosentarget context is active, ie when the userarrives at work or home. Targeting the

message allows, for instance, a work-related message to be sent to work on aweekend. The message will be deliveredonce the recipient enters the workplace.

Appropriate technological matches havebeen chosen for these concepts. Use ineveryday life requires a mobile devicethat is available all the time and is notlimited to a certain area. Consequently,the prototype has been designed to workin WAP-enabled mobile phones. It isoptimised for use in phones with colourdisplays and multimedia capabilities.Both WML and XHTML versions areavailable. The system is open withregard to positioning methods, and thecurrent system supports both cell-basedand proximity-based positioning. A trialon a WLAN platform will complementthe evaluation both conceptually andtechnologically. The WLAN trial will beused to study the promising concept ofcontext-aware services more thoroughly.

The development and studies are beingdone under the project Kontti (Context-Aware Services for Mobile Users). Thetwo-year project began in 2002 and willrun through 2003. The project is fundedin part by Tekes and its NETSprogramme. Other collaborative partnersin the project are Nokia, Radiolinja andTeamware. In addition, the FinnishBroadcasting Company (YLE) providescontent in the trials.

Links:http://www.vtt.fi/tte/projects/kontti/

Please contact:Juha Kolari, VTT (user studies)Tel:+358 3 316 3311E-mail: [email protected]

Timo Laakko, VTT (project manager)E-mail [email protected]: +358 9 456 4505

For many, a mobile device is a constant companion during daily routines. Informationand service needs vary according to the user's immediate situation. A context-aware service responds to this by offering contextually relevant information. Byidentifying these contexts and the varying information needs users have in them,mobile services can be made more topical, personal and easily available.

Kontti — Context-aware Mobile Portalby Juha Kolari

Context-awareness can bring about a

more personal and spontaneous way of

using mobile services.


15

Context aware computing refers to thetreatment of the user’s environment insystem behaviour. Aspects of context arewide-ranging: location, speed of travel,noise levels, device in use, task engagedin etc, and mobile computing exposesusers to the effects of these. Context hasmany uses in mobile computingincluding: description of context to theuser, eg location notification; context-sensitive resource use, eg location awareprinting; contextual annotation of data,eg noting other people present as metadata to meeting minutes. At ImperialCollege we have been developing tech-niques to specify contextual mediation.This is the use of context to modify thedata selected amongst the data offered,in order to make the best presentation ofinformation to the user given for contextof use. Techniques include: scalinggraphics and summarising text to fitdocuments within a screen; limiting thequality of images to reduce data sizewhen using a slow network; including orhighlighting data describing requiredfacilities such as induction loops for thedeaf; omitting attractions which are

closed when presenting a map of atourist site. Some of these techniqueshave been deployed in isolation, oftenhard-coded into applications andtargeted towards a few specific devices –the effect often being unsubtle andfailing to capture all the user’s needs.Our research provides a general way toassociate a description of needs and limi-tations with the context of use – and usethis to select the best available combina-tion of data.

Our work initially focused on maps. Mapdata may be large in size, causing longdelays on slow networks and it mayinclude a level of detail, which cannot beinterpreted on low-resolution displays.Vector map data lends itself to partialpresentation and processing of differentelements. We were able to use this toallow the prioritisation of data accordingto its semantic types, eg roads, rivers,buildings, distance from location ofinterest; and its properties, eg differentscale representations of the same feature.In this way mode of transport affects thepriority of representations of roads, rail-

ways, footpaths etc. Speed affects inclu-sion of distant or small features. Task(work, leisure etc.) affects inclusion oftourist symbols. We also describe limitsover managed resources in order toconstrain the selection. Screen resolutionand available bandwidth affect thecombination of data selected, in order tomeet specified goals: time to downloadmay depend on speed, drawingcomplexity may depend on vibration.Note that the selection of data to displayconsiders the combined effect of all thedata in the selection, rather thanprocessing data on an element-by-element basis.

Contextual mediation is illustrated in theFigure. The left-hand map shows thecomplexity resulting from no mediation.The mediated maps are for navigating ina car at 30 to 40mph, hence the concen-tration on major roads and the omissionof minor roads away from the immediatevicinity (the centre). The data wasloaded within a 25s deadline over asimulated GPRS network. The differ-ence between the tourist and worker’s

by Dan Chalmers, Morris Sloman and Naranker Dulay

Mobility produces a wide range of context, which has a significant impact onusers’ experience of computing. Some solutions exist for issues such as smallscreens and low bandwidth, although these generally offer rather indiscriminatemodification of data and support only a small range of devices. A more sensitiveapproach has been investigated at Imperial College London: causing a selectionof data in response to a wide range of contexts.

Contextual Mediation enables Appropriate Data Selection

Figure (left to right):

Unmediated map, all features included.

Mediated for tourist in a car, less peripheral detail.

Mediated for worker in a car, note omission of tourist sites, golf courses, footpaths etc.

ERCIM News No. 54 July 2003



maps is subtle, but helps the worker toscan the map without distraction. Wherefeatures have hyperlinks from them, egdescribing tourist attractions or access todelivery points, the ease of selectinglinks also improves. In tests we found theresulting maps to be more predictable intheir download time than unmediatedmaps. Predictable delays have beenshown to be a key factor in user satisfac-tion in the web. The omission of unnec-essary detail was shown to provideimproved clarity and faster navigation inuser tests.

Our approach is applicable to othersemantically rich data intensive applica-tions. We are now investigating applyingthis approach to collaborative tools suchas message services, shared whiteboards,engineering and maintenance plans andweb based documents. In collaborationwith the University of Southampton wehave been investigating the mediation ofhyperlinks and the use of mediation inpervasive information systems. There isalso related work starting on UbiquitousComputing for Healthcare in theCommunity, which will be focusing on

monitoring of patients with clinicalconditions as they go about their normalactivity.

Link:http://www-dse.doc.ic.ac.uk/Research/ubicomp.html

Please contact:Morris Sloman,Imperial College London, UKE-mail: [email protected]

Weblogs, or blogs for short, are the latestInternet phenomenon. Having started asan underground culture in the late 90s,the movement has now reached such anextent that these Web sites have becomean important medium for the generalspread of opinion and information on theInternet. Thousands of new blogs arecreated every day by ordinary people orgroups, and cover themes such aspersonal diaries, political or professionalcommentaries, fanzines around cultureand sports, group diaries around fishing,excursions, etc.

Technically speaking, blogs are an easyway to produce a Web page by filling outa form. The text entered is displayed on apage organised in reverse chronologicalorder: the most recent entry is at the topof the page and older entries follow.Blogs can provide an archive mechanismfor all entries, and a search feature tolook for specific information withinolder content. Typically, a blog will havea small number of authors and anynumber of readers. A commentingsystem will allow readers to react to theopinions expressed by the authors on theWeb page.

The simplicity of this formula makesblogs available to the mass of Internet

users, relieving them from thecomplexity of HTML editing. Apartfrom this simplicity, blogs are interestingboth for their sociological aspect and fortheir content or topic area. From thelatter point of view, blogs have two mainpurposes: they help maintain a commu-nity around any given topic (from a classtrip to war in Iraq), and they allow one tomake oneself publicly heard.

mBlog, developed by SICS in collabora-tion with Ericsson, takes blogs a stepfurther through the addition of a mobilitycomponent. With mBlog, people canmaintain any number of blogs simply bysending images and texts by MMS,SMS, e-mail or by filling out a form in aWeb browser. Furthermore, each pieceof information can be geographicallypositioned.

Mobility: the Future of BlogsThe very nature of blogs and their usagelead us to consider them as ‘real-time’entities. Indeed, for a blogger, no matterwhat the topic of the blog, it is necessarythat information be published as soon aspossible. For example, if a blog was tosupport a local football team, being ableto blog the main events of a match in realtime is a way to keep supporters whocould not join the game aware of its

progress, as well as an effective methodof keeping a record of the match.Consequently, the immediacy of infor-mation production and consumptionnaturally lends itself to extending theconcept of blogs towards mobile envi-ronments.

The idea behind mBlog is the ability topost from a mobile phone (MMS, SMS,WAP input, e-mail), and the ability toread entries at any time using a numberof mobile outputs (WAP browser,speech synthesis, etc). mBlog also intro-duces a new dimension to weblogs,namely space. Each blog can have ageographical dimension, with each entrybeing positioned within this space. Thisoffers novel ways to access the informa-tion.

Business and Sociological ResultsThe genesis of mobile blogs will supportthe building of communities and socialnetworks of many kinds, both at theprofessional and personal levels. Forexample, we foresee professional usagewithin mobile fleets of salesmen.

Being ubiquitous by nature, mobileblogs will generate traffic within severalparallel distribution channels (GPRS,3G, Internet), and will therefore benefit

Easy-to-use Web sites, called ‘weblogs’ or ‘blogs’ are the latest massive Internetphenomenon. mBlog, developed by SICS in collaboration with Ericsson, takesblogs a step further by adding a mobility component.

mBlog: a Mobile Information Service for Allby Emmanuel Frécon



the telecom industry. In particular,mobile blogs will drive MMS, SMS,WAP and data traffic, as well asincreasing the sales and usage of multi-media phones. Through the provision ofa service that is anchored in our sociallives and behaviours, mobile blogs couldact as one of the kick-off applications forMMS and 3G services.

How does it work? Users send SMS, MMS or e-mailmessages to a dedicated number or e-mail address. mBlog then creates a newblog entry for each message. This entryis placed at the top of the Web pagewhile previous entries are moved down.Upon posting, entries are time-stampedand automatically given a geographicalposition, ie that of the device. This posi-tion can later be changed if necessary.

Alternatively, users can direct theirbrowser to a specific Web page and fillin a simple form. Upon completion, anentry is created as above. The resultingblog can be viewed from any Webbrowser or WAP phone. A geographicalmap shows the location of the entries. Itis useful to note that new multimediaphones now include fully-fledged Webbrowsers.

Technology behind the SceneBlog systems are based on XML tech-nology, which makes their content easyto manage, arrange and exchange.Lately, a number of standard APIs haveemerged, meaning that blog systemshave now become fully-fledged Webservices. mBlog makes extensive use ofstandard Internet protocols and is imple-mented in a scalable and distributedmanner:

Each mBlog service can be run sepa-rately on any machine with Internetaccess. Upon reception, the content ofMMS and SMS is analysed and routed tothe appropriate blog through a communi-cation standard called XMLRPC.

WAP output can be applied to anyexisting blog, and is triggered throughcontinuous surveillance of any Webpage (in this case a blog). Upon modifi-cation, blog content is fetched throughXMLRPC, converted to WML andposted back through FTP.

Entries are automatically geographicallypositioned using a positioning servermade at SICS. Currently, we onlysupport accurate WiFi positioning, butwe are hoping to support GPS and GSMin the near future. The geographical mapof the blog is continuously updated asnew entries are posted.

Future WorkmBlog provides a simple means forreading and writing blogs on the run.Compared to other 'moblogging'attempts, mBlog takes a unifiedapproach through the support of mostexisting mobile devices. Additionally,mBlog adds the dimension of space toblogs, to enhance the association ofinformation with places.

We are currently looking at extendingthe system by building it on top of a scal-able P2P network infrastructure.Information can be replicated along thisnetwork as needed. This has two mainadvantages: it guarantees its very exis-tence against external ‘attackers’ of allsorts, and access to consumers iswidened.

Link:http://ice.sics.se/projects/mblog.html

Please contact:Emmanuel Frécon, SICSTel +46 8 633 15 34E-mail: [email protected]://www.sics.se/~emmanuel/index.shtml

mBlog takes blogs a step further through the addition of a mobility component.



Integration Engineers are often facedwith a requirement to display andanalyse the complex schemas of infor-mation systems to be merged. As thesesystems can be dispersed over a widegeographic area, a Portable DigitalAssistant (PDA) provides a flexiblemeans of viewing and displaying schemainformation. However the browsingprocess, which is often complex andproblematic on a workstation screen,becomes more difficult on the smallerPDA. Using our metadata middleware, amobile user can query metadata on anobject-relational database and automati-cally display its structure. To demon-strate this, an application was developedwhich exploits our interface to theextended object Object-Relational (O-R)schema repository, to manipulatecomplex metadata information.

The O-R model is a complex structurecombining features of relational andobject models. The relational modelconsists of tables, triggers, constraints,views and procedures, while the objectmodel consists of types, associations,aggregations, encapsulation, inheritanceand other complex structures. Thiscombination of features provides apowerful environment for representingdata, yet if this data is not fully specifiedand manageable through a metadatainterface, its reuse and sharing becomesa time consuming and expensive task.The Efficient Global Transactions forVideo media research project (EGTV)has defined a complete metadata inter-face to O-R metadata and developed atool to demonstrate and evaluate itseffectiveness.

Mobile LayerThe Figure illustrates the deploymentarchitecture for O-R metadata access. Inthe Mobile Layer, a PDA uses the meta-

data middleware in a specific applica-tion. The Schema Browser queries theschema using the Extended-SQL meta-data query language, developed by theInteroperable Systems Group (ISG).This includes metadata query options forany O-R schema, role views and multi-media objects. The extended-SQLqueries are wrapped in XML to provide arobust, non-proprietary, persistent andverifiable file format for the storage andtransmission of data. Using the XMLversion of the language, the SchemaBrowser automatically queries anddisplays the complex metadata requiredto display the O-R schema model.

XML Translation LayerThe XML Translation Layer resides atthe database server. Its purpose is toprovide a standard interface to theExtended-SQL query language. Whenreceiving a query it is unwrapped to forman Extended-SQL query, which is subse-quently passed to the SQL translationlayer. After execution, results received

from the SQL Translation Layer areXML-wrapped using a basic rule set, andthen returned to the application.

SQL Translation LayerThe SQL Translation Layer is wheremost of the metadata processing takesplace. Current approaches to interfacingmetadata for O-R databases were exam-ined before this layer was specified. Itwas found that relationships betweentypes very often complex and difficult toretrieve using SQL:99. In fact, SQL:99 isnot powerful enough to extract somemetadata (nested table associations) andprogramming language extensions werespecified to overcome this. The SQLTranslation Layer accepts an Extended-SQL query, which is parsed to invoke asequence of actions against the schemarepository. The results may becomprised of conventional O-R meta-data, role metadata and/or multimediametadata. After the results are restruc-tured to an O-R format, they are passedback to the XML Translation Layer.

by Gerald O’Connor and Mark Roantree

Middleware developed at Dublin City University allows a mobile user to querymetadata on an object-relational database and automatically display its structure.This is particularly useful for example for displaying and browsing complexschema information on a Portable Digital Assistant (PDA).

A Mobile Schema Browser for Integration Specialists

Mobile Object-Relational Schema Browser.



Database LayerThe O-R schema repository wasextended by other EGTV researchers toprovide new interfaces to role and multi-media metadata. Role metadata wasadded because it adds to the expressive-ness of O-R databases. Roles providetemporal aspects to entities, a feature thatis missing in conventional models.Without roles, a new object must becreated each time the structure of anobject evolves and many complex issuesare involved in maintaining such anoperation. The current popular O-Rrepositories fail to manage multimediadata adequately, as it is often represented

as a BLOB with no further metadata toexplore its application or use. This wasaddressed with the metamodel exten-sions. All of this metadata can beaccessed using this framework and theExtended-SQL:99 language. Thus, itbecomes accessible to integration engi-neers using mobile devices.

ConclusionsThe mobile display tool can automati-cally display schema, view, role, andmultimedia information in just a fewselection buttons. This provides themobile integration engineer with a fastcompletion of the first phase in the inte-

gration process. It also illustrates that themobile device can offer a viable platformfor some CASE tools. Current researchis focused on providing update facilitiesthrough the mobile device, in order tomake structural (metadata) changes tounderlying O-R schemas, and to definenew role views.

Link:http://www.computing.dcu.ie/~isg/

Please contact:Gerald O’Connor and Mark RoantreeDublin City University, IrelandE-mail: {goconnor, mark}@Computing.DCU.ie

Collaborative work is intrinsically peer-to-peer in nature. Members of a teamtypically interact directly with eachother, each responsible for a given set ofdocuments and carrying with them thesubset relevant for discussion. On theother hand, most of the currently avail-able tools supporting collaborationexploit a rigid client-server architecture.

This results in an 'architecturalmismatch' between the external viewprovided by the application and itsinternal software architecture. The effectof this mismatch is a lack of flexibility incarrying out the interactions, which mustall be funneled through the server. Thislimitation is even more evident whenmobility becomes part of the picture.People need to communicate and collab-orate even while in movement, and inde-pendently of their location. However, insimilar situations, server access is oftenprevented by technical or administrativebarriers.

We argue that a peer-to-peer approachholds significant advantages over tradi-tional client-server architectures. When apeer-to-peer architecture is adopted, dataand services are no longer gathered in asingle point of accumulation. Instead,they are spread across all the nodes of thedistributed system. Users may directlyhost the resources they want to sharewith others, with no need to publish themon a particular server.

Interestingly, these features are relevantnot only in mobile scenarios but also infixed ones, where the decentralizednature of a peer-to-peer architecturenaturally encompasses the case of multi-site or multicompany projects, whosecooperation infrastructure must spanadministrative boundaries, and is subjectto security concerns.

Unfortunately, most of the peer-to-peerapplications developed in recent yearsstarted from premises that are rather

different from those outlined thus far.They target the Internet and aim atproviding peer-to-peer computing overmillions of nodes, with file sharing astheir main application concern. Thedifference in perspective from thedomain of collaborative work is madeevident by their search capabilities,which typically do not guarantee tocapture information about all matchingfiles. In most cases they do not take intoconsideration features like security or theability to support reactive interactions,which are crucial in cooperative businessapplications. Moreover, they bring peer-to-peer to an extreme, where the logicalnetwork of peers is totally fluid, andnone can be assumed to be fixed andcontributing to the definition of a perma-nent infrastructure. This radical viewprevents access to resources exported bynon-connected peers, which is unaccept-able in the business world, where criticaldata is often required to be always avail-able, independently of its owner.

by Carlo Ghezzi, Gianpaolo Cugola, and Gian Pietro Picco

The advantages of a peer-to-peer architecture go well beyond the realm of Internetfile sharing, becoming crucial in supporting business processes and especiallycollaborative work involving mobile users. To support this view, we designedand experimented with PEERWARE, a core communication middleware forTeamWork applications.

A Peer-to-Peer Middleware for Mobile TeamWork



PEERWAREOn the basis of the above considerations,we have developed PEERWARE: apeer-to-peer middleware for teamworksupport specifically geared towards theenterprise domain.

PEERWARE is both a model and anincarnation of this model in a middle-ware. In developing both, our firstconcerns were minimality and flexi-bility.

The ModelThe PEERWARE coordination modelexploits the notion of a global virtualdata structure (GVDS), which is a gener-alization of the LIME coordinationmodel. Coordination among units isenabled through a data space that is tran-siently shared and dynamically built outof the data spaces provided by eachaccessible unit.

The data structure managed by PEER-WARE is a hierarchy of nodescontaining documents, where a docu-ment may actually be accessible frommultiple nodes, as shown in Figure 1.This structure resembles a standard filesystem, where directories play the role ofnodes, files are the documents, andUnix-like hard links are allowed only ondocuments.

When a peer is isolated, it is only givenaccess to its own tree (stored locally) ofitems (ie, nodes and documents).However, when connectivity with otherpeers is established, the peer has access

to the virtual tree constructed by super-imposing the trees contributed by all thepeers in the system, as illustrated byFigure 2.

In search of minimality, PEERWAREprovides only three main operations tooperate on the GVDS:• the execute operation allows peers to

execute an arbitrary piece of code on aselected set of items held by connectedpeers. The results are collected andreturned to the caller

• the subscribe operation allows peers tosubscribe to events occurring on aselected set of items, while

• the publish operation allows peers tonotify the occurrence of events.

By exploiting these primitives, peers canquery the GVDS and also subscribe toevents and receive the corresponding

notifications. The hierarchical structureof the GVDS provides a natural scopingmechanism, thus leading to an efficientimplementation of searches.

The MiddlewareCurrently, the PEERWARE model hasbeen implemented in two middleware:one, developed in Java, has been used asthe core of the MOTION platform, theother, developed in C# under theMicrosoft .Net infrastructure, is the coreof the PeerVerSy configuration manage-ment tool.

Both implementations are tailored to thebusiness domain and distinguishbetween a set of permanently availablebackbone peers, and a fringe of mobilepeers, which are allowed to connect anddisconnect as required. To optimizerouting, these peers are connected to

Figure 3:

The PEERWARE run-

time architecture.

Figure 2:

Building the GVDS in

PEERWARE.

N1

N3N2

N7

N9N8 N10

D6

D3

D1

D8 D7

N1

N3

N4

N6

N2

N5

D2

D5

D4

N1

N3

N4

N6

N2

N7

N9N8 N10

N5

D6

D2

D3

D5

D4

D1

D8 D7

Peer A(DS1)

Peer B(DS2)

GVDS

Figure 1: The data structure provided by PEERWARE.

N1

N3

N4

N6

N2

N7

N9N8 N10

N5

Node

Document

D6

D2

D3

D5

D4

D1

D8 D7



form an acyclic graph in which themobile peers represent the leaves, asshown in Figure 3.

Access control and security are criticalissues in the domain we target and theyare addressed by two separate modules.One provides mechanisms to establishencrypted channels among peers and tomanage the security informationnecessary to authenticate a peer. Theother embeds the actual security policythat determines the capabilities of agiven peer.

To increase flexibility, the functionali-ties provided by the security modulesand also by the repositories holding localdocuments are sharply decoupled from

the specific implementation provided forthese functionalities. Thus, the securityprotocols, as well as the format of thesecurity information used to performauthentication, and the repository effec-tively used can be changed easily, eg, toadapt them to the common practice of aspecific business environment.

ConclusionsCollaboration defines a scenario whereinteraction is intrinsically peer-to-peer.We exploited this idea by developingPEERWARE, a peer-to-peer middle-ware explicitly tailored for collaboration.The model of collaboration adopted byPEERWARE is minimal but expressiveenough to support complex collaborationschemes, including both reactive and

proactive interactions. Current incarna-tions of this model in a middleware havebeen tailored to the domain of enterprise-wide collaboration. We are now workingon a different implementation orientedtoward more dynamic scenarios,including ad-hoc networks.

Links:PEERWARE: http://peerware.sourceforge.netMOTION project:http://www.motion.softeco.it/pages/PeerVerSy website:http://sourceforge.net/projects/peerversy/

Please contact: Gianpaolo Cugola, Politecnico di Milano, Italy Tel.: +39 022 399 3493E-mail: [email protected]

Fluid Computing is a sub-area ofPervasive Computing. It denotes thereplication and real-time synchronisationof application states on several devices.Thus the application state flows as a'fluid' between devices.

There are three main application areas:

1. Multi-device applications: severaldevices may be temporarily coupled tobehave as one single device, forexample, a mobile and a stationarydevice.

2. Mitigation of the effects of variableconnectivity: applications on ubiquitousdevices can exploit full or intermittentconnectivity, or operate in disconnectedoperation - all in a seamless fashion.

3. Collaboration: multi-person applica-tions enable several users to collaborateon a shared document in synchronous

mode (at the same time) or in asyn-chronous mode (at different times)without the user having to switch theapplication.

The Fluid Computing middleware repli-cates data on multiple devices andachieves coordination of these devicesthrough synchronisation. Each devicehas a replica of the application state,which allows it to operate autonomously.A synchronisation protocol tries to keepthe replicas consistent, depending on thequality of the network connectivityavailable (weak consistency). The FluidComputing synchronisation protocoloperates in two modes. Batch modereplication occurs when connectivity isregained after disconnection,exchanging all updates that have accu-mulated during the disconnection.Trickle mode replication occurs as longas there is some connectivity, propa-

gating updates in real-time as soon asthey are generated at a replica.

The weak consistency approach enablesdisconnected operation: we want toallow the user to continue working witha device and making updates even whenthe device has no connectivity. Replicasare therefore allowed to become incon-sistent when disconnected. The propaga-tion of updates for synchronisation hastwo special properties, explained below.

Epidemic update propagation: A userwho carries several devices (mobilephone, PDA, watch etc) will often be in asituation in which the devices haveconnectivity between one another via awireless proximity network (egBluetooth), but no wide-area connec-tivity. We want to be able to use theproximity connectivity to make thedevices consistent among themselves.Updates can thus be exchanged with any

It has long been desirable to be able to let an application 'flow' from one userinterface to another, depending on the user's situation and the capability of theinterface being used. The IBM Zurich Research Laboratory has developedmiddleware which realises this wish. The service provided is called FluidComputing.

Fluid Computingby Marcel Graf



other replica, may even propagatethrough several other replicas, and arenot required to go through a centralserver.

Real-time update propagation: Mobiledevices are limited in theirhuman/computer interaction capabilitiesbecause of their small size. This problemcan be overcome by temporarilycoupling several devices tightly via awireless proximity network. Tightcoupling means that a user action (eg akeystroke) on one device is immediatelyreflected on the other devices, such thatthe user experience is that of a singledevice. This enables the combination ofdevices to be optimised for a certaininput or output modality, for example,the combination of a mobile phone witha large stationary display. Thereforewhen there is connectivity between a setof replicas, and the user performs anupdate, this update is sent immediatelyto the other replicas. The replicatedapplication state allows a user todecouple devices at any time withoutbreaking the application.

The Fluid Computing middleware is alightweight Java library that runs onPDAs such as Sharp ZaurusTM orPocketPCTM devices. The data structureof the replicated data is that of an XML

document (other data structures aresupported by the middleware as plugins).The API used by applications to manipu-late that data structure is the DocumentObject Model (DOM) API of the W3C.Application developers interact onlywith the XML data structure and areshielded from the complexities of thereplication mechanism.

A Sample ScenarioImagine yourself to be a doctor at arenowned Zurich Clinic on the shores ofthe Lake of Zurich, where the rich andfamous regain their health. All doctorsand nurses are equipped with PersonalDigital Assistants (PDAs) that give themaccess to their patients' data whereverthey are. A wireless LAN providesconnectivity to the PDAs throughout theclinic's premises.

You are walking along an aisle whilereading the data record of Annelie Gryffon your Compaq iPAQ *. You have todecide on her exercise program. Firstyou prescribe some swimming. Becausethe display is small you soon find your-self scrolling back and forth. You need alarger screen! You walk to a PC. Gryff'srecord appears on the PC, including theswimming exercises you have alreadyentered. You continue listing prescrip-tions using the PC. Every keystroke is

immediately reflected on your PDA. Inan emergency you could immediatelyleave the PC - it would not be necessaryto press a 'sync button' to synchronise thetwo devices.

For lunch you decide to go to Hiltl'srestaurant, a vegetarian restaurant nearBahnhofstrasse. Obviously there is nolonger a wireless connection to the clin-ic's intranet, but you still have Gryff'spatient record on your PDA. You decideto prescribe a vegetarian diet - maybethat will do some good for her weight.As soon as you return to the premises ofthe clinic, this prescription propagates tothe central patient database.

Link: http://www.zurich.ibm.com/csc/mobile/

Please contact:Marcel Graf, IBM Zurich Research LabSwitzerlandTel: +41 1 724 8318E-mail: [email protected]

A Fluid Computing application is no longer tied

to a particular device. It may run

simultaneously on several devices which the

Fluid middleware coordinates, for example a

small mobile PDA (left) and a big stationary PC

display (right). When wireless network

connectivity is available synchronization

occurs keystroke-by-keystroke, otherwise

updates are queued until connectivity is

regained.


A wide variety of new interactive plat-forms are now being offered on themarket, ranging from small devices suchas interactive watches to very large flatdisplays. This is changing the nature ofmany interactive applications,converting them to nomadic applicationswhich need to support user accessthrough different interactive devices,from various locations. Consequently,one fundamental issue is how to supportsoftware designers and developers inbuilding such applications. In particular,there is a need for innovative methodsand tools able to support the develop-ment of interactive software systems thatcan adapt to different targets whilepreserving usability.

In current practice the design of multi-platform applications is obtainedthrough the development of severalversions of the same application, one foreach platform considered. Such versionscan at most exchange data. This solutionis somewhat limited because it implieshigh implementation and maintenancecosts. Another solution proposed is theuse of style sheets. Each platform isassociated with a different set ofstylesheets. Thus, the same elements arepresented differently according to thetype of platform available. This can beuseful, but it is still insufficient; stylesheets only help by trying to bettersupport the same tasks through differentplatforms. However, this is not alwaysmeaningful because users often want todo different things according to the typeof platform, and there may be dependen-cies among tasks performed throughdifferent platforms.

We have developed a solution for thesethorny issues. Our approach can be

summarised in four words: One Model,Many Interfaces. This means that westart with an abstract description of theactivities to be supported and obtaindifferent user interfaces for each avail-able platform in order to support theseactivities. In particular, we start with atask model of a nomadic application,which describes the activities that shouldbe supported in order to reach the user’sgoals. Designers are able to create effec-tive user interfaces for the various plat-forms considered through a number oftransformations supported by our newtool TERESA (TransformationEnvironment for inteRactivE SystemsrepresentAtion).

This work is being carried out in the ISTEuropean CAMELEON Project. Theobjective of the project is to build

methods and environments supportingthe design and development of highlyusable context-sensitive interactive soft-ware systems by providing the means toexpress context-dependent informationand developing tools that support the useof the information contained in multiplelogical representations.

Within the context of this project, theHIIS lab of ISTI-CNR has been devel-oping TERESA, a publicly availabletool. The aim is to support the design anddevelopment of nomadic applications.The method underlying TERESA iscomposed of a number of steps (seeFigure) that allow designers to start withan idea of the overall task model of anomadic application and then deriveconcrete and effective user interfaces formultiple devices through multiple levels

Scientists at CNR propose a solution for the design and development of nomadicinteractive applications, accessible through heterogeneous platforms and fromdifferent locations. This solution is supported by TERESA (TransformationEnvironment for inteRactivE Systems representAtions), a mixed-initiative toolsupporting multiple transformations able to adapt the interactive part of anapplication according to the platforms employed.

Designing Usable Multi-Platform Interactive Applicationsby Fabio Paternò and Carmen Santoro


The main

phases of

the method

underlying

TERESA.



of abstractions and related in-betweentransformations.

A number of requirements have driventhe design and development ofTERESA: • Mixed initiative, the tool is able to

support different levels of automationranging from completely automatic tohighly interactive solutions wheredesigners can customise or even radi-cally change the solutions proposed bythe tool. This is important to satisfy avariety of needs. Situations where thetime available is short, the applicationdomain is narrow, or the designer hasno expertise call for completely auto-matic solutions. When designers areexpert or the application domain isbroad or has specific aspects, thenmore interactive environments areuseful because they allow the designerto make important design decisionsdirectly.

• Model-based, because the variety ofplatforms increasingly available canbe better handled through multipleabstractions that provide designerswith a logical view of the activities tosupport.

• XML-based, XML-based languageshave been proposed for every type ofdomain. In the field of interactivesystems there have been a fewproposals that partially capture the keyaspects to be addressed.

• Top-down, this approach is anexample of forward engineering.Various abstraction levels are consid-ered, and we support cases wheredesigners have to start from scratch,first creating logical descriptions, andthen moving on to more concreterepresentations until they reach thefinal system.

• Different entry-points, our approachaims at being comprehensive andsupporting the entire task/platformtaxonomy. However, there can becases where only a part of the task orplatform needs to be supported. Forexample, when only different devices(such as different brands of mobilephones) referring to the same type ofplatform are considered. In this case,there is no need for a nomadic taskmodel, given that only one type ofplatform is involved.

• Web-oriented, due to the spreading useof the web, we decided that Web appli-

cations had to be our first target.However, the approach is also validfor generating user interfaces for othertypes of software environments, suchas Java applications.

This approach has been applied to thedevelopment of a number of nomadicapplications and has emerged as apromising and effective solution whenaddressing the problem of generatingUIs for heterogeneous devices, capableof providing usable multimedia inter-faces for a broader set of mobile devices,including vocal interaction techniques.

Links:CAMELEON Project http://giove.cnuce.cnr.it/cameleon.html. HIIS lab, ISTI-CNR http://www.isti.cnr.it/ResearchUnits/Labs/hiis-lab/ TERESA: http://giove.cnuce.cnr.it/teresa.html

Please contact: Fabio Paternò, ISTI-CNRTel: +39 050 315 3066E-mail: [email protected]

Carmen Santoro, ISTI-CNRTel: +39 050 315 3053E-mail: [email protected]

The emergence of mobile computing hascreated new classes of applicationsdependent upon the user’s location,context and interaction with their currentenvironment. However, given theconstraints of the wireless environment(ie weak connection, poor networkquality of service and mobile deviceswith limited resources) developingdistributed applications within thisdomain is a complex task. Therefore,new middleware has emerged to masksuch problems from the user. Theseencompass synchronous (eg remote

method invocation) and asynchronous(eg publish-subscribe and tuple spaces)communication paradigms. However, itis the heterogeneous nature of the solu-tions that has created a new problem.Only applications and mobile servicesdeveloped upon the same middlewarestyle can interoperate with one another.For example, a tourist guide client appli-cation interacts with services available atthe user’s current location that providetourist information. If the client is imple-mented upon a particular middlewareplatform (for example, SOAP) it will

only function in locations that offertourist services across SOAP.

Similarly, services are advertised usingone of the contrasting service discoveryprotocols. At present, four main servicediscovery protocols exist: Jini, ServiceLocation Protocol (SLP), Universal Plugand Play (UPnP) and Salutation. In addi-tion, new protocols are emerging tobetter support the discovery of servicesin mobile environments and across ad-hoc wireless networks (eg ServiceDiscovery Protocol in Bluetooth and

by Paul Grace and Gordon S. Blair

Mobile applications are now developed upon a wide range of service developmentplatforms, commonly referred to as middleware. However, the diversity of thoseavailable presents a problem for mobile client development. How can a singleclient implementation interoperate with heterogeneous service implementations?

Interoperating with Heterogeneous Mobile Services



Salutation Lite). Mobile clients using asingle discovery protocol will missservices advertised by alternative proto-cols.

Hence, we argue that adaptive middle-ware is required to support the interoper-ation of mobile clients with heteroge-neous services. Using this approach,middleware dynamically alters itsbehaviour to: i) find the required mobileservices irrespective of the servicediscovery protocol, and ii) interoperateswith services implemented by differenttypes of middleware.

The ReMMoC (Reflective Middlewarefor Mobile Computing) project, whichbegan in October 2000 and concludes inSeptember 2003, is being carried out atLancaster University in collaborationwith Lucent Technologies UK. The aimof the project is to design and implementsuch an adaptive middleware, whichallows mobile clients to be developedindependently from the underlyingmiddleware implementations that maybe encountered at different locations.

For applications to dynamically operateacross different middleware they mustbe programmed independently fromthem. Hence, abstract definitions ofmobile services are required. The clientapplication, which requests this service,can then be developed using this “inter-face” in the style of IDL programming.A request of the abstract service ismapped at run-time to the correspondingconcrete request of the middleware

implementation. The emerging WebServices Architecture includes a WebServices Description Language (WSDL)that provides such an approach based onabstract and concrete service definitions.We employ WSDL as the basis ofservice description and programmingmodel.

To build the adaptive middleware, weutilise three key techniques: compo-nents, component frameworks andreflection. We use OpenCOM (devel-oped at Lancaster University) as ourcomponent model to develop compo-nents for Windows CE platforms.OpenCOM is a lightweight, efficient andreflective component model, built atop asubset of Microsoft’s COM. The motiva-tion behind component frameworks isthen to constrain the scope for evolution,ensuring all dynamic changes betweencomponent configurations are legiti-mate. Finally, reflection is a principledmethod that supports introspection of theunderlying middleware and providestechniques to dynamically adapt compo-nent architectures.

The ReMMoC platform consists of twokey component frameworks for bindingand service discovery, whose behaviourcan be dynamically altered using reflec-tion. The binding framework re-config-ures between different interaction proto-cols (we have implemented IIOP, SOAPand publish-subscribe bindings); thefigure illustrates a dynamic change ofpersonalities within this framework. Theservice discovery framework changes

between one or more of the differentservice discovery technologies (we haveimplemented the SLP and UPnP proto-cols).

ReMMoC has been fully developed andtested using simple applications eg chat,news and stock quote clients. Ongoingwork includes an evaluation of thismethod on larger, complex applications(eg ubiquitous computing and intelligenthome environments) and across a rangeof further middleware bindingsincluding data sharing and tuple spaces.An evaluation of memory use hasconcluded that each device cannot storeevery possible middleware component.Therefore, methods for dynamicallydownloading components when neededare required. Furthermore, techniques toensure components are available to start-up before they are used, eg predictivecaching based upon context information,are an interesting option.

Link:OpenCOM and ReMMoC Web Page -http://www.comp.lancs.ac.uk/computing/research/mpg/projects/opencom/

Please contact:Paul Grace, Lancaster University, UKTel: +44 1524 59 3801 E-mail: [email protected]

Gordon Blair (Project Co-ordinator)Lancaster University, UKTel: +44 1524 59 3801 E-mail: [email protected]

Dynamic change from SOAP client

configuration to IIOP client

configuration of OpenCOM

components.



Providing tailored content and presenta-tion for different clients in heteroge-neous environments is becomingincreasingly important. Devices such assmall palm computers, smart phones andpocket PCs are already common compo-nents of the computing infrastructure.Providing adapted content that takes intoaccount the complete range of devicesand user agents available is challenging,since the capabilities of these terminalsare evolving at a rapid pace. On the otherhand, Web content is mostly designed

for the capabilities (screen resolution,bandwidth etc) of desktop PCs and to alesser extent terminals with poorer capa-bilities such as first generation WAP-based systems. Designing an infrastruc-ture flexible enough to cope with thissituation requires the modification ofseveral components of the current Web.At the lower layers, one of the basicrequirements is the provision of aminimal knowledge about the differententities that compose the contentdelivery chain, ie client, server, docu-ment and media resources, etc. Startingfrom such knowledge, adaptation mech-

anisms and content-negotiation strate-gies could be applied to deliver to thetarget client content that takes advantageof its capabilities and satisfies itsconstraints.

The negotiation and adaptation core,called NAC, is an architecture developedin order to provide a solution for thedelivery of multimedia content in hetero-geneous environments. For contentgeneration, NAC uses dynamic andstatic adaptation techniques. The adapta-

tion is controlled using an adaptation andnegotiation module (called ANM), andan optional module, called UCM (usercontext module), that runs on embeddeddevices. UCM allows the ANM know-ledge about the client description to beenriched through the use of profiles.NAC default organisation is proxy-based(Figure 1), but the proxy entity can beomitted by using (installing) ANM at theserver side.

The proxy is a third entity that standsbetween servers and clients. It allowsbetter handling of the heterogeneity of

clients and the variety of content repre-sentations on servers. Indeed, proxiesallow content adaptation withoutaffecting the existing components in thesystem. For the content generation byadaptation, the proxy (Figure 2a) is theentity responsible for retrieving clientrequests and profiles and performingpossible adaptation on the contentreceived from the original server. Theadapted content is then sent to the targetclient with respect to its characteristics.The proxy can transform existing multi-media content, meaning content does notneed to be authored in multiple versions.In such situations and in order to avoidthe adaptation cost, a negotiationstrategy is used to select the best variantto be delivered. All the proxy tasks aredesigned to behave transparently toclients and servers.

In NAC, the description of the environ-ment (device capabilities and prefer-ences, network limitation, content char-acteristics etc) is described as a set ofconstraints that the content providershould satisfy. In our approach, theconstraints resolution strategy isachieved by adding constraints progres-sively to the original content.

The universal profiling schemata (UPS)was defined in the NAC framework tohave a central role in the generation ofadapted content by modelling thedifferent profiles of the environments.UPS extends CC/PP to describe not onlythe client capabilities and preferencesbut also the capabilities of servers andproxies, the characteristics of the content

In the current Web infrastructure, access to content is achieved from a variety ofterminals such as desktop PCs, smart phones, set-top boxes and personal digitalassistants (PDAs). The different capabilities of these terminals has led to thedevelopment of a variety of parallel Webs, each accommodating a particulardevice feature. This situation poses a serious problem to Web designers andmaintainers, since every piece of information needs to be authored for every typeof terminal with the associated protocols. Research in the WAM project at INRIAattempts to unify Web access to information with a more flexible and negotiatedapproach.

NAC: An Architecture for Multimedia Content Adaptation for Mobile Devicesby Nabil Layaïda and Tayeb Lemlouma

Figure 1: The NAC architecture.



and the adaptation methods that mayexist at the server or the proxy level.UPS identifies three main categories ofcontexts: the client category, the servercategory and the network category.Profiles in the first category contain thedescription of the client characteristics(software, hardware and user agent) andthe constraints related to the use ofdifferent media resources (supportedmedia formats, size, resolution etc). Theserver category includes the document

instance profile that describes the docu-ment characteristics and functionalities.It includes also the resource profile thatdescribes media resources used in docu-ments and the adaptation method profilethat describes available adaptationmethods that exist in the server or theproxy side. Finally, the network profiledescribes the network characteristics.ANM includes a set of adaptation tech-niques that ranges from real-time(re)encoding of media items such as

video, to structural transformations suchas XHTML to WML.

Links:http://wam.inrialpes.frhttp://opera.inrialpes.fr/people/Tayeb.Lemlouma/NAC.htmhttp://www.w3.org/Mobile/CCPP/http://www.w3.org/2001/di/

Please contact:Nabil Layaïda, INRIATel: +33 476 61 53 84E-mail: [email protected]

Figure 2: NAC entities: the ANM proxy, Pocket SMIL and the UCM module. From left to right: the proxy; Pocket SMIL, an embedded

multimedia player (centre); the User Context Module (right).

Three communication networks, namelythe emerging 'mobile' wireless, the clas-sical 'fixed' wireline, and the alternative'computer-centric' internet, are evolvingin a relatively independent way.However, their interactions increase dayby day. It is of paramount importance toinvestigate how these three systems couldbe integrated in order to form a commoncommunication space, in which the usersof any one of the networks can easily and'transparently' access the other two.

The main purpose of a project currentlyunder way at PisaTel is to investigate thefeasibility of such an integration. The

work is being carried out within the javaJAIN MAP API international project.

JAIN MAPThe aim of the JAIN (Java APIs forIntegrated Networks) initiative [1] is tointegrate wireline (PSTN), wireless(PLMN) and packet based (IP and ATM)networks, in order to modify the actualproprietary market into an open one. Thenovelty of JAIN lies substantially in twofundamental aspects: the service porta-bility obtained by inserting Java inter-faces over the protocols, and the networkconvergence ensuring the technologyindependently of the services.

We are interested in the JAIN MAP API.MAP (Mobile Application Part) is aprotocol in the GSM stack, for mobilitymanagement and other services. JAINMAP API provides an abstraction levelover the complicated MAP interface. Inthis way, a MAP service developer doesnot have to be aware of the specific MAPimplementation features. At present, fourMAP capabilities are specified in JAINMAP: transaction (corresponding to theSMS service), session (USSD service),position (MAP location service) andinformation. We have focused on sessioncapability, realizing the USSD(Unstructured Supplementary ServiceData) service, which allows information

by Rita Arrighi, Maurizio A. Bonuccelli, Francesca Lonetti and Francesca Martelli

A project at PisaTel, a laboratory located at ISTI-CNR in a joint collaborationbetween ISTI, Ericsson Lab Italy, Pisa University and the “Scuola S. Anna”, aimsat integrating wireless, wireline and Internet networks.

Integrating GSM Networks and Internet: New Unstructured Services



exchange between a mobile station and aGSM network application.

The USSD service is very similar to theSMS service. The main differenceconsists in the session-oriented nature ofthe USSD service which makes itnecessary to establish a session each timea customer (a mobile station) or a networkapplication approaches a USSD service.As the USSD service is not a store-and-forward one, it is faster than SMS. AUSSD session can be network initiated ormobile initiated: in the first case, thesession is started by a network applica-tion; in the second one, the session isstarted by a mobile station.

ApplicationsWe concentrated on integrating GSMnetworks and Internet. In particular, wedesigned and tested two new mobile initi-ated USSD services for GSM customersallowing them to easily interact with aserver connected to Internet. The assump-tion on the customer side is minimal:customers only need a simple mobilestation (eg cellular telephone handset)able to communicate on GSM cellularnetworks, and the network operator must

be able to support the new services. Thissecond assumption is not cumbersome,since the software for these two newservices is not large and is quite fast torun.

The first service is e-mail related. TheGSM customer first invokes the newservice by inserting an appropriate code,followed by an internet ID, the password,and the address of a mail server. Thesedata are delivered to an entity inside theGSM network, called GSMscf, which isalso connected to Internet. The GSMscfopens a session with the mail server and,after being authorized by entering the userID and password, requests the number ofunread emails that are in the targetmailbox. The answer received byGSMscf is then propagated to the GSMcustomer that asked for the service, and isshown on the display of the cellularphone.

The second service is designed for moresophisticated customers. It enables theactivation of a secure Telnet session witha specified computer connected toInternet. The data exchanged during thissession must be reproducible on the

display of the available handset. The dataexchange is performed similarly to theprevious service, with more roundsbetween the handset and the remotecomputer, and with a particular emphasison the security aspects in the Internet partof the data path (security in the GSM partshould be guaranteed by the networkoperator).

In this research we showed that GSMcustomers can effectively interact withInternet-connected devices by easilydeveloping new simple software applica-tions, without changes in the systemarchitecture. We shall continue in theproduction of new services for larger inte-gration of Internet and GSM systems.

Links:JAIN:http://java.sun.com/products/jain/api_specs.html

USSD:http://www.mobileussd.com/whatis.asp.

Please contact:Francesca Martelli, ISTI-CNRTel: +39 050 315 3468E-mail: [email protected]

The goal of several recent researchprojects has been to study the new gener-ation of mobile communication andinformation services, based on self-organisation (AODV, Terminodes,CarNet). Such systems have latelybecome very topical, with the advent ofthe peer-to-peer communicationparadigm and the emergence of ad hocnetwork technologies. Many funda-mental technical questions remain openhowever, and real business applicationsstill need to be rolled out. One specificdomain where the peer-to-peer

phenomenon has had a major impact isdata and media interchange on theInternet. Mobile devices will presum-ably be the next target, as the minimalrequirements for multimedia applica-tions are almost guaranteed.

An ad hoc network is “a transitory asso-ciation of mobile nodes which do notdepend upon any fixed support infras-tructure. [...] Connection and disconnec-tion is controlled by the distance amongnodes and by willingness to collaboratein the formation of cohesive, albeit tran-

sitory community.” [Murphy et al]. Butwhat are ad hoc applications? One couldsimply answer 'pieces of software thatwill run on ad hoc networks'. We believethat defining a class of applications (adhoc applications) by coupling it to onlyone possible implementation technology(ad hoc networks) reduces the generalityand effectiveness of the definition.Therefore we define an ad hoc applica-tion as a self-organising applicationcomposed of mobile and autonomousdevices interacting as peers, whose rela-tionships are made possible because of

A new framework for ad hoc applications is under development at the Universityof Lausanne in Switzerland. This thin layer, based on existing standards, will helpdevelopers build applications that include three features: mobility, peer-to-peeroperation and collocation. As the framework is independent of the underlyingnetwork technology, the proposed solution will be adaptable to the emergenceof the next-generation ad hoc network.

From Ad Hoc Networks to Ad Hoc Applicationsby Benoît Garbinato and Philippe Rupp


29ERCIM News No. 54 July 2003

relatively close physical distances(collocation). In addition, this dynamiccommunity, defined by geographicalproximity, needs to have a common(application-level) interest. Moreformally, three features must be presentin an application for it to deserve the adhoc label:• Mobility: in order to be able to use the

application everywhere, the usershould not be limited by range. Therange limit is set by the business logicof the application.

• Peer-to-Peer: direct communicationbetween peers is mandatory. Thismeans that the client/server relation-ship is defined in an ad hoc manner bythe application logic (direct interactionbetween the pieces of software).

• Collocation: all logical interactionsbetween applications must result in aphysical interaction between users.This means that in order to be called anad hoc application, the service has tobe location-based.

The above definition allows us toabstract the network completely, andfocus on the application aspects. In thisway, there is a clear decoupling betweenthe application and the network, and anytype of network — GSM, WLAN or ad

hoc — can be used. This decoupling isfundamental in understanding the keyissues underlying ad hoc applicationsand in providing reusable solutions tosolve these issues. This eases the devel-opment of ad hoc applications. Byabstracting the network level, we areable to build ad hoc applications in theabsence of any ad hoc network, as longas the underlying infrastructure canprovide support for the three basicaspects described above.

Here is an example of a possible ad hocapplication called the ubiquitous fleamarket. This application is availablewherever you are and at all times. It isavailable on many mobile devices andmatches buyers and sellers presentwithin a certain range, the latter beingpreviously defined by the user. As theuser is walking, this ad hoc applicationscans its surroundings for possible peersellers or buyers. It has to be noted thatany user can be buyer and/or seller.When the application finds anothermobile device that runs the same piece ofsoftware, it scans the shared items inorder to find a match. If there is a match,the user is alerted and can then ask thepeer to get in touch and make the phys-ical transaction.

The ubiquitous flea market (see Figure1) is based on a framework to ease itsdeployment. Our so-called ad hoc appli-cation framework is based on existingstandards, which makes it lightweight,and facilitates its adoption by devel-opers. Some frameworks close to oursactually redefine, for instance, a trans-port protocol (PeerWare and Proem).

This is not necessary as with the adventof Java-enabled mobile devices, most ofthese basic services are already availableand standardised. Figure 2 presents thedifferent layers for an ad hoc applicationwhen based on our framework.

The work is still in progress and the firstversion of the framework will be avail-able in summer 2003. It will be demon-strated using a first implementation of anad hoc application, the ubiquitous fleamarket described above.

Links: Inforge - Institute for Business Information Systems Lausanne: http://inforge.unil.ch

[Murphy and al.]:http://citeseer.nj.nec.com/murphy98exercise.html

Terminodes project:http://www.terminodes.org

AODV: http://www.cs.ucsb.edu/research/trcs/abstracts/2002-18.shtml

CarNet: http://www.pdos.lcs.mit.edu/papers/grid:sigops-euro9

PeerWare: http://www.elet.polimi.it/Users/DEI/Sections/CompEng/GianPietro.Picco/papers/esec01.pdf

Proem: http://www.cs.uoregon.edu/research/wearables/Papers/p2p2001.pdf

Full version of the paper: http://isi.unil.ch/radixa/usr/prupp

Please contact:Benoît Garbinato and Philippe Rupp, Inforge, University of Lausanne, SwitzerlandE-mail: [email protected], [email protected]

Figure 1:

The ubiquitous flea

market use case.

Figure 2: The different layers of an ad hoc

application.


30

ARES – the Architecture of Network ofServices – has been set up as a newINRIA Project in 2000 and is now hostedby the INSA of Lyon, France. ARESworks on the design, modelling and vali-dation of software and hardwareelements for deploying services andprotocols into networks. Network archi-tectures are becoming increasinglycomplex and heterogeneous both interms of the technologies on which theyare based and the areas in which they aredeployed. The notion of communicatingobjects and ambient networks, in whichinteractions between network elementsare spontaneous and transparent to theend user, significantly complicates thedeployment of these infrastructures.Thus, new models have to be realised,new approaches have to be validated,and new architectures proposed with twoprinciple concepts in mind: mobility andembedment.

ARES is currently developing dynamicsoftware components that would beuseful in middleware approaches, takinginto account the specific characteristicsof hybrid networks,namely:• in the design of elementary services,

mobility must be considered the rule,rather than the exception

• potential loss of availability of somenodes and strong variation in thequality of service is also the rule, notthe exception.

Many concepts have to be revisitedbecause of this particular context, espe-cially the service discovery engine andnotions of service deployment andadministration.

WaveCar The aim of WaveCar is the dynamicdeployment of services in a distributedenvironment. In this project, mobility is

one of our main interests, since deploy-ment of services will take place on a car-embedded platform. The idea is to equipthe car with a system which will providesynchronisation with the requisite clientservices through an Internet connection.

The other important point is that the caris not always connected to the network;in fact it must be near an access point toget a connection through radio waves.This off-line mode shows the importanceof synchronisation between the clientexpression of needs and the effectivedeployment of services.

Our architecture separates three impor-tant elements of our system - the clientadministration platform, the servicepoint (access point) and the autonomousphysical entity:

by Stéphane Frénot, Anis Krichen and Stéphane Ubéda

As everything becomes dynamic and pluggable, we must envisage applicationscreated through a combination of ambient elementary services without the helpof an infrastructure. The WaveCar project investigates the notion of elementaryservices in a highly dynamic environment, in order to create what are now referredto as ambient and hybrid networks.

Light Support for Dynamic and PervasiveServices on P2P Networks

Figure 1: WaveCar functional architecture. Figure 2: WaveCar general architecture.




• Client Administration Platform(CAP): This platform software isinstalled on the client computer. It hasthree main functions: to authenticatethe client, to localise the client’s carand to request management. The clientcan visualise the actual state of theembedded system and take manage-ment actions. These are performedimmediately if the car is connected tothe network, otherwise synchronisa-tion will be performed next time thecar connects.

• Service Point:Service points are thecore of the system. These are allconnected through a P2P network.This is an interface between the CAPand the autonomous physical element.

It essentially provides a networkconnection to the embedded system, acaching system for localisation andrequests management. This cachingsystem is essential for synchronisa-tion. All service points must commu-nicate in order to achieve homoge-neous management tasks like physicallocalisation of cars or event-sendingfor upgrades concerning a specialAutonomous Physical Element.

• Autonomous Physical Element (APE):This is the car's embedded system. Inour case it is a platform for the deploy-ment of services which is in confor-mity with the OSGi specification. Therole of this platform is to dynamicallyload new services on demand.

Services are represented by bundleswhich can be downloaded from a servicepoint and directly managed through theservice platform. These services couldbe MP3 players, vocal post-it, mailclients etc.

Links:http://citi.insa-lyon.fr/http://darts.insa-lyon.fr/http://www.insa-lyon.fr/

Please contact:Stéphane Frénot, CITI Laboratory at INSA-Lyon Scientific and Technical University, FranceTel: +33 04 72 43 64 22E-mail:[email protected]

The future success of wireless networkslike UMTS and WLAN depends on thelarge availability of innovative services.Such services will drive network usageup for data and voice services and willhopefully make network technologieslike UMTS worthwhile over the longterm. So far the services provided by aparticular operator have been stronglytied to the particular capabilities of theoperator’s network vendor. This verticalapproach to service provisioning under-mines innovation; services are excludedfrom specialised service developersbecause the services typically have torun in a proprietary execution environ-ment. In the last few years there hasbeen considerable worldwide turmoil inthe area of communication servicedevelopment and deployment, leadingto different service architectures,

different interfaces and different serviceenvironments. On the other hand, thethird-generation mobile networks arecurrently under intensive study by inter-national standardisation bodies such asthe ITU and ETSI, as well as theEuropean Commission. Where possible,Fraunhofer FOKUS has tried to mergethese resulting architectural andnetwork concepts in order to provide aglobal service-provisioning architec-ture, coming from the different solutionsand capabilities suggested and providedby various projects.

Starting from TINA-C’s service archi-tecture and appropriate inter-domainreference points, Fraunhofer FOKUShas developed a series of service execu-tion environments. These provide well-defined interfaces between services and

execution environments on the onehand, and between execution environ-ments operated by different providers onthe other. One lesson learned was that itis very hard to agree on a global stan-dard for the interfaces between servicesand their execution environment. Thusevery execution environment defines itsown internal interfaces to services.These interfaces become more and morecomplex when the execution environ-ment provides VHE support. Additionalinterfaces between the service and adap-tation and profile components have to beintroduced. Appropriate topics havebeen investigated in a series of VHE-related projects like Vesper, AlbatrOSSor VHE-UD.

A new category of problems arises in thecase of mobile, nomadic users. These

by Ali Hafezi and Klaus-Peter Eckert

A Virtual Home Environment (VHE) is a means for the context-dependent provisionof telecommunication-related services to nomadic users. The Fraunhofer instituteFOKUS has participated in several national and international projects that havedefined VHE architectures and implemented appropriate platforms. While thecurrent work focuses on the personalisation and adaptation of services, futurework will concentrate on operational aspects of VHE systems including the flexibledeployment of services, comparable to the deployment of components in asupporting application server.

Towards a Virtual Home EnvironmentApplication Server



users need personalised access to theirservices even if they connect from othernetwork or VHE providers. In such situ-ations, either interworking is necessarybetween service instances running in thehome and visited provider’s domain, orcross-domain access to profile informa-tion is required. These scenarios areinvestigated in the AlbatrOSS project.Again the lack of globally standardisedinterfaces between services and execu-tion environments prevents the provi-sion of a comprehensive solution.

One possible solution to these kinds ofproblems is the introduction of servicedeployment descriptors that describe therequirements and obligations of servicestowards their users and execution envi-ronments. As in well-known componenttechnologies like EJB or CCM, theservices as well as the execution envi-ronments describe their required and

provided interfaces and the appropriateprotocols. Thus it is possible to mapboth descriptions either manually orautomatically and to generate bridgingcomponents that adapt the technologyand protocols between services andexecution environments. First steps inthis direction have been made in theVHE-UD project.

This adaptation during the deploymenttime of a service needs only static infor-mation. It can be used to satisfy service-level agreements and to guarantee someagreed quality of service during thestart-up phase of a service. Much morecomplex is the task of controlling andjustifying the quality of service duringthe execution time of the service.Components are required that monitorthe status of the network and of theservices, the end-to-end evaluation ofthe provided service quality is necessary

and an appropriate end-to-end reconfig-uration of the service must be enabled.In a fully operational environment, themonitored status information will alsobe used for logging and accountingpurposes.

In current projects, Fraunhofer FOKUShas developed various building blocksfor a comprehensive VHE platformtogether with a series of component-based, adaptive middleware environ-ments. It will be the challenge of futureprojects to integrate these ideas, experi-ences, and software modules into aglobal VHE architecture built aroundthe concept of VHE application servers.

Link:http://www.fokus.fraunhofer.de/platin/projects

Please contact:Ali Hafezi, Fraunhofer FOKUSTel: +49 30 3463 7109E-mail: [email protected]

The market for mobile phones, handheldcomputers and wireless PDAs is increas-ingly driven by multimedia-basedInternet applications. New demandsrelevant to mobile commerce servicesinclude the electronic purchase oftickets, goods, or audio/visual content.The Telecommunications and NetworksLaboratory (Netlab) of ICS–FORTH isdeveloping and implementing an M-commerce API package for mobilephones, intended to provide mobilecommerce services, achieve widespreadusage, and offer unique benefits over andabove the alternatives. The design of thecorresponding API complies with securehardware modules already installed onmobile phones (eg SIM cards) andupcoming advanced secure memorymodules (SecMMCs).

To achieve this, M-commerce has to bebased on secure applications and solu-tions to gain trustworthiness. SecMMCat ICS-FORTH’s NetLab places strongsecurity at the centre of this project. Thechallenge is to implement a securityscheme that meets the end-user require-ments and which, at the same time, isconvenient and simple to use. The poten-tial revenue that multimedia and dataservices can bring to the industrydepends on the end-user perception ofsecurity and trust, ie, whether the mobilecan become a 'digital wallet'. M-commerce should be viewed in thecontext of the number of ways that endusers will be able to pay for goods andservices.

To that end, we are developing securemobile phone applications, and estab-lishing a framework for secure mobiletransactions to support a variety of appli-cations and services. In order tomaximise the potential of such services,a number of factors are crucial forsuccess.

Firstly, end users require device owner-ship: although anyone can pick up thedevice, only its owner must be allowedto carry out transactions involvingpersonal data. Secondly, end usersrequire secure and reliable networks.Thirdly, all brands and companiessupplying services must be able to betrusted. This is true for all components ofthe value chain, including the financialinstitution, the retailer, the network oper-

by Ioannis G. Askoxylakis, Diomedes D. Kastanis and Apostolos P. Traganitis

M-commerce, the new buzzword in today’s mobile industry, opens new horizonsfor mobile phones and services. Originally used to place voice calls, today’smobile phone uses data, multimedia, entertainment, and mobile commerceservices, which are expected to overtake voice in revenue generation for operators.

M-commerce API Package for Mobile Phones



ator and the device manufacturer. Otherimportant factors include allowing endusers to record and print transactions,ensuring that services are simple, easy touse, and reasonably priced.

Mobile commerce applications requirethe existence of a secure module. In ourcase we consider a variety of securitymodules, including:• the broadly used SIM-card module • the forthcoming Secure Multimedia

Card (SecMMC) module• similar smart memory devices (as

specified in the draft standard 'MobileCommerce Extension Standard-McEX; Core specification' by the 5Cgroup).

(The 5C Group is an industrial standard-isation group of the five mainMultimedia Card manufacturers:Infineon Technologies Flash Ltd,Hitachi Ltd, Matsushita ElectricIndustrial Co Ltd (MEI), ToshibaCorporation, and SanDisk Corporation.)

The SecMMC can be viewed as a clas-sical Multimedia Card with an integratedsmart-card kernel, which combines thehigh security of smart cards with thelarge memory capabilities of classicalMultimedia Cards. Obviously thesupported services depend on the secu-rity module being used (SIM module orSecMMC). For example, whereas bothmodule types can provide an electronicpayment system, memory-demandingmultimedia applications can only besupported by the SecMMC.

The SecMMC is being developed by aconsortium consisting of GuardeonicSolutions AG Germany (project coordi-nator), Infineon Technologies Flash LtdIsrael, IAIK University of Graz,Technische Universität München,Fraunhofer-Gesellschaft-IGD, FORTH,Infineon Technologies, FlashGmbH&Co KG and Mühlbauer AG.

Our development of the M-commerceAPI for mobile phones focuses on threebasic areas of application:• mobile payment• mobile entertainment (or infotain-

ment)• mobile business.

Mobile Payment applications allowpersonal payment transactions like e-ticket purchase, local merchantpayments, stock trading, and eventuallytelematic and transportation usage. Thistype of application can be supported byboth SIM and SecMMC modules. Keycharacteristics of this application are:

• Anonymity:during an economic trans-action nobody can extract any infor-mation concerning the correspondingtransaction pair.

• Off-line operation:two parties cancomplete a transaction without theintervention of a third centralisedintermediary entity.

• Transferability:each entity of thesystem should have a hybrid role, bothas a payer and as a payee.

• User mobility:the payment process isindependent of the location of thetransaction pair.

In addition, Mobile Entertainment appli-cations involve copyrighted content(music, video and eventually navigationcontent). Consumers want to personalisetheir mobile phones with ringtones,graphics and picture messages fromcontent providers. Games, downloadableapplications, music and video feeds aresure to follow, as secure modules like theSecMMC become available for integra-tion into mobile phones. In some regionsof the world this type of system isalready in use to a limited degree.

Mobile Business applications securelyconnect mobile users to both internalcompany and external public organisa-tions. This type of application is anessential part of comprehensive salesforce automation implementation.

All these emerging mobile servicesinvolve large volumes of multimediadata and require high levels of security toprotect the transmission and storage notonly of financial accounts and PIN codesbut also of the multimedia data that isdownloaded and stored.

Please contact:Apostolos P. TraganitisTel: +30 2810 391724E-mail: [email protected]

Ioannis G. AskoxylakisTel: +30 2810 391723E-mail: [email protected]

Diomedes D. KastanisTel: +30 2810 391723E-mail: [email protected]

The M-commerce API for mobile

phones focuses on three basic

areas of application: mobile

payment, mobile entertainment and

mobile business.



Wireless networks are rapidly becomingcommon place; upcoming technologiessupport wireless communication onchip-sets; new laptops have built-inWiFi cards; hotspots are being installedin airports, hotels and coffee shops; andoffices are converting their existing localarea networks to wireless to allow theirworkers take advantage of mobility.Soon numerous embedded devices - thatform the heart of the ubiquitouscomputing revolution - will join thisdomain and increase the number ofparticipants exponentially. Existing

network architectures that use dedicatedcentral nodes to coordinate the commu-nication between participants willbecome more and more infeasible as thenumber of participants increases byorders of magnitude and where deviceswill more likely converse with oneanother in an ad hoc fashion.

Ad hoc networks present the next greatchallenge for distributed systemsresearch. In an ad hoc network, theparticipants are used to route communi-cation traffic from senders to receivers.Every participating node in thesenetworks executes a routing algorithm

that allows messages to be directedtowards the next node along a route tothe receiver. Protocols that supportcommunication in ad hoc networks haveto take into account the mobility of theparticipants and the variation in theconnectivity between associated parties.New applications in these ad hoccommunication scenarios can takeadvantage of the peer-to-peer character-istics of the network and offer location-based services that address specificcommunities of users.

To support research in this area theDistributed Systems Group in theDepartment of Computer Science atTrinity College Dublin, in collaborationwith Media Lab Europe (MLE), isdeploying WAND, the Wireless Ad hocNetwork for Dublin, as a large-scale test-bed for ad hoc network protocols andapplications. The network will cover thecentre of Dublin along a 2km route fromTrinity College to MLE. This area willbe seeded with a number of custom-buildwireless-enabled embedded PCs. ThesePCs are 3x3x6 inch containers thataccommodate a stack of PC/104 boards,WiFi PCMCIA cards, and a set of

antennae. The embedded PCs will behosted in apartments and shops, ontraffic lights, and in phone kiosks alongthe route to provide a minimum level ofconnectivity.

The embedded PCs form a sparse popu-lation of wireless network nodes. Thissparse coverage is constantly availableand the embedded PCs can be config-ured to create a variety of networkmodels. The test-bed can be furtherpopulated through the introduction ofmobile nodes such as laptops, PDAs, andother mobile devices with wirelessconnectivity. This ability to be config-ured in various ways and to be populatedsparsely or densely enables researchersto develop and investigate protocols andapplications that are specific to the areaof ad hoc networking in variousscenarios.An example of such an investi-gation is the examination of routingprotocols for ad hoc networks. Therouting of messages in an ad hocnetwork represents a difficult problemthat is addressed by several mature orrecently proposed routing protocols.These protocols have distinct advantagesand disadvantages and it is important todetermine the exact circumstances underwhich each of these protocols representsa good choice. The WAND test-bedoffers a unique opportunity to explorethe behaviour and performance ofrouting protocols in a real-life environ-ment and to investigate requirementsthat are necessary to provide servicessuch as location-based information orquality-of-service provision.

Another area of research that is investi-gated with the help of the WAND test-bed concerns the adaptation of applica-tions to wireless ad hoc networks.Existing applications are rarely designedto take advantage of additional informa-

by Stefan Weber, Vinny Cahill, Siobhan Clarke and Mads Haahr

The migration of wireless networks from hotspots to multihop ad hoc networksis the first step towards self-organized global routing. The WAND project isdeploying a large-scale test-bed for ad hoc network protocols in the centre ofDublin to facilitate research in this emerging discipline.

Wireless Ad Hoc Network for Dublin: A Large-Scale Ad Hoc Network Test-Bed

WAND Route from Trinity College Dublin to Media Lab Europe.



tion and services that are offered by adhoc networks. The applications rangefrom peer-to-peer file sharing mecha-nisms that facilitate the exchange of datain ad hoc networks to location-basedservices that provide data to usersdepending on their physical position inthe network.

An example of the applications that willbe explored with the help of the WANDtest-bed is communication between vehi-cles. This application will help mobilenetwork nodes like cars, buses, andtrucks communicate with one anotherand with static network nodes such as

traffic lights, bus shelters, and trafficcameras. Mobile participants mayinform one another about obstacles onthe roads, actions that are going to betaken to avoid obstacles, or the routes ofemergency services vehicles. Staticnodes may provide information aboutthe state of the traffic in an area andexploit information from mobile nodessuch as the number of passengers or theestimated arrival time of buses.

The WAND test-bed offers researchers aunique opportunity to investigate low-level mechanisms such as routing proto-cols and low level services, as well as

user-level applications in a real-life envi-ronment that reflects all the randomnessand unpredictability that is extremelydifficult to be reproduced with simula-tion.

Links:http://wand.dsg.cs.tcd.iehttp://www.dsg.cs.tcd.iehttp://www.mle.ie

Please contact:Stefan Weber, Trinity College Dublin, IrelandTel: +353 1 608 1543E-mail: [email protected]

Lightning Active Node Engine (LANE)is an active network platform developedat VTT Technical Research Centre ofFinland. It is based on a decoupled activenode architecture which separates theactive applications from non-active IP-related routing. This decoupling wasused in former active network projectsCaspian and MAO.

The LANE architecture is depicted in theFigure. Decoupling is implemented bydividing the Active Node into an ActiveRouter and an Active Server.

The Internet lacks a number of servicesrelevant to mobile users such as auto-matic conversion proxies, dynamic fire-walls etc. Active Networking orprogrammable networks can answerthese service requirements. However,this raises some fundamental implemen-tation issues on Active Nodes.

These issues have been addressed in thesystem design of LANE, which is basedon the Application Layer Active

Networking (ALAN) concept. The workis divided between the two parts, theActive Router (AR) and the ActiveServer (AS). The routing in the ActiveRouter is dynamically configurable, andall time- or processor-consuming activi-ties (or Active Services) are located andexecuted in the Active Server.

The Active Server handles the creation,execution and management of activeservices. All active services are on-the-fly deployable Java binaries. For securityreasons, users can only referenceservices by name. Active services aredesigned and implemented for the usersrather than for the servers. As a proof of

by Sami Lehtonen

There is a strong trend towards wireless networking and smaller terminal devices.The advent of various wireless terminals such as PDAs, mobile phones, or similarhandheld equipment, will result in greater and greater volumes of Internet traffic.The key issues are support for seamless roaming, security and content adaptation.

Lightning Active Node Engine - An Active Network User Service Platform

The LANE architecture



concept we have implemented dynamicMobile IP support for our Active Serverframework.

Services needed by mobile users can belocated anywhere in the network, followthe user, and balance server loads. In oursystem, we designed an entity called theUser Agent. This entity is created by theActive Server according to the user’sprofile, which is located in a user profiledatabase. There are User Agents withdifferent privilege levels for differentpurposes (or users). The security of thedesign is ensured with separately imple-mented User Agent classes. Theseclasses implement only the interfacesthat are granted for a particular user.

The User Agent represents the user on aparticular Active Server. All services for

the user at that Active Server aremanaged by the User Agent, and the usermight have his/her User Agent instanti-ated in several Active Servers around thenetwork. This makes it possible to moveservices from one Active Server toanother, balancing the load betweenActive Servers. If the user then movesaway, abandoning the User Agent, theremoval of the User Agent removes allservices for the user at that ActiveServer.

In our system, the services are mainlyimplemented as atomic, meaning that asingle service instance serves only oneuser. We have followed a strict designprinciple, namely, there are multipleservices for a user and multiple users at aserver.

The research group included KimmoAhola, Mikko Alutoin, MarkusAroharju, Arto Juhola, Titta Koskinen,Sami Lehtonen, Marko Lyijynen, SamiPönkänen, and Pertti Raatikainen.

Future work will consist of imple-menting load-balancing between UserAgents located in different ActiveServers. We will implement a method totransfer the state of a User Agent and itsservices to another Active Server (whena Mobile Node moves to another foreignnetwork). Furthermore, implementingother useful services will prove thedesign of our concept.

Please contact:Sami Lehtonen, VTTTel: +358 9456 7240E-mail: [email protected]

In order to highlight some of the chal-lenges inherent to ubiquitous computing,we use the example of ubiquitousgaming. Based upon elements of tradi-tional multiplayer games like 'TheSettlers' or 'SimCity', we introducevarious ideas of how to adapt them foruse in a ubiquitous environment.

For instance, while travelling by train,passengers could play a game together.At train stations, some players mightleave while others join the game,possibly taking over existing roles.People could also collaborate, joiningforces as well as dividing work. In addi-tion to that, the environment itself couldhave an influence on the game. Forinstance, train stations might representmarketplaces in which a large variety of

goods are offered and collaboration withother players waiting for their trains ispossible. Aside from that, severalsophisticated services, hosted forexample on a powerful stationary node,might be available only at such speciallocations (see Figure).

Obviously, middleware support for ubiq-uitous applications is a very big field.Currently we are focusing on twospecific areas within this complexdomain, which will be discussedthroughout the subsequent sections.

Information BrokeringOne problem area to be tackled is that ofsharing and distributing informationbetween network nodes, ie, not onlybetween all participants in a single appli-

cation, but also across different applica-tions, where some information (eg,where other devices are likely to befound) might be of general interest. Forthat purpose, we introduce the concept ofinformation brokers. An informationbroker is a service running locally oneach device, collecting and dissemi-nating data in the background whilewithin the coverage area of otherdevices.

From a logical perspective, a single setof information is available via the localbroker. Physically, however, data is fullydistributed over all nodes, with a localcache holding copies of the most inter-esting information. Queries are issuedagainst the local broker, which can notonly return data stored locally, but also

Mobile devices such as Personal Digital Assistants (PDAs) are currently usedmostly for personal information management. However, these devices arebecoming powerful enough to run much more sophisticated applications not onlyin an isolated manner, but even, due to their wireless communication capabilities,in a cooperative way. We are looking at middleware support in the domain knownas ubiquitous computing.

Middleware Support for Ubiquitous Computingby Luc Hogie and Christian Hutter



send agents into the network. Theseagents collect information from devicesin the local surroundings, therebysupplementing the query results in anincremental way. For non-time-criticalqueries that are allowed to run for sometime, it is even possible to collect infor-mation from nodes that were notconnected to the network at the time thatthe query was initiated. In the gameexample, a player might want to look forthe best place where to buy or sell itemsof a specific type, dependent upon thecurrent price and other criteria. As thiskind of information might becomeoutdated after a period of time, appro-priate strategies, eg for limiting informa-tion lifetime, need to be provided. Inparticular, the trade-off between replicaconsistency and communication over-head is a complex research topic.

In the context of this task, we are collab-orating closely with the University ofTrier, Germany, where a macroscopicsimulation environment for multihop adhoc networks has been developed.Information brokers represent a serviceintegrated with that environment. Byusing the simulation, we can test ouralgorithms with a large number ofdevices. The Trier team in turn employsour service in their own applications,thereby giving us valuable feedback.

Service Architecture and Mobile GridAside from sharing information betweenthe nodes in the network, it is alsosensible to share resources, eg in termsof computing power and services. Up tonow, scientific and business applicationshave benefited from networking mostlyby using distributed resources forstorage and retrieval of information ordistributing the computational effort ondifferent nodes.Most research on distributed and parallelcomputing is based on an existing andfixed network infrastructure and there-fore leads to fairly static solutions suchas PVM, MPI, CORBA, RMI, etc.With the emerging ubiquitouscomputing paradigm, certain significantlimits relevant to current middlewaremust be addressed. In particular, middle-ware is:• usually tied up with a central naming

mechanism• built for use of workstations or servers,

meaning most current middleware isheavy and therefore unsuitable for useon lightweight devices.

Our aim is to define elegant models of aservice architecture for ubiquitous appli-cations. We are interested in the conceptof a large-scale operating system usingthe resources available not only on thelocal machine but everywhere in thenetwork. This might be considered as an

extension of current well-known peer-to-peer networks like Kazaa or Gnutella.These open networks focus on filesharing. The idea is to allow nodes insuch a network to act as serviceproviders. Assuming a device can makeavailable services of any type, thenetwork is able to represent a computa-tional grid, a data-sharing system, oreven a distributed database. We arecurrently working on the model (usingthe metaphor of human society andcommunication) and its implementation.Once completed, it will be the lowestlayer of (and a first step towards) amiddleware and runtime environmentdedicated to applications for the mobilegrid.

This research is done in collaborationwith the Computer Science Laboratoryat the University of Le Havre (France),where work is currently being done onthe specific issue of load balancing forad hoc networks.

Please contact:Luc Hogie, Christian HutterLuxembourg University of Applied SciencesE-mail: [email protected], [email protected]

Illustration of ubiquitous

gaming scenario.

temporary collaboration betweentrain and marketplace

networks

Ad-hoc gaming network

marketplace



The recent proliferation of cheap, small,and increasingly accurate sensor tech-nologies is creating a new informationrevolution where applications thatinteract with the physical environmentare becoming widespread. This aware-ness of the physical environment, and itsuse by applications in fulfilment of theirgoals, is known as context-awarenessand a number of promising applicationshave appeared which make use ofcontext information. Context-awarenessis of particular importance in mobileenvironments where the operating envi-ronment is constantly changing due tothe mobility of devices and the charac-teristics of wireless communication tech-nologies. In such environments, context-awareness can enable applications torespond intelligently to variable band-width, unreliable connections and theeconomy of different connections. Oneof the greatest challenges in context-aware computing, and one that has not

yet been adequately addressed, is theprovision of middleware and services tosupport the application developer. Themajor problem lies in providing genericsupport for the acquisition and use ofmultiple fragments of informationgleaned from (potentially unreliable)multi-modal sensors in a mobile envi-ronment.

The use of context information by appli-cations in a mobile environment poses anumber of challenges arising from thedistributed and dynamic nature ofsensors, the accuracy and resolution ofsensors, and the fusion of output ofmultiple sensors in order to determinecontext. In addition, the mobile environ-ment poses further challenges withregard to the dependability,predictability, and timeliness of commu-nication. Middleware is required thatprovides abstractions for the fusion ofsensor information to determine context,

representation of context, and intelligentinference. Essential services that providesupport for operation in a mobile envi-ronment, such as supporting the relia-bility of communication, are alsorequired.

The sentient object model defines soft-ware abstractions that ease the use ofsensor information, and associated actu-ation, by context-aware applications. Atthe heart of the model is an event basedcommunication model that permits loosecoupling between objects and conse-quently supports mobility and applica-tion evolution. The event-based commu-nication model includes mechanisms forthe specification of constraints on thepropagation and delivery of events. Asensor is a software abstraction thatencapsulates a hardware sensor deviceand produces software events inresponse to real-world stimuli. An actu-ator consumes software events andencapsulates a physical device capableof real-world actuation. A sentient objectis an entity that can both consume andproduce software events, and lies insome control path between at least onesensor and one actuator.

Internally, a sentient object consists ofthree major functional components:• the sensory capture component is

responsible fusing the outputs ofmultiple sensors, and uses proba-bilistic models, including Bayesiannetworks, to deal with inherent sensoruncertainties

• the context representation componentmaintains a hierarchy of potentialcontexts in which an object can exist,and the current active context

by Gregory Biegel and Vinny Cahill

Despite rapid advances in sensor technologies making a wealth of environmentaldata available to applications, programmer support for the development of context-aware applications remains poor. The majority of such applications are developedin an ad-hoc, application-specific manner and there is an urgent need formiddleware and services to support the development of such applications. Mobile,intelligent software components known as sentient objects provide onemiddleware abstraction that can ease the development of mobile context-awareapplications.

Sentient Objects: Towards Middleware for Mobile Context-Aware Applications

The sentient object model.



• the inference engine component is aproduction rule based inference engineand supporting knowledge base,giving objects the ability to intelli-gently control actuation based on theircontext.

Sentient objects are cooperative and inaddition to traditional forms of commu-nication, the sentient object model usesstigmergic coordination, or coordinationvia the environment. Stigmergic coordi-nation does not rely on direct communi-cation between objects, say via TCP/IP,but rather depends on objects being ableto sense and make changes to their phys-ical environment. Simple rules,embedded in the inference engine, thengovern actuation and consequentlybehaviour, according to what has beensensed from the environment. This typeof coordination is extremely robust andis particularly suited to mobile environ-ments where traditional communication

channels may not always be available oreconomical to use.

We are developing middleware andassociated services, based on the sentientobject model, to support context-awareapplications. Such middleware andservices will ease the task of developingapplications based on context perceptionin a number of key areas. Applicationswill be insulated from both the complex-ities of physical sensors, actuators, andassociated protocols, as well as thefusion of sensor data to infer context andreduce uncertainty. The middleware willalso provide a high-level rule specifica-tion language through which intelligencemay be added to objects withoutrequiring knowledge of the intricacies ofa specific production rule system.Furthermore, an intuitive visualprogramming tool is under developmentthat will permit the rapid development ofcontext-aware applications.

The work described in this article waspartly supported by the Future andEmerging Technologies programme ofthe Commission of the European Unionunder research contract IST-20000-26031 (CORTEX – CO-operating Real-time senTient objects: architecture andEXperimental evaluation).

Links:Ubiquitous computing at Trinity College Dublin:http://www.dsg.cs.tcd.ie/index.php?category_id=228

Cortex Project: http://cortex.di.fc.ul.pt/

Please contact:Gregory Biegel, Trinity College DublinTel: +353 1 6081531E-mail: [email protected]

Vinny Cahill, Trinity College DublinTel: +353 1 6081795E-mail: [email protected]

Pervasive computing is now a reality,and intelligent devices flood manyaspects of our everyday lives. As statedby the Semiconductor IndustryAssociation, semiconductors integratedin traditional computers currently repre-sent less than 50% of a market of$US204 billion. As new applicationsappear, the need for database techniquesembedded in various forms oflightweight computing devices arises.

Sensor networks gathering weather,pollution or traffic information havemotivated several recent works (eg theCOUGAR project at Cornell). Theyhave brought out the need to executelocal computation on the data, likeaggregation, sort and top-n queries,either to save communication bandwidth

in push-based systems or to participate indistributed pull-based queries. Otherforms of autonomous mobile hostsinclude hand-held devices (eg personaldatabases, diaries, tourist information,etc). These can be used to execute on-board queries on data downloadedbefore a disconnection. Personal folderson a chip constitute another motivationto execute on-board queries. Typically,smart cards are used in various mobileapplications involving personal data(such as healthcare, insurance, phonebooks, etc). In this context, queries canbe fairly complex and their executionmust be confined to the chip to preventany disclosure of confidential data.Thus, saving communication costs,allowing disconnected activities andpreserving data confidentiality are three

different concerns that motivate thestorage of data and the execution of on-board queries on lightweight computingdevices.

The resource asymmetry of lightweightdevices entails a thorough re-thinking ofdatabase techniques. Our research objec-tive is therefore to design embeddeddatabase components that can matchhighly constrained hardware resources.

Preliminary studies led us to design andvalidate a full-fledged DBMS, calledPicoDBMS, embedded in an advancedsmart card platform (powerful CPU, tinyRAM, fast read-time and dramaticallyslow write-time in EEPROM stablestorage). PicoDBMS is based on acompact and efficient ad-hoc storage and

Ubiquitous computing and ambient intelligence introduce the need for embeddingdata and evaluating database queries in a growing variety of ultra-light computingdevices. The SMIS project at INRIA focuses on the design of database componentsembedded in a chip. Applications range from mobile computing to dataconfidentiality protection.

Database Components on Chipby Nicolas Anciaux, Luc Bouganim and Philippe Pucheral



indexation model. The PicoDBMS queryevaluator takes advantage of the CPUpower and the fast read-time in theEEPROM to process any type of SQLquery without using RAM, whatever thevolume of the queried data.

Capitalising on this work, new andimportant research efforts need to beundertaken: • to better capture the impact of each

device hardware constraint ondatabase techniques

• to propose new storage, indexationand query techniques allowing ad-hocembedded database components to bebuilt

• to set up co-design rules helping tocalibrate the hardware resources offuture devices to match the require-ments of specific applications.

As a first step in this research agenda, weare conducting a comprehensive analysisof the RAM consumption problem. Wetry to answer three important questions.First, does a memory lower bound existwhatever the volume of the queried data?Second, how can a query be optimisedwithout hurting this lower bound? Third,how does an incremental growth ofRAM impact the techniques devised in alower bound context? This studyprovides helpful guidelines to calibratethe RAM resource of a hardware plat-form according to a given application’srequirements, as well as to adapt anapplication to an existing hardware plat-

form. In the mid-term, we plan to focuson energy consumption and on the usageof future stable-memory technologies.

Besides mobile computing and ambientintelligence, database componentsembedded in a chip may serve other

important purposes in a wide range ofapplications. Typically, tamper-resistantchips (eg smart cards or tokens) can beused to make information systems safer.Data security is particularly emphasisedin mobile environments. For instance,the increasing connection of travellingemployees to corporate databases andthe need to resort to Web-hostingcompanies and Database ServiceProviders to make personal data avail-able to mobile users introduce newthreats to data confidentiality. Usershave no choice other than to trust theprivacy policies of Web companies.However, according to the FBI, databaseattacks are more and more frequent (theircost is estimated to be more than $100billion per year) and 45% of the attacksare conducted by insiders.

While traditional database security prin-ciples like user authentication, commu-nication encryption and server-enforcedaccess controls are widely accepted, theyremain inoperative against insiderattacks. Several attempts have beenmade to strengthen server-based securityapproaches thanks to database encryp-tion. However, as Oracle confesses,server encryption has not proven to be

the 'armour plating' that was expected,because the Database Administrator (oran intruder usurping his/her identity) hasenough privilege to tamper with theencryption mechanism and get the clear-text data. Client-based securityapproaches, where decryption onlyoccurs at the client-side, have beenrecently investigated but they cannotsupport data-sharing among users havingdifferent access rights.

The client-based solution we are devel-oping at INRIA, called C-SDA (Chip-Secured Data Access), allows encrypteddata to be queried while controllingpersonal privileges. C-SDA comprisesan access right manager, a queryrewriter, a query evaluator and a crypto-graphic engine, all embedded into asmart card to prevent tampering. Aspictured in Figure 1, C-SDA acts as anincorruptible mediator between a client(potentially mobile) and the encrypteddatabase. To illustrate the benefits of thisapproach, C-SDA allows complexpermissions to be expressed onencrypted data, for example, 'give user Xthe privilege to compute the averagesalary of all employees without givingher access to the elementary data'. Thiscooperation of hardware and softwaresecurity allows the orthogonalitybetween access-right management anddata encryption to be reestablished.

This work is partly supported by theFrench ANVAR agency (AgenceNationale pour la VAlorisation de laRecherche) and by Schlumberger, whichprovides us with advanced smart cardplatforms. This initial work considersrelational data, pull-based queries and anSQL-compliant access-right model. Newsolutions will be devised to tackle morecomplex data (eg XML), push-basedmodels (selective information broad-casting) and sophisticated access-rightmodels. Some of these solutions will bestudied in the context of a MEDEA+European project scheduled for thefourth quarter of 2003.

Link:SMIS project: http://www-smis.inria.fr/

Please contact:Nicolas Anciaux, Luc Bouganim, Philippe Pucheral, INRIAE-mail: {Firstname.Lastname}@inria.fr

C-SDA functional architecture.



At the present, users must carry one ormore portable devices that act as dataand program repositories or as connec-tion devices. These devices need to beadministered, backed-up, and synchro-nized; their batteries need to be replacedand recharged. This is not a 'pervasivetechnology', but rather an 'invasive tech-nology'. The next step towards a newconcept of services that support usermobility is known as 'ambient intelli-gence' where computers are no longerrepositories for data and programs, butbecome windows to the environment:services and applications. Ubiquitousapplications adapt themselves to the enduser delivery context, becoming acces-sible from any device and giving the userthe perception of moving in an empow-ered environment. Such requirementsemphasize the concept of the 'position':'ubiquitous computers must know wherethey are'(Weiser).

User mobility implies that applicationsshould be accessible from a number ofdifferent channels: from a web browser,from a WAP phone or from anInteractive Voice Responder (IVR)machine. A given channel can providemultiple interaction modalities. Forinstance, a Windows-Icons-Mouse-Pointing (WIMP) desktop applicationcan make it possible to dictate texts bymeans of a voice recognizer.Applications should be designedaccording to the principle of: designonce, display anywhere.

The e-mate ProjectIntegrated development environmentsare aimed at building applications whosedelivery context is known a priori,limiting their reuse in different environ-ments. In the e-mate project, we have

designed and implemented an architec-ture to support the development of appli-cations which are seamlessly deliverablein multiple contexts. Developers candesign and implement the software,abstracting from the technology detailsand focusing only on applicationconcepts and relations between them.For instance, in the development of a'Personal Agenda', developers only dealwith conceptual details such as appoint-

ment, person, address, calendar, etc. andconceptual relations such as 'whichaddresses are relevant to a given person';'which persons are involved in a givenappointment' and so on. Any otherconcern about technology constraintssuch as network protocols, transactionmanagement, service discovery, datapersistency, adaptation of components touser terminals, and user interface presen-tation, should be of no importance asthey are already addressed by the e-mateinfrastructure.

The rationale behind e-mate is influ-enced by the object oriented paradigm.Objects are self-contained, behaviouralcomplete pieces of data with variablesand procedures. They are reactive enti-ties which are activated wheneversomeone invokes one of their proce-dures. The reaction of an activated objectis to produce another object (the result)and/or a side effect changing its internalstate.

Objects can be considered as buildingblocks, able to react and therefore tointeract. Our system provides aviewing/controlling mechanism to auto-matically fill the gap between the end-users and the application objectsallowing direct interaction betweenthem. This 'bridge' is implemented forany type of terminal used by the user.Thus, we can put into practice a designbased solely on the definition of businessrelated objects and their relations.

by Davide Carboni, Gavino Paddeu and Stefano Sanna

Techniques and frameworks that assist designers and programmers in thedevelopment of applications that are seamlessly portable from one device toanother are still a challenge. The e-mate project at the Center for Advanced Studies,Research and Development in Sardinia (CRS4) tackles this issue proposing atechnique based on the object-oriented paradigm with a strong focus on thedelivery of services on multiple platforms.

Object-driven Application Design: from Mobility towards Ubiquitousness

The Agenda application objects running on a desktop PC, a smartphone and a Personal

Digital Assistant.



The issue is to take business objects andgenerate a user interface aimed at inter-acting visually with them. To addressthis critical point we have developedMORE (Multi-platform ObjectREnderer). The rendering process usesreflective capabilities of the Javalanguage to extract the 'displayable'information directly from the code. Suchinformation consists of the set of publicprocedures and public variables that areincapsulated in the model object.

Complex models are viewed ascomposite graphical objects and therendering process is applied recursivelyuntil atomic interactors are obtained. Theadvantage of this approach is that thesame model can be accessed from anydevice for which a rendering engine hasbeen implemented (see the Figure).

The rendering process takes into accountthe end-user device. Since displayingand communication capabilities mayvary completely on different devices, theengine adapts the model description (interms of data format and size) to takeadvantage of the target user interface andavailable communication channel. The

current implementation uses XML as aprotocol language to convey the interac-tion between the end-user and the appli-cation business objects. Depending onthe processing capabilities of the clientplatform, the user can perform a simpleinteraction or eventually access anenhanced set of native graphical compo-nents that fully exploit the capabilities ofthe device.

Mobile devices rely on wireless (andoften expensive) connections, thereforesmart strategies have been implementedto reduce bandwidth occupation. Therendering engine sends only portions ofinterface that are truly renderable by theclient; on the other side, the client tries toperform most of the computing locally,using the communication channel onlywhen required. Finally, since mostmobile devices have small displays andone-hand-operations user interfaces, therendering engine sorts interface elementsby usage, giving faster access tofrequently used functions.

Some specific applications require bothinterface and data adaptation to supportmobile client delivery. For example,

Location-Based Services require mobilecartography functionalities and presentcritical issues when deployed on mobileterminals over wireless connections. Toaddress such issues, we have developeda lightweight version of the GeographicMarkup Language (called 'compactGML'), which is used by the renderingengine to encode maps according toclient displaying capabilities.

The e-mate project was successfullycompleted in early 2003 and some of theresults will be used in future activities.For example, we plan to release theMulti-platform Object REnderer asOpen Source software and improve it interms of usability, and the cGML will beevaluated and used in an EU Project(EurEauWeb IST-34182).

Links:CRS4 - Networked Distributed Applications:http://www.crs4.it/nda

E-mate project: http://www.crs4.it/nda/e-mate

Compact GML: http://www.crs4.it/nda/cgml

Please contact: Gavino Paddeu, CRS4, ItalyTel: +39 070 925 0284E-mail: [email protected]

In mobile computing, users are facedwith a dynamically changing environ-ment, and are offered a large and varyingrange of remote wireless devices. In thissituation, users may carry small 'pocketstyle' processing units, emphasisingmainly wireless networking (such asBlueTooth), an embedded OS, and somelocally running applications, while allinteraction may evolve dynamically byutilising such remotely available devices(ie the small machine offers no UI equip-

ment). Dynamic configuration and conti-nuity of interaction are key ingredients inthis world of mobile applications,supporting distributed user interfaceswhere the input and output (I/O) devices(such as buttons, text lines, rasterdisplays, loudspeakers etc) are discov-ered 'on the fly' during user mobility.

The dynamism and continuity of interac-tion inherent in these scenarios are at theroot of the 2WEAR project, for which

FORTH (coordinator), ETHZ(Switzerland), MA Systems & Control(UK), and NOKIA/NRC (Finland) havejoined forces to explore mobilecomputing as an agglomerate of small,wearable and physically distributeddevices. Funded by the EC, 2WEARtook off on January 1, 2001 with theobjective to build such a system, in termsof both software architecture and devicesthat communicate using short-range

by Anthony Savidis and Constantine Stephanidis

One of the main challenges in mobile computing is to maintain continuous high-quality interaction between users and interactive applications. The Voyager toolkit,developed at the Human-Computer Interaction Lab of ICS-FORTH, provides allthe necessary implementation components to build interactive applications thatare capable of utilizing different types of User Interface input-output resources,as they are dynamically discovered or lost during mobile interaction.

Development Tools for Dynamic Environment-Adapted Mobile Interfaces



radio technology, and deliver it byDecember 2003.

In dynamic, transient environments, themain architectural challenge is to main-tain high-quality interaction betweenusers and applications. This principle ofcontinuity emphasises the uninterruptedsequence of dialogue activities andensures that the human-computer inter-action dialogues transform gracefullyfrom the user’s perspective. Developedat the Human-Computer Interaction Labof ICS-FORTH, the Voyager Toolkit is aUser Interface (UI) development frame-work (delivered as a C++ softwarelibrary) which provides interfaceprogrammers with all the necessaryimplementation components to buildinteractive applications that utilisedifferent types of UI I/O resources, asthey are discovered or lost during inter-action. As shown in Figure 1, a 2WEARUI is implemented as a typical client UI,while the UI I/O resource distribution ishandled underneath, through the2WEAR communication technology.

The Voyager toolkit places abstractinteractive objects (the Selector andTextEntry) at the service of theprogrammer, hiding run-time coordina-tion and control details, and revealingonly semantic characteristics at the APIlevel. These objects are internally linkedto the lower-level management ofremote UI resources to cater for I/Ocontrol through various levels of special-isation that are completely hidden to theapplication programmer. Programmersthus have an effective instrument toeasily craft mobile dialogues withdiverse dynamic UI I/O resources.Furthermore, the single-focus policy forobjects and applications has beenenforced, meaning that a single applica-tion can have the focus in a mobilesystem and that only the focus objectwill actually allocate resources for inter-action purposes. This modality, which isconcurrently supported in Windowingsystems, turned out to be a necessarydecision in mobile interactions, since thephysical display 'breadth' of the deviceswhich are dynamically engaged does notallow concurrent dialogues to be practi-cally supported.

When new resources appear, if dialogueis 'suspended', no action is taken, sincethe application does not own the focus.Otherwise, the Voyager UI frameworksubstitutes the instantiation of the currentfocus abstract object with the bestpossible I/O instantiation withoutaffecting the state of the focus abstractobject. When new UI resources arediscovered, the UI framework finds thebest possible I/O instantiations by recy-cling the resources currently used by thefocus, plus those just discovered, as wellas those that are not granted and are thusavailable. If the best instantiations foundare the ones currently in use, no actionsare taken. Otherwise, the current I/Oinstantiations are released (assumingdialogue was not 'stalled', in which casethose instantiations are null), while thecurrent best-found input and outputinstantiations are activated (when outputinstantiations are activated, or when theyare associated with a different abstractobject, they call their display methodautomatically). In the case where thedialogue was previously in a 'stalled'state, and is now in a 'working' state,dialogue revival is automaticallysignalled.

When UI resources are lost (either due tonetwork failure, or to lack of responsive-ness, or because the services are grace-fully removed), if the dialogue is already'stalled', no further action is taken.Otherwise, the UI framework asks the

current focus object if it happens to useany of the lost resources. If not, no actionis taken, but if this is the case, the currentinput and output instantiations arereleased, and a configuration round isentered, seeking for the best possibleinput and output instantiations. If nosuch instantiations can be found with thecurrently available resources, thedialogue is signalled as 'stalled'.Otherwise, the input and output instanti-ations that have been found are acti-vated.

To conclude, the challenges in imple-menting mobile interactive applicationscapable of dynamically utilising the UIresources discovered in the surroundingenvironment are: (a) handling thedynamic resource coordination andcontrol requirements at runtime, and (b)offering a compact and invariantprogramming model which is intact fromthe variety of dynamic UI resources. TheVoyager UI development framework hasbeen expressly designed and imple-mented to address these two challenges.

Please contact:Constantine Stephanidis, ICS-FORTHTel: +30 2810 39 1741E-mail: [email protected]

The development role of the Voyager User Interface development framework in

developing mobile distributed applications in the 2WEAR project.

Client UIcode

Client non-UIcode

IOresource

IOresource

IOresource

IOresource

service service service service

2WEARnet / component

technology

Localcalls

Net messages orcomponent

calls / notifications

2WEARinteractiveapplication



Jenny works at 'Care4U', a homecareorganisation with about 1100employees. In order to control costs andboost productivity Care4U has providedevery employee with a Personal DigitalAssistant (PDA) that has a wirelessconnection (GPRS= General PacketRadio Service) to the central server.Jenny registers most of her activities onthe PDA. The data is used for hourlogging as well as updating the medicalrecords of the patients. Jenny trusts theautomated system better than the paperpredecessors. It saves her time andreduces mistakes as compared to theprevious paper forms.

Such a scenario sounds easy enough butit is far from that. Before it can berealised, a number of technical andorganisational problems have to besolved. Medical data is privacy sensitiveand needs to be highly accurate. In theCare4U scenario, data is entered andstored on the PDA. Then it is sent to thecentral database of Care4U in real timeto avoid synchronisation conflicts. Aback office application processes thedata to fit the administrative system andupdate the medical records of Care4U.The medical records in the ownership ofthe patients themselves are updated aswell. These records are stored andmanaged by an intermediary organisa-tion acting on behalf of the patients

Data integrity starts at the source, so thePDA application should be carefullydesigned to avoid mistakes (rememberthe presidential elections in the US of2000). The PDA application should checkif the data is correct (is the number withinthe logical range of this data type?).Security starts at the source as well, so inorder to be able to register, the user shouldbe identified. The connection betweenPDA and the central database of Care4U,

and between Care4U and the interme-diary organisation handling the patients’medical record, should be secure. Care4Ushould be authorised to update thepersonal medical records, and the datatypes and information structure of bothdatabases should match.

Solving these mostly technical problemswithin the Care4U organisation is rela-tively easy. In most cases, standard busi-ness applications can be adapted to thespecific situation and habits of Care4U.The problem becomes much more diffi-cult if integrated healthcare has to besupported using this kind of applicationsacross several institutes.

Without underestimating the technicalproblems to get these systems to work,this is only part of the problem if you wantto bring new mobile services to themarket. The Freeband - B4U project has

the ambition to accelerate the marketintroduction of new mobile services andtherefore has designed a method todevelop business blueprints for suchservices. This method approaches a busi-ness idea from four different perspectives.

To accelerate market introduction ofservices and service development youneed to look not only from a technicalperspective. You also need a clear viewon:• The end user perspective:The value

proposition describes the serviceoffered to specific customers in aparticular market segment. Mostimportant is the assumed customervalue of the service concept, ie, what isthe added value of the service ascompared to existing alternatives? Towhat customer needs does the serviceappeal? In the homecare example boththe care worker and the care institutehave to be taken into account whendefining the value proposition. Whatyou functionally want to offer deter-mines what you technically need.

• The organizational perspective:Thevalue network is the configuration ofactors that possess complementaryresources and capabilities, whichtogether perform activities to realisethe service. Due to financial consider-ations, increased specialisation in thewireless industry and increasedinvolvement of other sectors likemedia and IT, it is no longer possiblefor a single company to develop amobile service all alone. For examplein the Care 4U scenario you need atleast a mobile network operator, adevice manufacturer and a systemintegrator. Each actor in the networkhas some resources, among whichtechnical ones, which need to becompatible and connected.Responsibilities have to be clearly

The Business Blueprint gives a broader look at service development, integratingthe technical perspective with the end-user, organizational and financialperspective. This method was developed within the Freeband B4U project thataims at accelerating the introduction of new mobile services. The project is acooperative effort of the Telematica Instituut, TNO Telecom and Delft Universityof Technology, supported by the Dutch Ministry of Economic Affairs.

Introducing New Mobile Services Fasterby Timber Haaker and Oscar Rietkerk

Figure 1: Jenny registers her activities on

her PDA.



defined. In the Care4U case the situa-tion is complicated due to the existinglegacy systems at the care institute.Integration is hard to realise.

• The financial perspective:Therevenues generated from the servicewill depend on the price and the tariffstructure. Whether customers likeCare4U are willing to pay the pricedepends on the value they perceive.For Care4U that means the cost reduc-tion and productivity boost they mayrealise plus the advantages theyperceive for Jenny. For a sustainablecooperation in the value network thefinancial arrangements are critical.Only if a fair distribution of benefits,investments, costs and risks can bearranged a sustainable value networkcan be realised. Costs strongly dependon the chosen technical implementa-tion. Some of the risks are directlyrelated to the technical choices: proventechnology, future proof concept andscalability.

Finally, the necessary co-operationbetween parties and the interdependen-cies between the perspectives is acomplex process to direct through all thestages of service development. That is

why the Business Blueprint method wasdeveloped to manage this process moreeasily.

It is a structural approach to simultane-ously design the four perspectives anddevelop an integral business design. Theblueprint may be applied:• explorative: to assess a business

opportunity of an organization withinthe value network

• creative:to develop a business oppor-tunity with partners

• evaluative:to evaluate the strong andweak points of a particular businessdesign.

The Business Blueprint method has beensuccessfully applied to several businessventures, other Freeband projects andprototypes developed within B4U. TheBusiness Blueprint comes in a variety offorms: consultancy, workshops, check-lists and a business game. It gives abroader look at service development,integrating the technical perspectivewith the end-user, organizational andfinancial perspective. It speeds up theprocess and increases the success rate ofservice development while maintaininggrip on complexity.

Link:TNO Telecom: http://www.tno.nl

B4U project:http://www.freeband.nl/projecten/b4u/

Please contact:Timber Haaker, Oscar Rietkerk, TNO Telecom, The NetherlandsE-mail: [email protected],[email protected]

Figure 2: The four

perspectives for fast and

successful development of

new mobile services.

End user perspective Organizational perspective

Prices, tariffs

Revenues, benefits

Costs, investments

Risks

What are viable financial arrangements ?

Roles and actors

Relationship to oneanother

Influences and power

Which actors are necessary for realising this service ?

Service description

Value proposition

Assumed customer value

Context

Is there a need for this service ?

Financial perspective

Necessary technicalfunctions

Alternatives forarchitecture

How can I realise the newservice technically ?

Technical perspective

End user perspective Organizational perspective

Prices, tariffs

Revenues, benefits

Costs, investments

Risks

What are viable financial arrangements ?

Roles and actors

Relationship to oneanother

Influences and power

Which actors are necessary for realising this service ?

Service description

Value proposition

Assumed customer value

Context

Is there a need for this service ?

Financial perspective

Necessary technicalfunctions

Alternatives forarchitecture

How can I realise the newservice technically ?

Technical perspective



From a mobile network operator’sperspective, the success of broadbandmobile services is highly dependent onthe ready availability and acceptance ofattractive applications and content,which the end-user is willing to pay for.We have produced an intermediary facil-itating platform that is positioned for usein a business-to-business (B2B) setting,between mobile operators on the onehand, and content providers on the other(see Figure 1).

Uluru is a platform that hosts 3rd partyapplication logic and makes commonservices available to these applications.These services can potentially be quitecomplex and may also require a certain'trust relationship' between the specificservice and the application using it (egpayment through a mobile operator’sbilling system). Such trust generallyrequires contracts/agreements (SLAs)between the mobile operator and partiesthat wish to charge using the operator’sbilling infrastructure, which can be diffi-cult and expensive to arrange.

Uluru provides a set of coarse-grainedfunctions to the applications that mayuse these through well-defined APIs (seeFigure 2). These functions fall into thefollowing broad categories:• content management: handling content

and its meta-data• financial exploitation: accounting,

charging, payment and billing• personalisation (content recommenda-

tion): an engine that can be used tooptimise application usability

• authentication: used for controllingaccess of users and applications

• logging: system monitoring• event messaging: allowing asyn-

chronous messaging between applica-tions and system components

• subscription: managing user details• MMS and SMS office: services that

can send and receive text-messagesfrom mobile phones.

The Content Management ServiceProvisioning caters for the intake of thecontent feeds and performs encoding andstorage of these feeds. After transcoding(eg into MPEG-4) and enrichment of thecontent (eg making it fit for dedicatedapplications), the content is again storedand ready to deliver to the mobilenetwork.

End-users of Uluru’s hosted applicationscan be charged for the services theyconsume using existing mobile operatorbilling services through a Parlay(http://www.parlay.org) interface. Thisleverages the pre-existing customerbilling relationship between the end-userand the mobile operator.

For personalisation we have developedthe Duine Toolkit, which is an SDK thatallows developers to integrate predictionengines into their own applications. Incontrast to most current availablepersonalized information systems thatfocus on the use of a single predictiontechnique or a fixed combination of twoor three techniques, Duine combinesmultiple techniques in a dynamic andintelligent way, thereby providing moreaccurate and stable predictions.

Each of the above-mentioned platformfunctions are exposed to the hostedapplications through the Uluru platformAPI, and are implemented by 'backend'components using an adapter approach.This adapter approach for decoupling theCOTS (Commercial Off The Shelf)

by Johan de Heer, Andrew Tokmakoff, Henk Eertink and John Anijs

The ULURU project, initiated in 2001 by the Dutch Telematica Instituut andMobiview, has its focus on aspects that play a primary role in the value chainbetween content providers, mobile network operators and end-users. Researchhas been performed in personalization, payments, digital rights management,content management and content distribution.

Uluru: Mobile Interactive MultimediaExperimental Service Environment

Figure 1: Uluru: Roles and Services.

Figure 2: Uluru’s functions.

AccessNetworkproviderULURU

ContentMgt System

DigitalRights Mgt

Persona-lization

Application

Messaging

Streaming

WebDistribution

Contentprovider

Content

Accounting

Application provider

cont

ract



components from the platform avoidsvendor lock-in and insulates the platformfrom the underlying technologies.Furthermore, Uluru is communicationtechnology independent; its onlyassumption is that IP network connec-tivity is available. This means that theplatform can operate using network tech-nologies including 802.11x, GPRS, andUMTS. The implementation of theUluru platform is heavily based upondistributed Java technologies such asJ2EE and the platform itself isdistributed over multiple locations. Inone example, a NewsFlash applicationwas located at the Telematica Instituut,whilst the Uluru platform itself waslocated at Hilversum, the MediaPark. Inthis situation, the NewsFlash applicationutilized CORBA to access the Platformservices.

In order to provide good insight into theunderlying workings of the Platform, wehave developed a visualization toolkit,which allows system events to bedepicted on a graphical representation ofthe system. The visualiser (shown inFigure 3), depicts the interaction betweenplatform components (such as profiling,content recommendation, contentdelivery and financial exploitation), basedon input of real-time triggers (events) thatoccur during the process of deliveringinteractive multi-media services.

All that is needed of the developer is toinclude some “metering points” withinthe application and to make use of thepackage’s event reporting functionality.The implementation of the EventServerinterface is based on the Pushlets frame-work, in which a servlet or JSP streamslines of JavaScript code into a hiddenbrowser frame. The browser executes the

lines and performs visualization actionsaccordingly. In effect, this mechanismprovides a callback from Java in theserver to JavaScript in the clientbrowser. The client-side graphics havebeen implemented using Scalable VectorGraphics.

The Uluru platform provides the infras-tructure to deliver a wide range of multi-media content and interactive services tomobile end-users (subscribers). TheNewsFlash application demonstrates thecapabilities of the ULURU platform,including the possibility to offer person-alized content and to apply flexiblecharging for various kinds of content,which are regarded essential forsuccessful business (see Figure 4).NewsFlash offers the latest news in text,image and streaming video format. TheULURU platform dynamically adaptsthe news content offered to the userbased on preset preferences, previouslyrequested news articles and a calculateduser interest rating for news articles.

Another major programme goal, theimplementation of an experimental plat-form for mobile multi-media services,has now been achieved. The ULURUplatform is now available for user trials,which offers external parties, (such ascontent providers and mobile networkoperators) the ability to perform tests onapplications providing next generationmobile communication services. TheULURU platform will be available foracademic and pre-competitive user trialsat the Telematica Instituut site before theend of summer 2003. If you would liketo obtain access to Uluru’s platformfeatures, please contact us.

Links:Telematica Instituut: http://www.telin.nl

Mobiview: http://www.mobiview.nl

Uluru project: http://uluru.telin.nl/

Toolikit by Telematica Instituut:http://uluru.telin.nl

Event Visualiser by Telematica Instituut: http://uluru.telin.nl

Please contact:Johan de Heer, Andrew Tokmakoff, Henk Eertink, John Anijs, Telematica Instituut,The NetherlandsE-mail: {Johan.deHeer;Andrew.Tokmakoff;Henk.Eertink}@telin.nl; [email protected]

Figure 3: Visualization

of System Functions.

Figure 4: Personalised Teletext news for i-mode (left), and Sony/Ericsson P800 (right),

and NewsFlash with ANP content for IPAQ (middle).



Public Travel Information Systems(PTIS) are widely used in train, tramwayand subway stations. However, the inte-gration of mobile devices (such as PDAsfor personalised traveller information oronboard displays in trains) is difficultdue to the heterogeneity of these devicesand the fact that the status of connec-tivity changes rapidly (from high avail-ability of bandwidth to off-line mode).Moreover, configuration and manage-ment of these devices and the distribu-tion and delivery of content - increas-ingly of a multimedia nature - are diffi-cult issues.

During our work we have found anumber of requirements for middlewaresystems dealing with stationary andmobile devices in public informationsystems. The most important of theseare:• Heterogeneity of devices:PTIS must

cope with a multitude of devices,including personal computers, publicinformation terminals, videowalls,LED displays and also mobile devices,such as PDAs (Personal DigitalAssistants) and mobile phones.

• Variability in connection modes - theoff-line mode: The use of mobiledevices implies that not all deviceswill be available all the time; in mobileenvironments, devices are regularlyoff-line, and as a result, the off-linemode cannot be considered to be anexception. Consequently, the middle-ware infrastructure must be capable ofcaching data in the off-line mode andredistributing it upon reconnectionthrough the use of sophisticated datasynchronisation mechanisms.

• Centralised configuration and 'Plug &Work' capability: Another issueinvolves the accessibility of mobiledevices in large systems for configura-tion and management purposes. For

example, a huge number of informa-tion displays in train stations cannot bephysically accessed. Consequently,the configuration data for thesedevices should be minimised, and ingeneral they should be configured,maintained and monitored through acentral management tool. Using acentral management tool ('Control

Center') with automatic data distribu-tion allows the easy maintenance of ahuge number of devices with similarconfigurations. This also allows theso-called 'Plug & Work', ie thedynamic configuration of devicesduring run-time.

• Quality of information: The quality asperceived by the customer, ie traveller,is measured by the quality of the infor-mation delivered. This requires infor-mation to be synchronised and updatedregularly.

Under consideration of the requirementsand aims mentioned above, we haveadopted the following technical solu-tions. The middleware infrastructure isbased on an overlay network structure

with a hierarchical, two-tier networkscheme. This means that, despite theunderlying physical network, mobiledevices are dynamically assigned tological sub-domains within one globaldomain, enabling efficient devicemanagement, message routing and allo-cation of the necessary infrastructure. Inorder to provide a simple and efficient

programming model for developers,software infrastructure on devices had tobe structured appropriately. The conceptof a service-oriented architecture (SOA),defining services as coarse-grained func-tional units with well-known interfaces,was therefore adopted. Following therequirements for mobile devices,services are allowed to use differentcommunication modes like synchronousRPC and asynchronous 'Store andForward' messaging. Reliable datadistributed to mostly off-line devicesrequired the development of a data-distribution mechanism called VirtualDatabase (VDb). With Virtual Databaseit is possible to provide services with avirtual database connection withoutestablishing a real connection. VDb

by Karl Rehrl, Harald Rieser and Siegfried Reich

The Sun Technology Research Excellence Center of Salzburg Research and theAscom Center of Competence are jointly working on a next generation middlewarethat will allow the integration of mobile devices in Public Travel InformationSystems.

Next Generation Public Travel Information Systems

Public Travel

Information

Systems in Graz.

©S

alzb

urg

Res

earc

h



Even though this process is, in principle,repeated every night, varying distur-bances and problems can occur eachtime. By utilising mobile communica-tion, people working in different loca-tions can be reached, and problemssolved. It also enables reporting of up-to-the-minute information on the entireproduction process, and communicationbetween supervision and employees aswell aswithin a team. With the help ofreal-time control and tracking, the entiredelivery process can be optimised andthe quality of the process improved. Pilotsystems and solutions were developed toenable work-related communication innewspaper distribution organisations,and experience was gained in practicalfield tests. The tests made clear whatkind ofbenefits can be offered by the useof these new tools, as well as showinghow currently available technology isable to meet the challenges of profes-sional usage in a real working environ-ment.

At present, the flow of electronic infor-mation in newspaper production ends at

the post-press department. Tailoring ofnewspapers, new subscription alterna-tives, new products and predictedshortage of deliverers will increase thechallenges for accurate and reliabledistribution. In practice, distribution willevolve, and is predicted to function moreand more through teamwork, whereteams will be autonomous and the needfor supervision will decrease. Thesechallenges create a demand for newsystems of communication and transmis-sion of data.

The objective of the mobile communica-tion pilots was to plan and implementmobile control and tracking in order toobtain empirical information on the suit-ability of the system and the portablewireless devices used. The pilot systemincluded following functions:• on-time and deviation reporting to the

distribution organisation• instant messaging to deliverers• communication between and among

the foreman and the deliverers.

The preliminary requirements specifica-tion was made based on interviews withdeliverers, their foremen, van driversand management from several distribu-tion companies. In particular, opinionsconcerning portable devices and featuresof the system were gathered. Typicalworking methods and the need for bilat-eral communication were also examined.At the pilot stage, the operation of thepilot system was first presented to thepilot companies and deliverers, and feed-back was gathered accordingly.

Web-application architecture waschosen since individual applications todifferent terminals, as well as the controlof the state of the application, wouldhave been troublesome, the latterresulting from at the time not so reliableGPRS-connection. The application canbe run on any general Web browser.Typical layers in such architecture areclient application, www-server anddatabase. Web browsers served as theclient application at the different termi-nals. HTML pages were created dynami-

by Olli Kuusisto and Janne Pajukanta

Newspaper distribution and transport are time-critical operations in which a largenumber of people are required to work in a coordinated manner. VTT has developeda pilot system based on mobile communication to solve production andcommunication problems in the newspaper distribution and transport process.

Mobile Communication in Newspaper Distribution and Transport

hides the necessary communicationdetails from the developer and improvesthe efficiency of data distribution.Services ask the VDb for database frag-ments and get these either from the localcache or upon network connectiondirectly from the database. VDb is usedfor the distribution of configuration dataand management data, and is also wellsuited for multimedia content.

Another of our technical aims is thesupport of a broad variety of heteroge-neous devices. This was achievedthrough the use of widely adopted stan-

dards such as Web Services, and throughbuilding upon platforms which are alsoavailable for embedded devices. Whileour first prototype was based on the Java2 platform, requirement changesdemanded that the current prototype bedeveloped using the .net Framework. Inthe future, we are planning to support the.net Compact Framework on embeddedWindows.CE devices as well.

In conclusion, we have argued forrequirements for next-generation trans-port information systems and haveoutlined how we are addressing these

issues with respect to the middlewarelayer. The concept of Service-OrientedArchitecture was carefully chosen inorder to overcome platform borders andto simplify the process of porting code toother platforms.

Link:http://www.suntrec.at

Please contact:Siegfried Reich, Salzburg Research Forschungsgesellschaft m.b.H./AARITTel: +43 662 2288 200E-mai: [email protected]



cally with help of Java Servlet tech-nology based on the database.

The pilots took place in 2002-2003. Thepilot system made communication andthe flow of information easier to managebetween deliverers and their foremen,and within the team of deliverers. It isexpected that the schedules of the entiredelivery process could be optimised and

the quality of the process improved withthe aid of real-time control and tracking.The users found the pilot system easy touse and navigate. Information from theproduction stage at the printing housewas useful and helped to plan the deliv-erers' work. The communication systemwas found to be useful, as there alreadyexists a high frequency of communica-tions between deliverers. At the moment

this is handled by cell phones, but itcould be more easily managed withPDAs.

Pilot-like real-time control and trackinggives accurate data on sources of distur-bance, which enables a fast reaction andincreases quality. Reporting of the startand stop times of delivery gives specificdata to foremen and to customer service.If the deliverer has not registered thestart time as scheduled, he/she can becontacted without delay. Traditionally itis necessary to either physically checkthe starting points of delivery, or to waituntil customers complain. Fast and accu-rate feedback can also be given toprevious working phases. Registering ofthe extra copies needed by deliverersenables better and more accurate controlof the edition and makes it easier totackle the causes for ordering extracopies. Also, an electronic controlsystem enables the optimisation of thedaily schedules of distribution, whichbrings savings in decreased waiting timecosts.

Link:http://www.vtt.fi/tte/projects/JALAN/

Please contact:Olli Kuusisto VTT Information TechnologyTel: +358 9 456 4760E-mail: [email protected]

Mobile applications are undoubtedly thenext wave in the evolution of e-business.Possessing features and functions thatare unique to mobile devices, such asmobility, personality, and flexibility,mobile applications are able to provideend-users’ added values, includinganytime, anywhere access, ability topinpoint users’ locations, and flexibility

in arranging tasks. It is predicted that themobile users will increase dramaticallyin the near future and the rate for mobileservices will drop significantly.However, the promising future of mobileapplications has been inhibited by theinfant stage of mobile applications, thedrawbacks of mobile devices, and thelimitations of mobile services.

Nevertheless, mobile applications aregaining attention both from academicsand practitioners. Understanding thevalues of mobile applications hasbecome particularly important. It wouldbe helpful to explore the values ofmobile applications from the end-users’perspectives (customers and company’semployees) and determine how the

What do end-users expect to have in mobile applications? What are the valuesof mobile applications that are important to end-users? How these values can beachieved? The Mobile Commerce Research team at the University of Nebraska-Lincoln examined the values of mobile applications using the 'Value-FocusedThinking' approach. This research developed means-ends objective networksdepicting how mobile applications would be able to achieve values expected byend-users.

Values of Mobile Applications to End-Usersby Keng Siau, Fiona Nah and Hong Sheng

Mobile communication pilot system overview.



values could be achieved, especially forcompanies embarking on implementingmobile applications or customers whoare embracing mobile applications.

To understand end-users’ perspectivesabout mobile applications, the best wayis to ask them. We conducted tworesearch projects to examine the valuesof mobile applications. One studies thevalues of mobile commerce tocustomers, whilst another one examinesthe values of mobile applications in alarge public utility company in the US.The same Value-Focused Thinkingapproach was used in both studies.Value-Focused Thinking approachallows us to explore the end-users’values and define relationships amongthe values — ie, how one objective canbe achieved by other objectives. Open-ended interviews were conducted tocollect values of mobile applications.The approach consists of four steps:1. Identify user wishes, concerns, prob-

lems, and values in mobile applica-tions.

2.Convert user input into objectives. Anobjective is something one wants tostrive toward.

3.Distinguish between fundamentalobjectives and means objectives. Ifone objective is important because itwill influence another objective, it is ameans objective. Otherwise, it is afundamental objective.

4.Build a means-ends objective networkdepicting specific relationshipsbetween means objectives and funda-mental objectives.

The results of the two studies presentedvalues of mobile applications from end-users’ perspectives and determine therelationships among the values. Some ofthe interesting findings include:• Convenience and efficiency are the

main advantages of mobile applica-tions both for customers andcompany’s employees. They are thetwo primary fundamental objectivesthat end-users expect mobile applica-tions to provide.

• Mobility is the basis for the values ofmobile applications. The ability to beused at anywhere, and at any place washighlighted by the interviewees as themain reason to adopt mobile applica-tions. It is also the main advantage

mobile applications have over tradi-tional personal computers. In a way, itis a 'necessary condition' for mobileapplications.

The researches also suggested someareas that need to be enhanced to ensurethe wide diffusion of mobile applica-tions. They are:• Limitations of mobile devices. Being

much smaller, mobile devices havesmall screen and complicated inputmechanism, and are considered to bemore difficult to use than personalcomputers. The low display resolutionand small display screen have inhib-ited information to be displayedcompletely and clearly. The limitedbattery life has restricted the mobilityof end-users.

• Mobile service quality. The technicalrestrictions have posed a challenge forthe wide diffusion of mobile applica-tions. Low bandwidth, unstable

connection, and limited coverage areaare the main drawbacks of currentmobile technology and services. In ourinterviews, these are listed as the mainobstacles to overcome when usingmobile applications. Although webelieve that with the advancement ofmobile technology, these problemswould be alleviated or resolved in thenear future, more research needs to beconducted at the current stage.

• Security options. Security is a majorconcern for mobile applications.Wireless transmission, in a way, biasesend-users to perceive mobile applica-tions to be more vulnerable and unse-cured. Thus, more security enhance-ments and options must be provided toensure the security of data and busi-ness transactions.

Our research not only shows the causalrelationships among the values of mobileapplications, but it also suggests areasfor further research and development.This research steam is ongoing and weare currently investigating the human-computer interaction aspects of mobiledevices and mobile applications.

Please contact:Keng SiauDepartment of ManagementUniversity of Nebraska-Lincoln, USATel: +1 402 472 3078E-mail: [email protected]

Further Reading

• Siau, K., Lim, E., Shen, Z., “MobileCommerce – Promises,Challenges, and ResearchAgenda,” Journal of DatabaseManagement, Vol. 12, No. 3, Jul-Sep, 2001, pp. 4-13.

• Siau, K., Shen, Z., “MobileCommerce Applications in SupplyChain Management,” Journal ofInternet Commerce, Vol. 1, No. 3,2002, pp. 3-14.

• Siau, K., Shen, Z., “BuildingCustomer Trust in MobileCommerce,” Communications ofthe ACM, Vol. 46, No. 4, 2003, pp.91-94.

• Siau, K., Shen, Z., “MobileCommunications and MobileServices,” International Journal ofMobile Communications, Vol. 1,Nos. 1/2, 2003, pp. 3-14.

• Krogstie, J., Lyytinen, K., Opdahl,A., Pernici, B., Siau, K.,Smolander, K., “Research Areasand Challenges for MobileInformation Systems,”International Journal of MobileCommunications (forthcoming).

• Siau, K., Sheng, H., Nah, F., Davis,S., “Trust in Mobile Business: AnEmpirical Investigation,”International Journal of ElectronicBusiness (forthcoming).

R&D

and

Tech

nolo

gy T

rans

fer


Articles in this Section53 Algorithmic Clustering of Music

by Rudi Cilibrasi, Paul Vitányi and Ronald de Wolf, CWI

54 Extension of Electronic Documents to enableServer-Independent Architecturesby Paola Carrara, Giuseppe Fresta, Daniela Poggioli,CNR

55 COSMOS: Safe and Optimal Operation in ComplexIndustrial Systemsby Péter Inzelt, SZTAKI

57 Leading-Edge Knowledge Management for PublicEmployeesby Simon Lambert, CCLRC

58 Radiosity for Realistic Real-Time Renderingby François Cuny, VSP Technology, France

59 Ultra-High-Speed Image Processing: Over 10 000 Frames Per Secondby Ákos Zarándy and Róbert Fekete, SZTAKI

60 Internet-Based Learning and Collaboration for Dispersed Co-Workers by Dian Tan and Peter Mambrey

62 Internet and Isolated Villages: An Opportunity for Developmentby Diego Puppin, ISTI-CNR

52




Recognition of similarities betweenmusic pieces is crucial for the design ofefficient music information retrievalsystems. The ever increasing wealth ofdigitized music on the Internet calls foran automated organisation of the mate-rial based on similarity, which enablesusers to navigate and which gives themadvice. Current methods of automated

classification utilize specific musicalfeatures related to pitch, rhythm,harmony, etc., just as a human expertcategorizes the material manually. Onecan extract such features using forinstance Fourier transforms or wavelettransforms, and classification is done byexisting software, based on various stan-dard statistical pattern recognition classi-fiers, Bayesian classifiers, hiddenMarkov models, ensembles of nearest-neighbour classifiers, or neuralnetworks. However, this approachrequires specific and detailed knowledge

of the problem area, since one needs toknow what features to look for.

Our approach does not require specificmusical knowledge, as it is based on ageneral mathematical theory of simi-larity. It uses a metric derived from thenotion of ’information distance’.Roughly speaking, two objects (repre-

sented as strings of bits) are close if wecan significantly compress one given theinformation in the other. Compression isbased on the ideal notion of Kolmogorovcomplexity, which unfortunately is noteffectively computable. Hence wereplace this ideal version by standardcompression techniques (the programbzip2). We lose theoretical optimality insome cases, but gain an efficientlycomputable similarity metric. This newmetric has been shown to work well invery different applications, including thefully automatic construction of the

phylogeny tree based on whole mito-chondrial genomes and of a languagetree for over 50 Euro-Asian languages,as well as the detection of plagiarism instudent programming assignments, andthe phylogeny of chain letters.

Now we have applied our compression-based method to the classification ofpieces of music, and performed variousexperiments on sets of mostly classicalpieces given as MIDI files. Wecomputed the distances between all pairsof pieces, and then built a tree (using theso-called quartet method) containingthose pieces in a way that is consistentwith those distances. After the successfulcompletion of three controlled experi-ments we turned to the real world. Themethod distinguishes quite well betweenvarious musical genres (classical, jazz,rock), and in the classical genre evenbetween composers — a remarkable factindeed, given the method’s completeignorance of music. The Figure showsthe result for a set of 12 classical pianopieces: the 4 movements from Debussy’sSuite Bergamasque, 4 parts from Bach’sWohltemperierte Klavier II, and 4preludes from Chopin’s op.28.

Future applications may include the useof the program as a data mining machineto discover hitherto unknown similaritiesbetween music pieces of differentcomposers or genres, and the selection ofa plausible composer for a newly discov-ered piece of music of which thecomposer is not known.

Please contact:Rudi Cilibrasi, Paul Vitányi, or Ronald de Wolf, CWITel: +31 20 592 9333.E-mail: {Rudi.Cilibrasi, Paul.Vitanyi,Ronald.de.Wolf} @cwi.nl

A fully automatic method for music classification has been developed at CWI. Itis based only on compression of strings that represent the music pieces. Themethod uses no background knowledge about music whatsoever: it is completelygeneral and can, without change, be used in different areas like linguisticclassification and genomics. It is based on an ideal theory of the informationcontent in individual objects (Kolmogorov complexity), information distance, anda similarity metric.

Algorithmic Clustering of Musicby Rudi Cilibrasi, Paul Vitányi and Ronald de Wolf

The result for a set of 12 classical piano pieces.



The two dialoguing units of traditionalWeb architecture are a data delivery unit,which is usually in charge of computingactivities, and a data display unit, whichrequests, receives and browses informa-tion. JavaTM opened the way to a newconcept of the Web, by sharing/distributing computation on both sides ofthe Web architecture, but it deliversbinary code which is not extensible andcan be only executed.

XML (eXtended Markup Language)makes it possible to exchange open andtextual information: It defines structuredtext files which can be manipulated byprogramming languages acting at anycomponent of the Internet architecture.XML tags can be freely defined to reflectthe semantics of any applicationuniverse and the structure of the docu-ments to be dealt with. One example isthe Geographic Markup Language(GML), defined by the OpenGISConsortium to provide a common modeland common format for the descriptionand expression of geographic entities.

Recent literature addresses the relation-ship between semantically-rich MarkupLanguages, such as GML or GraphML,and SVG (Scalable Vector Graphics),the W3C recommendation for vectorgraphics. Architectures have beenpresented in which information isdescribed and managed at the server sideby the language(s) suitable for the datamodel(s) under consideration, and isthen translated into a graphic-orientedlanguage at the browser side for visuali-sation and user interaction purposes.

These approaches maintain the tradi-tional asymmetry between the units ofthe Internet architectures, and make itimpossible to conceive application envi-ronments that can manage, manipulate or

even exchange information indifferentlyat either of the two logical units of thenetwork. To overcome this situation, wepropose to evolve the concept of elec-tronic document to an eXtendedElectronic Document (XED), which isan information aggregation gatheringtogether data items, templates for theirrepresentation, and tools for theirdescription/manipulation.

An XED document is an XML-docu-ment consisting of: • a set of fragments describing data to be

exchanged (for example GML frag-ments)

• a set of fragments describing transfor-mation rules

• a set of SVG fragments (or otherpresentation/interaction languagessuch as: XHTML, VRML, etc.) speci-fying the representation templates

• a set of tools (typically ECMAscriptcode) for the user.

Original data/template fragments can bestored in different repositories, and becompliant with various data models,formats, languages, etc. They can be

enriched by meta-information to accountfor different aspects such as semantics,quality, validity, languages, access rules,etc. The XED can be exchanged betweendifferent Internet components, trans-formed and manipulated at any networknode; the entire XED, or its components,can be searched for retrieval or filteringpurposes, and aggregation can thus bedynamic. An XED should contain all theinstructions necessary for its rendering,manipulation and/or storage, by thepreferred Internet data displayer such asa Web browser, a mobile or asound/voice rendering device.

Our proposal strengthens the possibilityof reusing data in different situations andpurposes as we do not conceive XEDs as‘static’ entities, created by the designer.For example, one ‘primitive’ XED canbe used both by a well-defined group ofend users (such as Italian students usinga Web browser) and intermediate userswho have to enrich it with more informa-tion (for instance, translators who mustadd the English version of the data orprogrammers who must unite transfor-

An eXtended Electronic Document (XED) is an XML document which aggregatesdata, templates, transformation rules and manipulation tools. It can be used todesign complex systems on the Internet in a server-independent way, modifyingthe traditional asymmetry between the logical units of Web architecture.

Extension of Electronic Documents to enable Server-Independent Architecturesby Paola Carrara, Giuseppe Fresta, Daniela Poggioli

Web-based GIS system architecture with browser activities and delivery of eXtended

Electronic Documents.

Query Processing

Atomic queries +aggregators

Results lists(GML)

Data display unit withSVG player

SIRS query interface

Aggreg. results (SVG) Query

User

XEDs

Data delivery unit

Archived docs(GML)

HTTPProtocol

XSLT

Query examinationenvironment.

XEDs

Matching

Resultscomputa-tion

Query forumulation.environment.



mations and rules to allow visualisationon a PDA).

In this way, it is possible to designcomplex systems on the Internet aschains of computing activities onspecific XEDs: each activity transformsthe document (the data, the templatesand/or the tools) and prepares it for asuccessive activity to be performed byother users at any Internet component.The designer can statically assign thedifferent activities on the basis of criteriasuch as performance, data dimensions,available hardware platforms, etc.; adynamic assignment can be also madeon the basis of the document status, userprofiles, etc.

We tested these ideas, taking as casestudy, an Information Retrieval systemfor geographic information to be imple-mented on the Web. The system designis performed through distinct steps, ie:

a)analysis of the computational activitiesto be performed, and the data andinteraction they require;

b)identification of the criteria guidingtheir assignment to the two logicalunits of the Web architecture;

c)identification of the users (not only theend users but also other users whoparticipate, with different roles, in thesystem creation, tuning, maintenance,etc.), and of the devices they prefer, inorder to identify the necessary XEDcomponents and/or the chain ofactions to create evolved XEDs fromprimitive ones.

Figure 1 illustrates the result of steps a)and b), ie the activities we identified(only macro-activities are shown), andthe assignment to the units where userscan formulate queries and look forresults. The criterion adopted was thelevel of interaction with the users: someactivities, such as query processing,matching between queries and docu-

ments, and the first aggregation of resultlist(s) are delegated to the data deliveryunit, as they do not require direct usermanipulation. At this step, data aredescribed by GML. The data display andinteraction activities (mainly queryformulation and results examination) areincluded in the display unit and areexecuted directly on the transformedSVG version of the documents.Therefore the architecture also includesenvironments to perform the languagetransformations. The XEDs, deliveredbetween the units by a plain HTTPprotocol, are not limited to the GML datadescription but also contain templatesand code to guide presentation, and thenecessary interaction tools.

Please contact:Paola Carrara or Daniela Poggioli, IREA-CNRE-mail: {carrara.p, poggioli.d}@irea.cnr.it

Giuseppe Fresta, ISTI-CNRE-mail: [email protected]

Safety and optimality of operation aredecisive factors for large and complexindustrial systems due to theircomplexity and the uncertainty associ-ated with their functioning. The projectCOSMOS aims to develop a novel infor-mation technology applicable incomplex industrial systems for processparameter monitoring, early detectionand diagnosis of equipment and tech-nology faults, and fast detection ofmalfunction roots.

The project partners include SZTAKI,the Atomic Energy Research Institute atthe Hungarian Academy of Sciences(MTA KFKI), the Department ofControl Engineering and Information

Technology and the Department ofMeasurement and Information Systemsat the Budapest University ofTechnology and Economics, PaksNuclear Power Plant Ltd, and MOLHungarian Oil and Gas Plc.

The SystemThe technology to be developed willoffer operational support for processdevices through advice given to the oper-ator. It will present user-friendly andefficient interactive tools for storingtechnology-related knowledge, on-linemodeling of process phenomena, visual-isation of diagnostics information, andthe on-line display of advisory documen-

tation to be used during fault-correctiveactions.

The kernel of the software system isrepresented by a central database, with astatic part consisting of the followingthree main elements: • the 'Object Structure Definition

(OSD)', which describes the featuresand attributes of the different objectshandled by the system

• the 'Abstract Object and KnowledgeRepository (AOR)', which acts as therepository for the objects described inthe OSD

• the external databases, handled in theiroriginal form, but using standard inter-faces.

COSMOS, a three-year project within the framework of the National Research andDevelopment Programs, commenced in the middle of 2001 with the aim ofdeveloping a novel, knowledge-intensive technology for safe and optimal operationin large and complex industrial systems. The project, led by SZTAKI, is carriedout by a consortium consisting of both academia and industry.

COSMOS: Safe and Optimal Operation in Complex Industrial Systemsby Péter Inzelt



The running programs perform theiroperations over an instant (fast) database(IDB), which constitutes the dynamicpart of the central database. An IDB canbe formed only from the objects stored inthe abstract repository and it alwayscontains as many objects as are requiredby the processing modules running at thetime.

The modules are used for establishing astructured storage for the engineering-technical information concerning thecomponents of the complex industrialsystem, and for the preliminary analysisof the safety, operational and ergonomicconsequences of the planned reconstruc-tions and changes in the applied tech-nology.

The 'Configuration and DocumentHandling Module' is used for storingbasic, technically relevant informationconcerning the different objects (build-ings, technological systems, equipmentetc). For documents directly connectedto the operation of the plant, the systemensures access to the on-line database ofthe process information system, in orderto retrieve and display the processparameters (measured or calculated

signals) which are referenced in thedocument.

The 'Virtual Reality Module' isconnected to VRML models convertedfrom the 3D CAD drawings of the roomsand the equipment, and is used forperforming the following basic tasks:• the VR Training Componentmodels

potentially dangerous activities, trains

work operations, and plans the optimalwork

• the VR Design and ValidationComponentis used for planning andmodeling plant modifications andreconstructions in an optimal way, andfor checking their conformance to thestandards and requirements

• the VR Decommissioning Componentis used for planning and modeling thedismantling of hazardous facilities andequipment.

Monitoring the technological process,the 'Early Fault-Detection Module' iden-tifies deviations from the normal stateusing the principle of analytic redun-dancy, ie, process measurements and on-line calculations based on the compo-nent’s model.

The 'Fault Diagnostic Module' identifiesthe root causes of equipment faults orsystem malfunctions. Having the faulttrees, which are defined by an interactivefault-tree editor, the reliability of theminimal cuts can be determined. Thisinformation can be used for identifyingthe most 'fault-sensitive' components,and for determining the relative proba-bilities of root causes creating specificfault symptoms. For large subsystems,the root causes are determined by an on-line fault-tree evaluation procedure.

The actual operational conditions fortechnological subsystems and equipmentefficiencies are determined by the on-line 'Process Optimisation Module',based on on-line process measurementsand background calculations. Using thismodule, the experts can assess the posi-tion of the actual working points withrespect to the optimal states.

In its final form, the developed tech-nology will constitute a general-purpose,open-architecture software shellproviding the operators of various indus-trial systems with efficient support. Itwill realise the following principle: anindustrial system should be able tooperate safely while simultaneouslyworking close to its optimum capacity.

Results and ApplicationsThe results will be implemented at theNuclear Power Plant in Paks, firstly forexperimental use at the central supervi-sory computer, and later at the supervi-sory systems of the nuclear block.Another implementation will be used byMOL Ltd, in the supervisory system ofthe network of high-pressure natural-gaspipelines in Hungary. According to theschedule of the project, the developmentof the prototype database, prototypeinterfaces, prototype functional modulesand final interfaces has been success-fully finished.

Link: http://www.sztaki.hu/sztaki/projects/html/cosmos.jhtml

Please contact:Péter Inzelt, SZTAKITel: +36 1 279 6184E-mail: [email protected]

The Nuclear Power Plant in Paks, one of the two application sites.

cou

rtes

y b

y N

PP



Public administrations carry out a hugevariety of tasks in the life of theircountry, but despite this variety theyoften have certain things in common intheir working practices. A frequentphenomenon is ‘organisationalmobility’, in which employees movefrom one department or unit to another,taking some of their acquired knowledgewith them, but at the same time needingto acquire more knowledge about thenew working environment. This is whatsets the experienced member of staffapart from the newly arrived one, and isa classic knowledge managementproblem.

The Pellucid project has chosen to useautonomous software agents as the basisfor its customisable platform. Thisapproach is extremely well suited to acomplex problem such as knowledgemanagement. Different classes of agentshave distinct responsibilities in seekingand presenting timely and appropriateinformation to the user. The Role Agentis responsible for assessing the user’sneed based on the context in the workprocess; the Task Agent and InformationSearch Agent are responsible forsearching for information to satisfy thatneed; while the Personal Assistant Agentdeals with presenting it in a personalisedway to the user. A Capitalisation Agenthas the job of building new knowledgefrom the organisational memory: forexample, using data mining to detectpatterns of usage of documents inprocesses.

The initial focus of the knowledgemanagement offered by Pellucid is ondocument management, contact manage-ment and critical timing management.Knowledge in these areas can be key toeffective and efficient working in manyroutine and non-routine tasks: for

example, where to find documents thatcan be used as models for a new case;who to contact in external organisations,or internally someone who has done asimilar task before; how long to expectcertain interactions to take. The threepilot applications within the project,though very different in character, allexhibit these three needs for knowledgemanagement. The pilot applications are:• Comune di Genova (Italy), Traffic and

Mobility Directorate: planning andinstallation of traffic lights in the city;

• Mancomunidad de Municipios delBajo Guadalquivir (an association oflocal governments in the south ofSpain): management of projects andservices;

• Consejería de la Presidencia, Junta deAndalucía (regional government in thesouth of Spain): management andresolution of problems with fixed tele-phony systems.

Underlying the operation of the Pellucidagents is a collection of ontologies,representing the structure and relation-ships of aspects of organisations andprocesses. These ontologies allow theagents to communicate together and toperform reasoning. A generic ontologyplus sub-ontologies for documents andcontacts have been developed. A keyline of research for the remainder of theproject will be the use of domain-specific ontologies and reasoning toallow retrieval and suggestions to theuser based on domain-specific know-ledge—for example, knowledge aboutsimilarities of traffic light installationsbased on their properties (type of roadjunction, traffic density, …). The projectaims to develop a generic, customisableplatform, and the separation betweengeneric and domain-specific parts willbe of great importance.

Implementation is being conductedusing the Java-based agent toolkit JADE.The possibility of using JESS (JavaExpert Systems Shell) is being exam-ined. Java Data Objects, an object-oriented interface to data storage, hasbeen chosen as the basis of the organisa-tional memory. Interfaces are beingprovided to workflow managementsystems that are compliant with theWorkflow Management Coalition stan-dards. In addition, the concept of a‘workflow tracking system’ is beingintroduced, to allow a lightweight provi-sion of context information to thePellucid system in cases where a fullworkflow management system is notneeded or desired.

The project is funded under theEuropean Commission’s ISTprogramme, and began in March 2002.There are eight partners from acrossEurope, including three local or regionalgovernment bodies as end-users, and twomembers of ERCIM: CCLRC and theSlovak Academy of Sciences Institute ofInformatics (part of SRCIM). Theproject is currently (May 2003)preparing for the preliminary user evalu-ations at two of the pilot sites.

Link: http://www.sadiel.es/Europa/pellucid/

Please contact:Simon Lambert, CCLRCTel: +44 1235 44 5716E-mail: [email protected]

The English word ‘pellucid’ means ‘transparent or translucent; extremely clearin style and meaning’ (Collins English Dictionary). That describes the vision ofthe Pellucid project in aiming to create a software platform for knowledgemanagement for public employees.

Leading-Edge Knowledge Management for Public Employeesby Simon Lambert

58


Radiosity and ray tracing are twomethods to simulate global lightingeffects. Although diffuse inter reflec-tions are crucial to produce realisticimages, radiosity-like methods arerarely used in production rendering:slow performance, robustness prob-lems, and control difficulty. VSPTechnology succeeded in resolvingall the industrial constraints.

Founded in 2001, based on researchconducted by INRIA, VSPTechnology develops and markets aportfolio of lighting simulation soft-ware products. With their technolo-gies, it is now possible to physicallysimulate and render the lightingeffect on a 3D scene in real time.

VSP Technology uses wavelet func-tions in its algorithm in order toresolve the global illuminationproblem. The use of wavelet func-tions ensures to have the most accu-rate approximation of the unknownradiosity function B(x). These recenthierarchical methods combined witha virtual mesh approach have greatlyspeeded up the computation ofradiosity. By carefully designing aparallel algorithm that minimizessynchronization and memory access,VSP Technology was able to obtainspeedups of 54 on a 64-processormachine.

Also, an important problem is to providean efficient solution to display thecomputed radiosity approximation. Botha meshing and a texturing technique givea solution of this problem.

VSP Technology uses a texturing tech-nique to store the photometric informa-tion. This technique consists in sepa-rating lighting information from thegeometry by storing the illuminationinformation as a texture map. This is

commonly called light map. At therendering step, this set of textures isblended with the initial geometry torender as accurately as possible thelighting information. Since graphicboards can render up to eight textures inonly one pass, this technique is actuallyopportune and allows 3D interactivemanipulation.

To ensure that the lighting information isdisplayed correctly, the mapping processmust avoid any overlap off the 3Dmodel. We are unfolding the modelusing a segmentation method to decom-pose the model into charts with naturalshapes. This method is based on a newoptimization-based parameterization

method: the least squares conformmaps (LSCM).

The choice of the LSCM parameteri-zation method is motivated by theproperties proved by its author (Levyet al.):

Their conformal criterion minimizesangle deformations, which allowsoften finding a valid parameteriza-tion. The parameterization is fastenough to be applied recursivelyuntil a valid and low distorted map isfound. The natural border extrapola-tion avoids the distortions implied bythe fixed border of other methods.And the result is independent of theresolution of the mesh. This propertyreduces texture swimming whenapplied on a multi-resolution struc-ture such as progressive mesh.

In order to avoid overlapping trian-gles in parameterization space, anextra segmentation is computed.Combined with a validation step, oursegmentation method ensures to finda valid solution with a controlleddistortion.

The result of our wavelet-based radiositycomputation is an illumination functionfrom a 2D space (the parameterizationspace, also called virtual mesh) to a 3Dspace (colour space) defined on a localnormalized mapping of each meshelement. Two main steps drive to thesolution:• the first step of our algorithm approxi-

mates the radiosity function on theinitial geometry or its triangulation inthe parametric domain with coloursstored at the vertices

• the second step is the light map gener-ation to render the illumination func-tion on each mesh element. Then theset of radiosity textures can be mapped

What influences have different surfaces and materials on the impression of theobserver? How do different lighting situations influence the perception ofarchitecture or a design? In which areas is additional artificial illuminationrequired?

Radiosity for Realistic Real-Time Renderingby François Cuny

Multiple light maps generation on a same geometry.




Machine vision systems are widely usedin industry for quality assurance andprocess control, among other applica-tions. Such systems usually consist ofone or more cameras, a framegrabbercard, a PC, and the image-processingsoftware. However, compact 'smart'cameras are also appearing on themarket, which integrate the abovecomponents into a single, compact unitthat can autonomously perform certaintasks such as fault detection. In thesenovel devices, the image-processingalgorithms are most often performed bya digital signal processor (DSP), or as inour case, by a Cellular VisualMicroprocessor (CVM).

Until now, the automation of multipleprocesses, production steps, or productinspections was hindered by two mainobstacles:

• the algorithms to be utilised werecomputationally too intensivecompared to the capacity ofcomputers;

• often the inspection requirement of theprocesses required a much higher rateof sampling than the sensors andcameras were capable of handling.

These barriers have now been overcomewith the introduction of the CVM tech-nology. CVMs have the ability tocapture and process several thousandimages per second in real time, whilecommon digital vision systems usuallyprovide only one to two hundredprocessed images per second.

CVMs are essentially CMOS sensorswith an integrated image processor. The

image processor is a cellular neuralnetwork-based image-processing array,and consists of several identicalprocessor elements. When performingan operation on an image, an everydayprocessor (like that of a PC) performs theoperation on the pixels one by one. In theprocessor array of the CVM, every pixelof the picture is handled by a differentprocessor element, meaning the opera-tion can be performed simultaneouslyfor every pixel. The individual processorelements are also directly interconnectedwith their nearest neighbours, enablinghigh-speed local operations.

by Ákos Zarándy and Róbert Fekete

Cellular Visual Microprocessor (CVM) technology allows extremely high processingspeeds to be reached in industrial applications of machine vision systems. Theseapplications are being developed as part of a larger series of development projectscarried out by AnaLogic Computers Ltd, a spin-off company of SZTAKI’s Analogicaland Neural Computing Systems Laboratory.

Ultra-High-Speed Image Processing: Over 10 000 Frames Per Second

on the initial geometry like otherexisting textures.

VSP Technology has applied the genera-tion of radiosity textures to industrialmodels such as architectural and manu-facturer design industries. Our radiositylight maps can easily be integrated intoexisting rendering software and dramati-cally increase the realism of a virtualscene without slowing down therendering speed.

Furthermore, we can group light sourcestogether to generate a set of light maps:one light map for each group of light

sources. Then, the light maps can beblended together during the renderingstep, allowing driving each group oflight sources independently. This allowsto interact in real-time on the lightingdesign and to visualize the impact ofeach light source group in the scene, bycontrolling independently the intensityof each light group.

The implication of this technology ishuge: significant reduction of projectplanning, enormous cost-saving for thevisualization of projects, physicallyaccurate simulation results in combina-tion with breathtaking image quality.

This aspect of physical accuracy cannotbe overemphasized in the domain ofdesign: looks-as-if-real counterfeitsobtained by any means - and approachused in video and movies - are hardlyacceptable for serious design jobs.

Links:http://www.vsp-technology.fr/

Please contact:François Cuny, VSP Technology, FranceTel: +33 3 83 58 17 51E-mail : [email protected]

Figure 1: Spark plug and the detected discharge (gas and surface discharge).



Owing to the parallel architecture andthe local connectivity, the CVM canperform operations on the full image in afew microseconds. Its computingcapacity is the equivalent of 1012 digitaloperations per second. The CVM canstore programs like common computers,and can be programmed from commondevelopment environments, using eithernative or the well-known C and C++languages.

As an example, a CVM’s extremely highspeed is capable of detecting and classi-fying the millisecond-long discharges inthe spark plugs of cars (Fig.1). The high-quality time resolution also makes itpossible to inspect the build-up of thedischarges. The images of the figurewere captured by the first prototype ofthe CVM, a binary system with 20x22pixel resolution. The testing measure-ments of a new 128x128 pixel CVM,which is capable of capturing andprocessing greyscale images, arecurrently in their final phase (Fig.2).

Both of these CVM chips were designedin IMSE-CNM Seville, Spain.

Based on this novel technology, a widerange of applications in a number ofareas is being developed. CVMs can be

utilised in any context requiring highframe-rate and fast processing. Someexamples include: • monitoring of passenger position and

airbag control for the automotiveindustry, eg, determining if the airbagcan be opened without harming thepassenger

• quality control in the textile industry,eg, detecting faults in yarn andweaving, as well as knots, dirt etc

• terrain recognition and classificationfor autonomous navigation ofunmanned vehicles

• position detection for robot control.

Link:http://www.analogic-computers.com

Please contact:Ákos Zarándy and Róbert FeketeAnaLogic Computers Ltd., BudapestTel: +36 1 279 6131E-mail: [email protected]

Figure 2:

The 128x128 pixel Cellular Visual

Microprocessor and its testing board.

The concepts of learning and informa-tion sharing in organisations are rapidlychanging: from the idea of learning inadvance and the false belief that know-ledge can be stored in repositories, toself-organised information sharingwithin groups and the construction oforganisational ‘knowledge’ bases. Weconducted a longitudinal case study toinvestigate how a group of dispersed co-workers acquires a new technical plat-form for the purposes of sharing infor-mation and self-reliable organisationallearning. The daily use of technologyand its options evolved over time basedon the user’s voluntary activity, withoutdirect intervention. We explicitly

focused on the self-organisation of theseprocesses.

The users were trainers within an insur-ance company. They were concernedwith internal vocational training, and theappendant back office - 22 people intotal. The back office was located at thecompany's headquarters, while thetrainers were dispersed all over thecountry. Since the back office wasresponsible for the trainers’ assignments,a lively exchange between back officeand trainers is implied.

As a technical basis, we chose theBSCW system (Basic Support forCooperative Work), developed by the

Fraunhofer Institute FIT. This group-ware system facilitates the coordinationand accomplishment of group work inshared workspaces. Since it is web-basedit works independently of a specificoperating system and can be accessedfrom any computer with an internetconnection. The user can create foldersand various types of objects (charts,URLs etc), and download and maintainthem (version control, conversion).Object-specific information (eg who hasread or changed a document) can berecalled anytime by the members of theworkspace by a simple mouse click on asymbol next to that object. They canreceive a daily activity report that lists allactivities which occurred the day before,

The internet-based collaboration platform NetKM, developed at Fraunhofer FIT,supports self-organised learning and knowledge-sharing. It was established inan insurance company to assist a geographically distributed team in organisingits daily work practice.

Internet-Based Learning and Collaboration for Dispersed Co-Workersby Dian Tan and Peter Mambrey



and the system also offers flexible allo-cation of rights. Plenty of other function-alities are available. The BSCW systemwas modified to meet the requirementsof the insurance company, and was intro-duced using the metaphor 'NetworkedKnowledge Market' (NetKM). Onlyinvited and registered members couldenter this portal. Advised by thecompany, we developed a folder struc-ture consisting of nine main folders, egorganisation, presentations, teachingmaterials, FAQ or personal folders. Weprovided them with some material andall users could add their own material.Our aim was to support self-organisa-tion, rather than regulating the contentand ways of using the system. Thecustomary methods of communicationpersisted, and the NetKM complementedbut did not replace them.

EvaluationThe period of use started in June 2001with an introductory workshop andtraining to explain the handling andfunctionalities of the system. Our find-ings refer to the one-year period from thestart to July 2002, but the system is stillin use by the company. The exploratorystudy was evaluated using qualitativeand quantitative methods. All available

data on the use of the NetKM (eg dailyactivity reports) were documented andanalysed. Additionally, all users wereasked to answer some questions aboutthe NetKM. We also conducted inter-views with seven carefully chosenpeople.

Users were encouraged to establish asmany objects as they wanted within themain folders. In the first year thesetotalled 337 folders on different levelswithin the main folders; the folderscontained 1043 documents. The numberof new objects varies greatly over theperiod. Back-office employees, trainersand researchers all contributed materialto the NetKM. Interestingly, the datasuggest that trainers who left the depart-ment stored their 'knowledge' in theNetKM shortly before leaving. In thisway, important information was trans-mitted to colleagues and remained avail-able to the company.

The number of downloads (readingactivities) per month from the NetKM iseven more irregular. The data show thatthe number of downloads is obviouslynot dependent on the number of newdocuments. For example, there was onemonth (March 2002) when only two new

documents were uploaded, but 101downloads were registered. This may beexplained by the fact that this analysisaccounted only for the upload of newdocuments, not the update of existingdocuments. Documents that were oftenupdated contained current informationand were often downloaded by users.

The written and oral questioning broughtup some interesting findings. The usersmentioned plenty of things they appreci-ated about the NetKM and also someaspects that could be improved. Theyalso provided us with useful hints for thefurther development of mobile self-organised methods of knowledge-sharing in dispersed groups.

Links: http://www.fit.fraunhofer.dehttp://www.olvio.de

Please contact:Dian Tan, Fraunhofer FITE-mail: [email protected]

The main structrure

of the 'Networked

Knowledge Market'.



Information and TelecommunicationTechnologies (ICT) are tools that canboost development in emerging nations.In a developing country, ICTs can giveaccess to information resources that aredifficult or expensive to reach otherwise.Besides the relevant governments, inter-national agencies such as the UnitedNation Development Programme

(UNDP) and International Telecommu-nication Union (ITU) are working tobring forth the best technologies to givedeveloping countries a boost in theirdevelopment and to narrow the digitaldivide. Nonetheless, the lack of cheapInternet connection, or of electric power,can prevent the implementation ofdigital services in rural and isolatedcommunities. Simple technological solu-tions can be effectively deployed to facethese limitations: the transfer of solu-tions from research laboratories to the

villages can bring opportunities forgrowth.

Here, we want to highlight some of thecontributions of a recently publishedstudy on the use of ICT solutions to offera reliable and inexpensive Internetconnection to isolated villages inBhutan. The analysis was completed by

a group of four MIT students, and is to beproposed to the UNDP and to theGovernment of Bhutan for implementa-tion.

In our analysis, we quickly realized thataccess to the Internet in a developingcountry can be extremely costly. Variousstrategies to reduce the required band-width or the on-line time have beendeveloped. Offline Internet is a verypromising approach, explored, amongothers, by the TeK (Time equal toKnowledge) group at MIT. The idea is

simple: when browsing the Internet, auser will only have access to pageslocally stored in a big cache in the localdisk. If a page is not locally stored, arequest for the page will be queued, andcompleted the first time the machinegoes online.

But a second way is offered by TeK.When users are looking for some data onthe Net, they will fill a form with thenecessary keywords describing theirinterest, the same way they would dowith a common search engine. Thisinformation will be sent to a TeK serverwhen the machine goes online. Later on,when the machine goes online a secondtime, the compressed results will beretrieved from the TeK server, and theusers can navigate them, as they woulddo from the result page of any searchengine. All the results of previousqueries, and queries from other users, arekept stored, and they are available to allthe users. Over time, TeK will build alocal library with knowledge of interestto that community.

This system offers a series of clearadvantages, including reduced commu-nication costs, increased reliability,increased opportunities for policyenforcement and control. It can be usedas the main tool to offer a reliableconnection in the presence of unreliablepower supply or Internet connectivity.

When the village is online, a number ofinitiatives can be planned:

• Tele-education: The wealth of know-ledge available on the Internet can bean important element to supporteducation in small communities. High-quality syllabi and class materials arecheaply available on the Internet.Educational partnerships can be set up:

A recent study proposes a reliable and inexpensive solution for Internet accessin isolated villages in Bhutan. The results of the study are to be proposed to theUnited Nation Development Programme and to the Government of Bhutan forimplementation.

Internet and Isolated Villages: An Opportunity for Developmentby Diego Puppin

When the village is online, a number of initiatives can be planned.

cou

rtes

y o

f P

er K

. So

ren

sen



big schools in developed areas canoffer educational support to smallerschools and communities, by offeringremote assistance for grading andteaching. On the other hand, real prob-lems of the village can be brought toclasses for solutions.

• Tele-jobs:In the globalized worldeconomy, many opportunities for tele-jobs are present. The availability oflow-cost, English-speaking, skilledlabor, can be exploited by interna-tional companies to reduce costs.Customers’ data can be sent digitallyfrom abroad to the workers in thecommunity, be processed, and sentback. A small isolated village withsome computing capabilities couldboost its economy this way.

• Assisted health care: Data about anisolated population could be analyzedby a remote staff of doctors in a large

hospital. The presence of epidemicproblems or rare diseases could beobserved by expert doctors, who canwarn the local health facilities. Goodresults have been observed in India fora similar project, focused on eye-care.

• E-post: The opportunities for cheapand reliable communication areimportant even for small communities,where a piece of mail can save a dayotherwise spent traveling. Mailmessages can be scanned, sent digi-tally and printed at the destination,with big savings in time, money andenergy.

In the study mentioned above, wedescribe how the TeK software, alongwith other solutions (wireless connec-tion, local networks, peer-to-peer datatransfer, solar panels) can be used withinthe context of the five-year Bhutan

Education Master Plan to reduce thecosts of the introduction of Internetconnection to the schools. But more thanthat, we show how simple IT solutionscan be used to connect villages withlimited or no electricity, and limited orno Internet connection, to the informa-tion available on the Internet.

The opportunities offered by the Internetto developing communities are endless,and the means for implementation areavailable. ICTs can open the way to a de-materialized and more sustainable pathof growth for the world’s poor: it is theresponsibility of the research communityto direct its energy where it can be themost effective to the world population.

Please contact:Diego Puppin, ISTI-CNRTel: +39 050 315 3003E-mail: [email protected]

SCILAB is a scientific software packagedeveloped by INRIA and ENPC (ÉcoleNationale des Ponts et Chaussées).SCILAB is one of the numericalcomputing tools such as Matlab thattotally revolutionized research anddevelopment in engineering sometwenty years ago. It provides a powerfulopen computing environment for engi-neering and scientific applications.Distributed freely via the Internet since1994, SCILAB is currently being used ineducational and industrial environmentsaround the world.

SCILAB includes hundreds of mathe-matical functions with the possibility toadd interactively programs from variouslanguages (C, FORTRAN, C++…). Ithas sophisticated data structures(including lists, polynomials, rationalfunctions, linear systems…), an inter-preter and a high level programming

language. SCILAB has been conceivedto be an open system where the user candefine new data types and operations onthese data types by using overloadingmechanism.

A number of toolboxes are availablewith the system:• 2-D and 3-D graphics, animation• Linear algebra, sparse matrices• Polynomials and rational functions• Simulation: ODE solver (ODEPACK)

and DAE solver (DASSL)• Scicos: block-diagram simulator for

hybrid dynamic systems• Classic and robust control, LMI opti-

mization• Differentiable and non-differentiable

optimization• Signal processing• Metanet: graphs and networks• Parallel SCILAB using PVM• Statistics

• Interface with Computer Algebra(Maple, MuPAD)

• Interface with TCL/TK• a great number of contributions for

various domains.

SCILAB works on most UNIX systemsincluding Linux and on Windows9X/NT/2000/XP. It comes with sourcecode, on-line help and English usermanuals. Binary versions are available.

Today there are more than 10,000monthly downloads of SCILAB softwarefrom SCILAB web site. This success ismainly due to the fact that, by integratingnumerous contributions, SCILAB offerstools pertaining to scientific computing(simulation, graphics, modelling, control,etc) that are not only of interest toteachers and researchers, but alsoincreasingly of interest to persons whowork in industrial sectors such as the

Following the tremendous success of INRIA's Open Source Software for NumericalComputations, called SCILAB, a consortium has been created to guarantee thequality and perpetuation of the software and to promote Scilab as worldwidereference software in academia and industry.

SCILAB Consortium launchedby Claude Gomez


R&D AND TECHNOLOGY TRANSFER / EVENTS

automotive industry, aerospace, chem-istry, finance, etc. Moreover, SCILAB isused all over the world. Books and arti-cles already exist in French, English andChinese. Workshops and lessons tookplace in India, Africa and China. Since2002 an annual SCILAB contest is orga-nized in China.

SCILAB ConsortiumTo face this success, INRIA proposed tocreate a Consortium to group academicand industrial members all over theworld. The Consortium will guaranteethe quality and perpetuation of the soft-ware and turn SCILAB into worldwidereference software in academia andindustry. Members of the Consortiumcan contribute to the evolution ofSCILAB based on new needs. A scien-tific committee and a steering committeecomposed of elected Consortiummembers will decide on the directionsfor new developments and play the roleof SCILAB architect. The Consortium ishosted by INRIA.

The main objectives of SCILABConsortium are:• to group the community of developers

to turn SCILAB into a platform to inte-grate all scientific advances in scien-tific computations, simulation andautomatic control. So SCILAB will bea privileged link between companiesand new scientific trends and will relyon a large community of contributors.

• to group the community of users toturn SCILAB into quality software foroperational use as well in industry asin education and research.

• to provide resources for a dedicatedteam for the development of SCILAB.

The dedicated team has already beencreated as a 'Development Project-Team'at INRIA-Rocquencourt. It is managedby the chief technical officer of theconsortium and will be eventuallycomposed of 14 members. In addition tothe development of SCILAB, the teamwill promote SCILAB and animate andfederate the community of contributors.

INRIA committed to finance 5 people inthe team.

On May 16, the SCILAB Consortiumwas launched with an initial meeting ofits members under the chairmanship ofProfessor Alain Bensoussan, formerchairman of INRIA, CNES and Councilof ESA. The Consortium foundingmembers are French companies andacademics: AXS INGENIERIE, CEA,CNES, CRIL TECHNOLOGY,DASSAULT-AVIATION, EDF, ENPC,ESTEREL TECHNOLOGIES, INRIA,PSA Peugeot Citroën, RENAULT, andTHALES. They firmly intend to increasethe European and international visibilityby attracting new members abroad.

Links:http://www.scilab.orgNewsgroup: comp.soft-sys.math.scilab

Please contact:Claude Gomez, Chief Technical Officerof Scilab Consortium, INRIAE-mail: [email protected]

ERCIM-Sponsored Events

ERCIM continues sponsoring up to ten conferences,workshops and summer schools per year. The fundingfor all types of events is in the order of 2000 Euro.

ConferencesERCIM invites sponsorship proposals from establishedconferences with an international reputation, wheresubstantive overlap can be shown between the conferencetopic and ERCIM areas of activity. Typical cases wouldinclude annual conferences in computer science withinternational programme committees, substantialinternational participation, and proceedings published withan established international science publisher.

Workshops and Summer SchoolsERCIM sponsors workshops or summer schools (co-)organised by an ERCIM institute. The additional fundingprovided by ERCIM should be used to enhance theworkshop by, for example, increasing the number ofexternal speakers supported.

Forthcoming events Sponsored by ERCIM:• ECOOP – European Conference on Object-Oriented

Programming, Darmstadt, Germany, 21-25 July 2003

• ECDL2003 – 7th European Conference on Research andAdvanced Technology for Digital Libraries, Trondheim, Norway, 17-22 August 2003

• FM 2003 – 12th International Formal Methods EuropeSymposium, Pisa, Italy, 8-14 September 2003

• CP 2003 – 9th International Conference on Principles andPractice of Constraint Programming, Kinsale, Ireland, 29 September-3 October 2003

• WISE 2003 – 4th International Conference On WebInformation Systems Engineering,Rome 10-13 December 2003

• SOFSEM 2004 – The Conference on Current Trends inTheory and Practice of Informatics,Merin, Czech Republic, 24-30 January 2004

For detailed information about ERCIM's eventsponsorship programme, see: http://www.ercim.org/activity/sponsored.html


EVENTS

Beginning with the first InternationalWWW Conference in 1994, this presti-gious series of the International WorldWide Web Conference Committee alsoprovides a public forum for the WWWConsortium (W3C) through the annualW3C track.

The conference with more than 800attendees from 57 countries started witha day of tutorials and workshops andfollowed by a three-day technicalprogramme. The fifth day was a

'Developers Day.' The tutorials andworkshops provided in-depth looks atspecific areas of current interest. Thetechnical programme included refereedpaper presentations, alternate trackpresentations, plenary sessions, panelsand poster sessions. Developers Day wasdevoted to in-depth technical sessionsdesigned specifically for Web devel-opers.

The W3C presented three days ofpresentations on the Web, future Webbrowsers, W3C architectural principles,the XML family, Web services, theSemantic Web, new devices, and hori-zontal essentials.

A number of 602 submissions wasreceived to the refereed papers track thisyear, a 33% increase over the previousrecord. 77 papers were accepted(compared to 72 in the previous year) forthe proceedings and presentation at the

conference, an acceptance rate of 12,8%which makes this World Wide Webconference the most competitive ever.All papers were rigorously reviewed bythe technical program committee, whichwas divided into the following twelveareas:• Applications• Browsers and User Interfaces• Electronic Commerce• Hypermedia

• Languages• Mobility and Wireless Access• Multimedia• Performance and Reliability• Search and Data Mining• Security and Privacy• Semantic Web• Web Engineering

Several changes were made in the arealineup this year: Web Engineering as anew area, and the Electronic Commerceand Security were split into two tracks,Search and Data Mining, and SemanticWeb.

One of the main points of the conferencewas the harmonic symbiosis of the WebServices and Semantic Web in a longerrun.

The Conference Co-Chairs wereGusztáv Hencsey (SZTAKI) and BeboWhite (Stanford Linear AcceleratorCenter), and the Programme CommitteeCo-Chairs were Yih-Farn Robin Chen(AT&T Labs - Research), László Kovács(SZTAKI) and Steve Lawrence (Google,Inc.). The main sponsoring institutionsof the event were Ministry of Informaticsand Communications, HungarianNational Tourist Office, IBM Research,Alvarion, AT&T Labs, ERCIM, W3C,NIIF and VOA News, and the confer-ence partners were InternationalFederation for Information Processingand W3C.

Links:WWW2003: http://www2003.org/IW3C2: http://www.iw3c2.org/

Please contact:László Kovács, SZTAKI Tel: +36 1 279 6212E-mail: [email protected]

The Twelfth International World Wide Web Conference took place in Budapest,Hungary, 20-24 May 2003, organized by the International World Wide WebConference Committee (IW3C2) and SZTAKI.

WWW2003 — The Twelfth International WorldWide Web Conference in Hungary by László Kovács

Word Wide Web conference in Budapest.


ERCIM News is the magazine of ERCIM.Published quarterly, the newsletter reports onjoint actions of the ERCIM partners, and aimsto reflect the contribution made by ERCIM tothe European Community in InformationTechnology. Through short articles and newsitems, it provides a forum for the exchange ofinformation between the institutes and alsowith the wider scientific community. This issuehas a circulation of over 9000 copies.

Copyright NoticeAll authors, as identified in each article, retaincopyright of their work.

ERCIM News online edition is available athttp://www.ercim.org/publication/ERCIM_News/

ERCIM News is published by ERCIM EEIG, BP 93, F-06902 Sophia-Antipolis CedexTel: +33 4 9238 5010, E-mail: [email protected] 0926-4981

Director: Bernard LarrouturouCentral Editor:

Peter [email protected]

Local Editors:AARIT: Karin Hraby

[email protected]: Martin Prime

[email protected]: Michal Haindl

[email protected]: Annette Kik

[email protected]: Carol Peters

[email protected]: Yannis Pachoulakis

[email protected]: Michael Krapp

[email protected]: Patrik Hitzelberger

[email protected]: Bernard Hidoine

[email protected]: Truls Gjestland

[email protected]: Harry Rudin

[email protected]: Kersti Hedman

[email protected]: Gabriela Andrejkova

[email protected]: Erzsébet Csuhaj-Varjú

[email protected]: Ann McNamara

[email protected]: Pia-Maria Linden-Linna

[email protected]

Free subscriptionYou can subscribe to ERCIM News free ofcharge by: • sending e-mail to your local editor• contacting the ERCIM office (see address

above) • filling out the form at the ERCIM website at

http://www.ercim.org/

EVENTS

CALL FOR PAPERS

IZS — 2004 International ZurichSeminar on Communications

Zurich, Switzerland, 18-20February 2004

Original technical contributions aresolicited on topics including:Access:• Wide area mobile radio• Wireless local area networks• Wireless local loops• xDSL and cable modem technologiesTransmission:• Channel modeling• Applied information theory• Coding and modulation• Joint detection, decoding, and ... (you

name it)• MIMO wireless and space-time

coding• Ultra wide band• Antenna design• Fundamental hardware issuesNetworking:• Algorithms, protocols, and systems for

routing and switching• Ad-hoc and peer-to-peer networking

of all scales• High-mobility architectures and proto-

cols• Next-generation internet• Optical networks• Quality of service• Fault-tolerance and reliability• Security

As a novelty in 2004, roughly half thecontributions will be invited.

Due to its traditional format with at mosttwo parallel sessions, the IZS is an idealopportunity to learn from, and tocommunicate to, leading experts in areasbeyond one's own speciality. All papersshould therefore be presented with thiswide audience in mind. Authors ofaccepted papers will be asked to producea manuscript not exceeding 4 pages thatwill be published in the Proceedings. Allaccepted papers will be allowed twentyminutes for presentation.

Deadline for submissions: 15 September 2003.

More information: http://www.izs2004.ethz.ch

CALL FOR PARTICIPATION

FM 2003: 12th International FMESymposium

Pisa, Italy, 8-14 September 2003

FM 2003 is the twelfth in a series ofsymposia organised by Formal MethodsEurope, an independent association whosepurpose is to promote and support theindustrial use of formal methods forcomputer systems development. The FMEsymposia have been notably successful inbringing together an international andvaried community of users, researchersand developers of precise mathematicalmethods for software development.

The scope of FM 2003 includes allaspects of the use of formal methods forsoftware development, with particularattention to new application areas thatare becoming critical in the new genera-tion information society. The sympo-sium will attempt to cover a wide varietyof themes, ranging from fundamentaltheories to practical experiences. In addi-tion to the three-day symposium, FM2003 will host 7 Workshops, 8 Tutorialsand system and tool demo sessions.There will also be a special day dedi-cated to industrial concerns: theFME/CoLogNet Industry Day.

More information: http://fme03.isti.cnr.it/

CALL FOR PARTICIPATION

SIGCOMM 2003Karlsruhe Germany, 25-29 August 2003

This year, SIGCOMM broadens it tradi-tional scope into a data communicationsfestival. The SIGCOMM conferencecontains full papers on a variety oftopics, including peer-to-peer, overlays,and denial of service, as well as positionpaper and poster sessions. A set of one-day workshops are part of the festival:networking research methods, scope,education, network-I/O convergence,future directions in architecture, andrevisiting QoS. Tutorials will be held onnetwork security and self-similar traffic.

More information: http://www.acm.org/sigcomm/sigcomm2003

ERCIM News 54, July 2003 67

IN BRIEF

SARIT — Professor Ambros P. Speiserdied on May 10, 2003. He was one of thelast of the early pioneers of digitalcomputer design and an insightfulinitiator and staunch supporter of scien-tific research. Ambros Speiser was the

leading hardware designer for theconstruction of the legendary ERMETHcomputer built at the ETH (SwissFederal Institute of Technology) inZurich 1950 until 55, ie, during the yearswhen Konrad Zuse's Z4, was used atETH as the first digital computer at acontinental European university. TheERMETH was then used until 1963. TheETH later honored Speiser for this workby awarding him an honorary doctor’sdegree. Well aware of his talents,Ambros Speiser was chosen by IBM tofound its Zurich Research laboratory in1956; he was its director until 1966.Then BBC (Brown Boveri & Company,now ABB) asked him to found and leadits corporate research laboratory inBaden. This he did until his retirement in1987. Speiser was also a professor at theETH in Zurich and often assumed bothSwiss and international positions asleader and advisor. He was president ofIFIP, the International Federation forInformation Processing, from1965 until1968, and of the Swiss Academy ofEngineering Sciences. Ambros Speiserwas not only a researcher with thehighest aspirations but, in his retirement,added another talent to his repertoire —that of a most articulate and respectedscientific journalist.Carl August Zehnder, Professor of Computer Science, ETH Zürich

Stelios Orphanoudakis(FORTH) willbe the new President of ERCIM fromJanuary 2004. He succeeds Gerard vanOortmerssen(CWI), who was Presidentfrom 1998. Van Oortmerssen wasappointed director of TNO Telecom fromMay 2003. He was director of CWI foralmost twelve years. At CWI he will besucceeded by Jan Karel Lenstra whoworked previously at CWI (1969-1989),Eindhoven Technical University (1989-2002) and Georgia Institute ofTechnology in Atlanta, USA. Lenstracomes from a family of very renownedDutch mathematicians.

CCLRC — Rutherford AppletonLaboratory hosted a launch eventincluding a live launch feed from ESA atthe occasion of the Mars Express mission.

Mars Express was successfully launchedat 18:45 BST on 2 June 2003 fromBaikonur, Kazakstan. Mars Expresscomprises a number of essential compo-nents – the spacecraft and its instruments,the lander, a network of ground and dataprocessing stations, and the launcheritself. RAL has been responsible for thethermal design and analysis of Beagle 2,which transports the lander during itsdescent from Mars Express.

CNR — GÉANT, the pan-Europeanmulti-gigabit research network, andGARR, the Italian research andacademic network are interconnectedat 10Gbit/secfrom 6 May 2003. GARR(http://www.garr.it) is the first NationalResearch Network in Europe to beconnected at 10Gbit/sec to GÉANT(http://www.geant.net), replicating theevent of July 1989, when GARR was thefirst European network to be connectedover an international link at 2Mbit/sec

INRIA — 'Internatics ' is the recentlylaunched newsletter of INRIA's Office forEuropean and International Relations.Through short articles and news items, itreports on INRIA's involvement inEuropean and international activites.

Besides the sections dedcated to coopera-tions in Europe, the Americas, Asia,Africa and Middle East, Internatics regu-larly features topics related to theEuropean research policy, making it agood source of information for scientistsand research managers. Internatics ispublished quarterly. The printed edition ispublished in French, the online version isalso available in English and offers freesubscription to both printed and electronicversions. http://www-direction.inria.fr/interna-tional/internatics/

©R

oger

Nic

kel/U

nive

rsity

Zur

ich

between Bologna and Geneva. GÉANTenables European scientists to competeinternationally by providing them with aworld-class backbone that offers thebandwidth and the quality of servicerequired for research and developmentactivities at this level. It represents thebasis for the introduction of 'virtual labo-ratories' and 'virtual institutes' in Europe.This will be extremely beneficial for theItalian research community participatingin international projects in many domains.The services provided by GÉANT, willalso allow the implementation of apowerful network research infrastructureto support GRID applications.

Beagle 2.

Fraunhofer ICT GroupFriedrichstr. 6010117 Berlin, GermanyTel: +49 30 726 15 66 0, Fax: ++49 30 726 15 66 19http://www.iuk.fraunhofer.de

Consiglio Nazionale delle Ricerche, ISTI-CNRArea della Ricerca CNR di Pisa, Via G. Moruzzi 1, 56124 Pisa, ItalyTel: +39 050 315 2878, Fax: +39 050 315 2810http://www.isti.cnr.it/

Centrum voor Wiskunde en InformaticaKruislaan 413, NL-1098 SJ Amsterdam, The NetherlandsTel: +31 20 592 9333, Fax: +31 20 592 4199http://www.cwi.nl/

Institut National de Recherche en Informatique et en AutomatiqueB.P. 105, F-78153 Le Chesnay, FranceTel: +33 1 3963 5511, Fax: +33 1 3963 5330http://www.inria.fr/

Council for the Central Laboratory of the ResearchCouncils, Rutherford Appleton LaboratoryChilton, Didcot, Oxfordshire OX11 0QX, United KingdomTel: +44 1235 82 1900, Fax: +44 1235 44 5385http://www.cclrc.ac.uk/

Swedish Institute of Computer ScienceBox 1263, SE-164 29 Kista, SwedenTel: +46 8 633 1500, Fax: +46 8 751 72 30http://www.sics.se/

Technical Research Centre of FinlandP.O. Box 1200FIN-02044 VTT, FinlandTel:+358 9 456 6041, Fax :+358 9 456 6027http://www.vtt.fi/tte

Foundation for Research and Technology – HellasInstitute of Computer ScienceP.O. Box 1385, GR-71110 Heraklion, Crete, GreeceTel: +30 2810 39 16 00, Fax: +30 2810 39 16 01http://www.ics.forth.gr/FORTH

Czech Research Consortium for Informatics and MathematicsFI MU, Botanicka 68a, CZ-602 00 Brno, Czech RepublicTel: +420 2 688 4669, Fax: +420 2 688 4903http://www.utia.cas.cz/CRCIM/home.html

Swiss Association for Research in Information Technologyc/o Prof. Dr Alfred Strohmeier, EPFL-IC-LGL, CH-1015 Lausanne, SwitzerlandTel +41 21 693 4231, Fax +41 21 693 5079http://www.sarit.ch/

Slovak Research Consortium for Informatics andMathematics, Comenius University, Dept.of ComputerScience, Mlynska Dolina M, SK-84248 Bratislava, SlovakiaTel: +421 2 654 266 35, Fax: 421 2 654 270 41http://www.srcim.sk

Magyar Tudományos AkadémiaSzámítástechnikai és Automatizálási Kutató IntézetP.O. Box 63, H-1518 Budapest, HungaryTel: +36 1 279 6000, Fax: + 36 1 466 7503http://www.sztaki.hu/

Trinity CollegeDepartment of Computer Science,Dublin 2, IrelandTel: +353 1 608 1765, Fax: 353 1 677 2204http://www.cs.tcd.ie/ERCIM

Austrian Association for Research in ITc/o Österreichische Computer GesellschaftWollzeile 1-3, A-1010 Wien, AustriaTel: +43 1 512 02 35 0, Fax: +43 1 512 02 35 9http://www.aarit.at/

Norwegian University of Science and Technology Faculty of Information Technology, Mathematics andElectrical Engineering, N 7491 Trondheim, NorwayTel: +47 73 59 80 35, Fax: +47 73 59 36 28http://www.ntnu.no/

Fonds National de la Recherche20, Montee de la PetrusseL-2912 LuxembourgTel. +352 26 1925 31, Fax +352 26 1925 35http:// www.fnr.lu

ERCIM – The European Research Consortium for Informatics and Mathematics is an organisation

dedicated to the advancement of European research and development, in information technology

and applied mathematics. Its national member institutions aim to foster collaborative work within

the European research community and to increase co-operation with European industry.

Order Form

If you wish to subscribe to ERCIM News free of charge

or if you know of a colleague who would like toreceive regular copies of

ERCIM News, please fill in this form and wewill add you/them to the mailing list.

Send, fax or email this form to:ERCIM NEWS

Domaine de VoluceauRocquencourt

BP 105F-78153 Le Chesnay Cedex

Fax: +33 1 3963 5052 E-mail: [email protected]

Data from this form will be held on a computer database. By giving your email address, you allow ERCIM to send you email

Name:

Organisation/Company:

Address:

Post Code:

City:

Country

E-mail:

✄

You can also subscribe to ERCIM News and order back copies by filling out the form at the ERCIM website at http://www.ercim.org/publication/Ercim_News/

ERCIM is the European Host of the World Wide Web Consortium.