Upload
lucinda-m
View
214
Download
1
Embed Size (px)
Citation preview
Kathl K.81...LuclndaM .........
Managing Organizational Handoffswith Empowered Teams
High morale, strong teamwork, and proactive thinking (instead ofreactivethinking) areessential catalysts for the development ofproducts thatmeetor exceed customer expectations. Thetraditional organizational structureinto which development work is segmented can either impede or enhancethe ability ofdevelopers toform effective teams. Because oftheir usual hierarchical form and often rigid boundaries, organizational structures can alsoinhibit the development ofa strong sense ofownership, thereby stiflingcreative, proactive problem solving. This paper describes a way toformempowered teams around traditional organizational boundaries, Le., byorganizing the team into a subproject team. Thesubproject concept wasused effectively inthe development ofa hardware, firmware, and operatingsystem platform for a new PBX. We did notcreate theusual artificial deliveriesbetween hardware and software organizations through contractual obligations. Instead, we formed a subproject thatencompassed all the development work needed toformulate, test, and deploy the new PBX vehicle. Quality metrics data for the subproject showed anincrease incustomer satisfaction, high team morale, and a reduction innecessary rework.Introduction
Many development handoffs betweenorganizations are achieved viawritten contracts, in which the sending and receivingorganizations specify their delivery expectations. These contracts are appropriate at somehandoff points. However, webelieve that suchcontracts canalso have a downside. Bytheirverynature, contracts are often legalistic, andcancausefinger pointing andaccusationswhena certain clause is not met. This,in tum,can leadto teamwork andownership that areless thanoptimum among developers.
The subproject teamconcept is analternative to rigid, organizational productdevelopment. In this paper, wedescribe thesubproject concept andhowitwasdeployedin the development ofa large, complex PBXforthe AT&T DEFINrrv® CommunicationsSystem product line. Wealso give a briefoverview ofthe subproject methodology andpresentquality metrics from the PBX development. In addition, weshare somecomments
from developer andcustomer retrospectivesto highlight the keensenseofownership andteamwork within the subproject. (panel!defines acronyms andterms.)
It is important to understand thatmany ofthe subproject methodologies wedescribe are techniques thatgoodteambuilders anddevelopers use everywhere inAT&T.Butthe subproject concept goesonestepfurther. It allows us, in essence, to form anorganization that is basedon the actual work athand, rather than try to fit the work into theexisting organizational structure.
Although thispaperdescribes theuse ofthe subproject concept fordevelopingproducts acrossthe boundary between hardware andsoftware organizations, the conceptcanandshould be usedat other traditionalorganizational boundaries as well.
Project BackgroundThroughout this paper, wewill draw
on examples to illustrate the subproject
22 AT&TTECHNICALJOURNAL. MAY/JUNE 1992
concept. Because wetookthese examples from a realPBX product development, wepresentsomebackgroundinformation on the product before describing the subprojectorganizational technique.
SCope of the Project. The RIse (reduced instruction set computer) platform1 subproject waschartered inlate1987. Its purpose wasto coordinate the design,development, testing, and delivery ofthe RIse processorcomplex for the Generic 3 Release ofthe DEFINI1Y Communications System.
The hardware for the RIse processor complexconsisted ofthe:- Processor board (which is basedon RIse technology).- Memory boards.- Duplication interface board. (This boardprovides the
connection between the two fully redundant processorcomplexes that are used to ensure highly reliableoperation for the DEFINI1Y system.)
- Massstorage boards (i.e., controller, disk, and tape).- Network interface boards.- System maintenance board.Several ofthe boardshad largefirmware packages (i.e.,built-in programming for the processor, network interface, system maintenance, and massstorage).
In addition, a real-time operating system wasported to the hardware. (Ported meansexisting softwarewas adapted andtranslated to work on the new processor.) Significant development effort wasneeded on theoperating system (i.e., kernel, drivers, etc.) to accommodate the newprocessor and peripheral boards. (Thekernel is a set ofprograms forexecuting an operatingsystem's mostprimitive or basic functions. Adriver isthe set ofinstructions a computer uses to transfer datato andfrom a particular peripheral, e.g.,a memorydevice or a printer.)
The operating systemalso wasenhanced witha newdebuggerthat wascompatible with the RIse compiler's optimization techniques. (A debugger is softwarethat is designed to helpdetecterrors in a computer'sprograms.)
The RIse subproject teamdesigned anddeveloped all the hardware, firmware, andsoftware (i.e., operatingsystem) needed for the new processor complex.
The SubproJect'. Custom..... The RIse subprojectteamhad several internal customers to whom deliveriesweremade (see Figure 1).These internal customerswerefunctionally organized teamswithin ourbusiness
Panel 1. Abbreviations, Acronyms, and lennacustomer-supplier model- a work process that
emphasizes relationships between the customerandsupplier, processinputandoutput, andrequirements andfeedback
KNCSL - thousand noncommentary sourcelinesofC-Ianguage code; doesnot include comment linesin the code
methodology - the processes, metrics, anddocumen-tation developed fora particular taskor technique
MR - modification requestPBX - private branchexchangePQMI - process quality management improvement; a
seven-step methodology forprocessmanagementandcontinuous improvement
RISC (pronounced risk) - reduced instruction setcomputer; an architecture in which the instructionset is reduced to a minimum to increase processingspeed. The instructions used mostoften arebuiltin;othersare donebycombining the built-ininstructions.
unit, who added additional development before the finalproduct wasreleased to market.
Ourlargestcustomer wasthe applicationsoftware development community forthe DEFINI1Y system, which consisted ofabout 115 developers. Thiscustomerrequired that the subproject software, firmware,andhardware deliveries be stable, so that developmentofthe extensive application software could proceed onschedule.
Another major customer wasthe PBX testingorganization. Its rolewasto certify the quality oftheentirePBX product before the product wasshipped toexternal customer sites.
Finally, weviewed the AT&T Denver Worksin Colorado as an important customer. Thiscustomer(which wereferto as the factory) expected our designsto fit smoothly into itsjust-in-time manufacturing processes." andto be easyto build and maintain.
The RIIC SUbproject Team. Overthe life ofthe project,the subproject teamconsisted of40to 60peoplewho came from about 15 supervisory groups. We (theauthorsofthis paper) werethe co-supervisors fortheentiresubproject, andinterfaced with allthe developers
AT&TTECHNICALJOURNAL.MAY/JUNE1992 23
Systemtest
Subprojectcustomers
Subprojectrelease
--- Deliveries
---- Feedback
Denver Worksfactory tools ,
II
Software -Idevelopment
III~
I IL ..J
Operatlnc system software
•••
Hardware and flnnwarecroaa-functlonal
eubproJect teams
• Suppliers to RiSe subproject
[:=J Rise subproject customers
r,II~III~
I IL .J
Figure 1. The subproject functionalstructure shows thedeliveries to the subproject team fromthe various crossfunctional teams andfrom the operatlngsystem softwaredevelopers. Thesedellverables areIntegrated togetherand then tested as aunit before beingreleased to the subproject customers.
andmanagers involved in the development.The RIse platform and the DEFINIlY systemsoft
ware weredeveloped in parallel, and the software wasportedto the newhardware reasonably latein the development program. Hence, the software development community used interim hardware to develop the software ona parallel development track, while the subproject teamdeveloped the RIse platform. The parallel effort wasdesigned to reducethe overall project-development interval, yet have state-of-the-art hardware incorporated intimeforthe product releasedate.
What •• a Subproject?Now that wehaveprovided background on the
DEFINIlY systemproject, weare readyto discussthemorestructural and managerial aspectsofsubprojects.Hereand in the sections that follow, wepresentthe genericconcept ofa subproject and illustrate it with examplesfrom the RIse subproject.
Asubproject is nothing morethan a teamthatis managed by a small numberofsupervisors and isresponsible fora broad, easily separable, functional partofa product. Hence, a largeproduct-development effortcanhave multiple subprojects that operatewithin theproduct boundaries.
In a subproject, organizational boundaries aremeant to be ignored, andteam members frequently
comefrom many different organizations. The subprojectteammembers own the development andare empoweredto solve issues. Unfettered by rigid organizational boundaries, this sense ofownership leadsto creative problemsolving, highexpectations, and improved employee satisfaction and morale.
In our casewith the RIse processor complex,everyteammember's jobwasto identify and resolve anyissue that wasrelated to introducing the new processorcomplex boardsinto the DEFINIlY system product. (Theissuescould be maintenance codeandalgorithms, duplication, factory tools, application interfaces, etc.) Ourprimary mission wasto makesure that allissuesabout theprocessorcomplex wereresolved, and that our subproject teammadehigh-quality deliveries on time.
Subproject Prlnclp••• and Technlqu••It is important to acknowledge that just forming
a teamandcalling the teama subproject is notenough.Just as important is to have teamnorms andvalues.Whensubprojects are started, suchguiding principlesshould be statedexplicitly bythe subproject management. These principles tend to be reiterated many timesbyteammembersand, thus, become the main themesofthe subproject.
Asan example, these weresomeofthe guidingprinciples for the RIse subproject:
24 AT&TTECHNICALJOURNAL. MAY/JUNE 1992
- There will be no "throwing thingsoverthe wall" toanotherorganization. Wewant to minimize the myopictendencies oftraditional handoffs between hardwareand software, andwewant to maximize accountabilityand ownership of issues. (Throwing things over thewall refers to the practice ofcreatingandhanding offa product or processwhose design does notconsiderthe needsor limitations ofthe other design, development, or manufacturing organizations and processes.)
- Because teamwork is ofthe essence,supervisors mustact as the rolemodels. Minimal escalation ofproblemsbeyond the supervisory level is a keyingredient toteamwork.
- Technical decisions mustbe madeat the developerlevel, with consensusand buy-in. Empowered teamsown their work!
- Wewill makeuse ofsubject-matter expertsto improvequality and shorten intervals. Asan example, while thedeveloper is designing the circuit, independent peoplecanbe enteringthe circuit schematic intothe computer, simulating circuit operation, and designing theboardlayout. Likewise, allmodels (including the firsthardware prototypes) canbe orderedand assembledbyfactory personnel.
- Wewill strive forperfection in allwedo,providingexcellent response-timely, knowledgeable,professional-to issuesand questions that our customersraise. There is no substitute fora competentteam-people makethe difference!
Along with the guiding principles, subprojectsmustuse specific organizational techniques to structuretheirwork. The specific subproject structureis not asimportant as the agreed-on organizational techniquesthat have beenwell communicated andhavethe team'scommitment.
The Rise subproject used the following organizational techniques:- Ourwork will be viewed as a combined delivery to our
customers and consistsofhardware, firmware, operatingsystem, and software tools. The subproject deliveries will have been completely tested as a unitby independenttesters, according to extensive test plans.
- To promote furthercooperation among subprojectteammembers, wewill form cross-functional teamswithin the subproject. Eachsuch teamwill have milestonesand demonstrations forwhich allmembers ofthe teamare responsible.
For example, weformed a cross-functional teamforeachboardofthe Rise processor complex. The appropriatehardware, firmware, and software developers were members ofeach such team. For the mass storage team, forexample, a keyfunction to be demonstrated was: savingPBX translation information to disk. Even though, at thehighest level, this was a software function, the hardwareand firmware members ofthe teamwere just as responsible as the software memberforachieving the milestone.
These demonstrations wereimportant focalpoints forteammembers, andallowed themto broadentheir scope beyond their traditional developmentresponsibilities.
Figure 1 shows the functional organization forthe Rise subproject. This structurefollowed thecustomer-supplier model,3 whereeach member ofthesubproject teamwasa supplier and the subproject servedas the single point ofcontact forthe customers (i.e., theDenver factory, the software development organization,and the test organization).
Notice howthe structurein Figure 1 minimizedorganizational interfaces andproduced a package ratherthan pieces. This package wasthen delivered as a platform to our customers in software development, testing,and the factory. Early in the development cycle, boththecustomers and the suppliers agreedto the quality leveland delivery dateforplatform releases.
Subproject Planning, Tracking, and MethodologyIn addition to guiding principles anda team
organizational structure, subprojects-like mostAT&Tprojects-need rigorin their planning and tracking. Thissection describes someofthe planning andtracking policiesused by the Rise subproject.
While the ideasthatwepresentin this sectionare considered a BestCurrentPractice in many AT&Tdevelopment organizations, sometimes these ideasareshoved aside or overlooked. Hence, reiteration is appropriate. (InAT&T, a Best Current Practice is a practicethat covers one part ofthe total development process,has substantial favorable experience associated with it,andis considered competitive bymostpractitioners.AT&T Bell Laboratories has a quality initiative thatidentifies, documents, and deploys these practices.)
Oneofthe first thingsthat the subprojectsupervisors should do is use processquality management improvement (PQMI) 4 to chart the subproject's
AT&TTECHNICALJOURNAL.MAY/JUNE1992 25
Reiddocuments
Hardware andsoftware release
to factory
Factorytests(diagnostic,
x-ray)
Product-designdocuments
FllAi.............U.C.,...
deYeIopment
Allocateheadcount
Set productschedules
Budget
Technicalprospectus
Existingfirmware anddevelopment--------..........,environment
Market needs
Figure 2. The Rise subproject PQMI diagram shows organizational handoffs for the Rise subproject development. Extensive cooperation between organizations was needed during
design and development. Application softwareand the platform were developed In parallel, so a subproject test wasneeded before the platform was released to the application.
26 AT&TTECHNICALJOURNAL. MAY/JUNE 1992
organizational handoffs. Such diagrams helpmanagersbroaden their thinking, leading to betterplanning andorganization.
Asan example, Figure 2 contains a high-levelPQMI diagram that wascreatedby the RIse subprojectsupervisors.
First, notice the extensive cooperation neededbetween organizations (i.e., forhardware, firmware, andoperating system) duringthe design anddevelopmentphases. Bylooking at the numberofnecessary interactions, werealized that the concept ofnot throwing thingsoverthe wall and the concept ofa subproject forthehardware, firmware, andoperating system effort couldgreatly improve the quality ofour design andthe initialdevelopment.
Second, werealized that,because the applicationsoftware wasbeingdeveloped in parallel on this project,weneeded a form ofsubproject test before wereleasedthe platform to the application customers.
Finally, as a resultofdrawing this diagram, weextended our concept ofcustomers, broadening it tocover systemtest and the factory.
Besides having an understanding ofthesubproject's organizational boundaries, eachsubprojectshould also have a basicsubproject development plan.This plan describes such thingsas deliverables, responsibilities ofeachsubproject teammember, andexternaldependencies on people outside the subproject. The planshould also document the subproject's demonstrationsofkeyfunctions (e.g.,jirst phone call on RIse hardware)andwhentheyshould occur. Subproject integrationschedules, which showkeysynchronization points forthe subproject, should also be included. In addition,methodology should be briefly described in this plan.
Werecommend that the subproject supervisorswrite this plan and that supervisors in the subproject'scustomer organizations review it.However, webelieveit is critical that the members ofthe subproject teamsetthe datescontained in the plan.
Another important area forsubprojects is theexchange ofdesign andimplementation ideasamongteammembers. Much ofthe team'ssense ofownershipstemmed from the weekly, self-managed, design teammeetings. In this forum ofpeers,teammembers discussedarchitectural issues, tracked action items, andrefined the methodology. The designers followed the
BestCurrentPractice development methodology (i.e.,requirements, architecture, design documents, inspections, etc.) andwereheavily involved in peer-levelreviews at every step.
Finally, subproject teamsmusttakethe initiativeinhelping to headoffcustomer issuesbefore theyhappen. Therefore, it is imperative that the subproject supervisors be intimately involved with the various subprojectcustomers. For the RIse subproject, forexample, wehelped write the overall PBX project plans, attended allproject meetings, andtracked milestones in the projecttracking database.
Qu.11Iy Metrics from • R••I SubprojectBynow, readersshould have some understand
ingofthe organizational andplanning aspects ofa subproject. Therefore, wecantakea quantitative look at thequality improvements that come from suchanempowered subproject team.
This section provides a rangeofmetrics fromour RISC subproject. When selecting the quality metricsandresultsto be presented here,wedecided to show avariety ofmetrics thatwefelt reflected overall quality.Forexample, wepresenttraditional fault ratesandcustomer-filed modification requests (MRS) that documenttroubles or problems found byproduct users. (Wereferto these as nonenhancement MRS. An MR may alsobe filed to requestan enhancement.) Butwealso believethat the high-quality design work wedid upfront wasreflected in our ability to demonstrate keyfunctionsaheadofschedule. Hence, wepresentsome scheduledataas well.
PartAofPanel 2 illustrates the size (i.e., thecomplexity) ofthe RIse platform. This information isneeded to understand fully the scope ofthe developmenteffort and the significance ofthe quality metric results.
PartBofPanel 2 contains the subproject's MRdata. It representsallnonenhancement MRs thatwerefiled bycustomers afterthe firstsubproject delivery onFebruary 1,1990. (MRS filed to requesta productenhancement are not included in the data.)
In PartC ofPanel 2,weshow cumulative software fault rates (since February 1,1990) forthe RIseplatform firmware, operating system, anddebugger.These ratesare aboutten timesless than the typicalsoftware fault rateoftwo to three faults per thousand
AT&TTECHNICALJOURNAL. MAY/JUNE 1992 27
B. Cumulative MRDataSince February 1,1990 (i.e., sincethe first
subproject delivery), customers have filed the followingMRs forchangesother than enhancements:
Package MRcount
Hardware 0Firmware 16Operating system 38Debugger 7
Toad 61
Panel 2. RISCSubproject Quality MetricsHere,wepresentquality metries for the RISC
processor development effort to illustrate the benefitsofthe subproject concept.
A. Project ComplexityThis tableillustrates the sizeor complexity of
the RISC platform. KNCSL standsfor thousand noncommentary source lines ofCcode.
D. RISC Platform DemonstrationsThe subproject approach has enabled the
developers ofthe RISC processor complex to meetorbeatschedule dates:
Fault rate
0.12/KNCSLO.34/KNCSL0.18/KNCSL
Unit
FirmwareOperating systemDebugger
Original ActuaIMilestone date date
Operating systemboot 5/6/89 5/2/89Operating systemoperational 5/15/89 5/2/89Bootfrom disk 6/15/89 5/22/89Bootfrom tape 7/1/89 5/22/89Start application port 8/15/89 8/14/89Primary portnetwork phone call 9/15/89 8/17/89Extended portnetwork phonecall 11/1/89 8/24/89Start application development 1/1/90 11/1/89
C. RISC Subproject Fault RatesSince February 1,1990, the cumulative software
fault ratesfor the RISC platform firmware, operating system,anddebuggerhave beenaboutten timesless thanthe typical software fault rate forsimilar software:
Size
18boardcodes131.5 KNCSL
110.17 KNCSL38.14 KNCSL
Area
HardwareFirmwareOperating systemDebugger
noncommentary sourcelines (KNCSL) ofC code.Another important metric centersaround the
models conversion program wherethe newRISC hardware replaced the interim hardware that the softwaredevelopment community had been using. The hardwareconversion wentflawlessly. About 15 systems wereinvolved; andeachwasconverted successfully, withonly about 2 hours ofdowntime needed to replace aprocessor complex. Since then, the models have beenup and running 100 percentofthe time. This representsa vastimprovement overprevious development effortsthat required substantial rework duringthe prototypeprogram, resulting in longspansofinstability and modeldowntime.
Afinal metric shows the subproject team'sabilityto meet (orbeat) expected schedules. PartD ofPanel 2contains our startupview ofthe important subprojectmilestone dates. Notice that it tookonly three days (i.e.,
August 14 to 17) to bringup an entirePBXapplication,instead ofthe projected one month (i.e., August 15 toSeptember 15). Also notice that the subproject customers(i.e., the application software developers) wereabletostart work two months soonerthanexpected.
Retrospective. on Subproject.Wehave discussed subproject concepts and
philosophies andlooked at somesubproject quality metrics. Now, it is timeto share somecustomer and developercomments about working with andona subprojectteam. These remarks point outthe customer focus thatsubprojects canhave andthe sense ofempowermentteammembers canfeel.
The DEFINITY system's project manager providedthe following piece ofcustomer feedback, which is representative ofwhatour othercustomers have told us:The quality ofthe delivery was outstanding. Historically,
28 AT&T TECHNICALJOURNAL • MAY/JUNE 1992
the creation ofa newplatform of thiscomplexity is aprojectmanager's nightmare because it requires close integration ofhardware. firmware. andsoftware. The traditional approach ofeach organization doing its partandhanding it off to the nextgenerally leads tomanyunpleasant surprises. As project manager, I was neveraware ofanyproblems ordisconnects within the RIsesubproject or between the subproject andthe main project. When the RIse productwas finally delivered, it metorexceededallourgoals andexpectations forqualityandperformance. Interfacing with the subproject was apleasure. The teamhadapersonalgoalof striving forexcellence in everything it did-not onlytechnically, butorganizationallyas well.
After wehad completed our first phaseofdevelopment (i.e., aroundApril 1990), weconducted a developerretrospective, or survey, to obtain feedback aboutthe subproject. These remarksweredrawn from that retrospective, and from a later survey:- While this successful development usedseveral
tried-and-true qualitypractices, the overriding commitment andambition to drive fora qualityproduct wasachieved through genuinecaring andteamworkamongpeers.
- I enjoyed beinginvolved with thisproject. Therewas seemingly lowoverhead andindependence of subgroups. I liked the fact thatwedecided the milestonedates and were leftresponsible formanaging the timein between. I also liked working in small, containablegroups.
- The team became essentially self-motivating and. toagreatextent. self-governing.
- We started asaratherunorganizedgroup, tryingtofigure out what to doandhow toproceed. Perhaps.in a way. lettingthe teamstruggle through thisinitialperiod of uncertainty onits own was itselfa teambuilding exercise. Later, writing documents thatrequired alotofinteraction between people from different departments anddisciplines generated a strongsenseof teamparticipation.
- Organizing the project by functional subprojects,ratherthan along departmental lines, fostered teamworkby encouragingpeople from different areas toworktogetheratanearlier stage in the project.
- It may soundsilly, butanalmostfamily-like feelinghas developed amongthe teammembers.
ConclusionThere are many ways to manage a team. This
paperdescribed the benefits oforganizing around a subproject whereorganizational boundaries are minimizedand the customer becomes the major focus. By usingthis approach, the DEFINIlY system's RISC processorteamachieved excellent resultsin terms ofquality deliveries,customer satisfaction, and improved moraleamong teammembers.
Otherswho are interested in usingthe subproject approach can do so by scrutinizing their productdevelopment and identifying large, easily separablepiecesofdevelopment work that spanorganizationalboundaries. These work areascan then be organizedinto subprojects.
AcknowledgmentsWedidnot invent the subproject concept. The
technique has been usedsuccessfully at AT&T on otherPBX development projects, perhapson a smaller scale. Itis alsojust a good, common-sense technique thatgoodteambuilders intuitively understand anddeploy. Weacknowledge previous uses ofwhatweare calling asubproject.
Wewould also liketo thankTonia Wright,Ken Roberge, andTomFisherfor their keen management insights; and the members ofthe RISC processorsubproject for their sense ofownership, energy, andhard work. TomFisherdeserves special thanksformanaging our subproject testingefforts.
References1. I. R. Mashey, "RISC, MIPS and the Motion ofComplexity," Proceed
ings ofthe UniForum Conference, February 4-7,1986, Anaheim, California, USR Group, SantaClara, California, 1986, pp. 115-124.
2. ]. D.Carboy, G.Foo, L. P.Jones, L. E. Kinney, and D.C. Krupka,"Striving forManufacturing Excellence at the Denver Works: ASummary," AT&TTechnical Journal, Vol. 69,No.4,July/August1990, pp.5-18.
3. R. B.Ackerman, R.]. Coleman, E. Leger, and]. C.MacDonnan,Process Quality Management andImprovement Guidelines: Issue 1,SelectCode500-028, AT&T Customer Information Center, Indianapolis, Indiana, 1987.
4. AT&T, PQMI: Tips, Experiences, andLessons Learned, Select Code500-446, AT&T Customer Information Center,Indianapolis, Indiana, 1986.
(Manuscript received October 23, 1991)
AT&TTECHNICALJOURNAL.MAY/JUNE1992 29
Kathleen K. Glass is a supervisor in the Integrated SystemsDevelopment Department with AT&T Bell Laboratories inDenver, Colorado. She is responsible for the development ofhigh-performance processor subsystems for AT&T's line ofDERNITYCommunications Systems. Ms. Glass joined the company in 1979. She has a B.S.E.E. from Virginia PolytechnicInstitute and State University (Blacksburg, Virginia) and anM.S.E.E. from the University of Colorado (Boulder, Colorado).Lucinda M. Sanders is a supervisor in the Advanced SystemsDepartment with AT&T Bell Laboratories in Denver, Colorado.Her group develops software for the Generic 3 Release of theRISC platform. Ms. Sanders joined the company in 1977. Shehas a B.S. in computer science from Louisiana State University (Baton Rouge, Louisiana) and an M.S. in computer science from the University of Colorado (Boulder. Colorado).
30 AT&TTECHNICALJOURNAL.MAY/JUNE 1992