An Introduction to FluidDB

Preview:

DESCRIPTION

The slides and notes used when presenting to Manchester Geek Night.

Citation preview

An Introduction to FluidDB

(A social database in the cloud).

http://fluidinfo.com/

Nicholas Tollerveyntoll@fluidinfo.com

The Agenda:

What?Why?How?

Where?

A high-level, non-technical overview of the concepts behind FluidDB

What?...is FluidDB

FluidDB is a platform for the web of things,

each represented by an openly writable “social”

object.

But what is an openly writable social object..?

http://www.flickr.com/photos/psd/1428129861/

An example of an openly “writable” social object in the real world. FluidDB just allows this in the digital world.

Why?

http://www.flickr.com/photos/jointhedots/4622191119/

Most interesting information is currently in walled gardens sitting in databases behind applications, APIs or in weird proprietary formats.

People I know

Often you have to duplicate information in each of the different walled gardens.

WHERE location NEAR pubs SERVING

SELECT self-catering-cottage FROM

AND pub HAS rating > 7 IN

And you usually can’t request information that combines data from the different walled gardens. For example, selecting all the self-catering cottages endorsed by the Cornish Tourist Board that are close to good pubs that serve my favourite beer. So how is FluidDB different..?

Share, annotate, augment and re-use information.

At Fluidinfo we want to make it much easier to share, annotate, augment and re-use information. FluidDB is designed to make this possible.

How?

There is only one FluidDB. All users and applications share the

same database.

• Objects - represent things

• Tags - define objects’ attributes

• Namespaces - organise tags

• Permissions - control access

FluidDB is conceptually very simple.

http://ntoll.org/images/93.jpg

Openly writable objects are simply tagged with information.

Data Structure

entity / attribute / value

Data Structure

object / tag / value(in FluidDB parlance)

Data Structure

object / tag / value(optional)

Namespaces/Tags

ntoll/ratingterrycojones/books/reviewamazon.com/book/titleamazon.com/book/authoramazon.com/book/isbnesteve/books/covertimoreilly/has_read

Namespaces and tags are pre-defined by users and applications. When you sign up for FluidDB you’re assigned a namespace corresponding to your username. Every namespace and tag can have a description and is itself represented by an object within FluidDB so meta-tagging is possible.

An object

about = “book:Dune”

id =“e127475b-eb93-48e1-a24f-dde6825b9ed8”

Every FluidDB object has a Universally Unique ID (this allows applications to operate on a specific object) and an optional unique “about” tag that can be used to indicate what the object represents.

An object

about = “book:Dune”

ntoll/rating = 5ter

rycojones/

books/rev

iew =

“I love

sandworm

s”

amazon.com/books/title =

“Dune”

amaz

on.co

m/b

ooks

/aut

hor =

“Fra

nk H

erbe

rt”

amazon.com/books/isbn =

“87-418-7517-6”

timoreilly/has_read

estev

e/boo

ks/co

ver =

id =“e127475b-eb93-48e1-a24f-dde6825b9ed8”

An object

about = “book:Dune”

ntoll/rating = 5ter

rycojones/

books/rev

iew =

“I love

sandworm

s”

amazon.com/books/title =

“Dune”

amaz

on.co

m/b

ooks

/aut

hor =

“Fra

nk H

erbe

rt”

amazon.com/books/isbn =

“87-418-7517-6”

timoreilly/has_read

estev

e/boo

ks/co

ver =

booleanstring (& set)numericopaquenull

id =“e127475b-eb93-48e1-a24f-dde6825b9ed8”

There are two “types” of information: primitive (boolean, string, numeric, null etc...) and opaque (that have an associated mime-type).

snoissimreP(seem to be back-to-front)

How does FluidDB ensure that users and applications retain control over their information..?

about = “book:Dune”

ntoll/rating = 5ter

rycojones/

books/rev

iew =

“I love

sandworm

s”

amazon.com/books/title =

“Dune”

amaz

on.co

m/b

ooks

/aut

hor =

“Fra

nk H

erbe

rt”

amazon.com/books/isbn =

“87-418-7517-6”

timoreilly/has_read

estev

e/boo

ks/co

ver =

Traditional:

id =“e127475b-eb93-48e1-a24f-dde6825b9ed8”

The traditional mechanism is to control the thing being written to: a database table or document for example. But FluidDB’s aim is to make the world writeable so Objects have no owners or associated permissions.

An object

about = “book:Dune”

ntoll/rating = 5ter

rycojones/

books/rev

iew =

“I love

sandworm

s”

amazon.com/books/title =

“Dune”

amaz

on.co

m/b

ooks

/aut

hor =

“Fra

nk H

erbe

rt”

amazon.com/books/isbn =

“87-418-7517-6”

timoreilly/has_read

estev

e/boo

ks/co

ver =

FluidDB:

id =“e127475b-eb93-48e1-a24f-dde6825b9ed8”

Permissions only apply to the namespaces, tags and tag-values. Therefore users and applications retain control over who sees, reads, writes or creates data without inhibiting the writability of FluidDB.

You control who can see, read or use your tags.

Anyone can tag information to an object.

http://www.flickr.com/photos/fixe/3415776837/

Permissions

• Apply to namespaces, tags and tag-values not objects

• Scoped by actions (e.g. “see”, “create”, “read”)

• Either “open” or “closed” with a list of exceptions

It’s a bit like the unix filesystem where permissions on directories and files are also scoped by action (read, write and execute) and apply to specific users or groups.

FluidDB’s Query Language• Equality & Inequality: To find objects based on the numeric values or

exact textual values, e.g., tim/rating > 5, or geo/name = "Llandefalle".

• Textual: To find objects based on text matching their tag values, for example, sally/opinion matches “fantastic”.

• Presence: Use has to request objects that have a given tag. For example, has sally/opinion.

• Set contents: The contains operator can be used to select objects with a matching value in a set of strings. The query mary/product-reviews/keywords contains "kids" would match the object with a tag called /mary/product-reviews/keywords and value ["cool", "kids", "adventure" ].

• Exclusion: Exclude objects with the except keyword. For example has nytimes.com/appeared except has james/seen. The except operator performs a set difference.

• Logic: Query components can be combined with and and or. For example, has sara/rating and tim/rating > 5.

• Grouping: Parentheses can be used to group query components. For example, has sara/rating and (tim/rating > 5 or mike/rating > 7).

Use the very simple query language to search for information. The current specification fits onto this slide.

{“tagPaths”: [" “twitter.com/username”," “twitter.com/fullname”," “tunkrank.com/score”," “ntoll/met”," “terrycojones/met”," “python.org/member”," “ub.edu/alumni”," “ub.edu/bsc”," “uoc.edu/msc”," ... etc ...]}

Tags attached to the “about:esteve” object.

Another way to find information is to ask FluidDB for tags that are attached to interesting objects.

https://fluiddb.fluidinfo.com/objects/OBJECT_ID/NAMESPACE(s)/TAG

HTTP ‘GET’ returns the value, HTTP ‘PUT’ adds/updates it etc...

Full API: http://api.fluidinfo.com/fluidDB/api/*/*/*

A RESTful API

Users and applications interact with FluidDB via a RESTful HTTP API.

Social data?

But what exactly do we mean by social data? Let’s look at a few examples:

http://www.flickr.com/photos/8001265@N08/2259250324/

Let me demonstrate!Tickery, WeMetAt, TunkRank.

Where?...can you find out more information and get involved?

http://fluidinfo.com/

• Twitter: @fluidDB @terrycojones @esteve & @ntoll

• http://youtube.com/fluiddb

• #fluiddb on Freenode IRC

• http://groups.google.com/group/fluiddb-users

• http://groups.google.com/group/fluiddb-discuss

Lots of 3rd party libraries:

Java, Clojure, Lisp, .NET, Perl, Python, Ruby, PHP & Javascript (all open-source)

http://fluidinfo.com/developers/libs

We’re listening to youthough the usual caveats apply ;-)

Tell us what you think!

Questions..?

Recommended