NODAL: A Filesystem for Ubiquitous Collaboration

1
NODALA Filesystem forUbiquitous Collaboration

• Lee Iverson
• SRI International
• leei_at_ai.sri.com

2
High-performance Teams

• Teams should be more than sum of parts
• Rapidly make decisions
• Make the right decisions
• Respond quickly to new tasks
• Rapidly allocate resources
• Reuse previous work in context
• Integrate remote members effectively

3
Ubiquitous Collaboration

• Development of rich, complete, adaptive,
  exploitable knowledge archive for teams.
• Facilitation, recording and indexing of dialogue
  and processes of decision making
• What?
• Who and when?
• Why?

4
Barriers

• High cost of entry
• Extra work to get data into shared repository
• Much relevant data is never entered
• Failure to interoperate with legacy applications
  and work models
• High cost of exploitation
• Inadequate search and reuse
• Not integrated with tools
• No support for dependencies/relationships

5
ExampleSoftware Development

• Products
• Design documents
• Source code
• Documentation
• Papers
• Marketing documents
• Discussion
• Email
• Chat

6
ExampleSoftware Development

• Relationships and dependencies are critical but
  implicit
• Maintaining dependencies requires great effort
• Expensive, unreliable
• Combination of products and discussions form a
  knowledge base
• Currently unexploitable

7
Example Proposals

• Products
• Background research
• Proposal proper
• Costing
• Bios
• Integrated, online product and knowledge base
  more critical when collaborators physically
  distributed

8
OHS/DKR

• Open Hyperdocument System
• Open standard
• All kinds of documents
• Tools for creating and manipulating
• Dynamic Knowledge Repositories
• Recorded dialog
• Collective IQ

9
OHS/DKRProposed Organization
Brain metaphor
10
OHS/DKR Data

• Filesystem Database
• Data Model for Documents
• Content and Structure
• Communication model(s)
• Security/Privacy
• Auditing Attribution
• Neurons

11
OHS/DKR Knowledge

• Semantics
• Explicit knowledge
• RDF/DAML
• Topic maps
• Object models
• Long term memory

12
OHS/DKR Context

• User interfaces
• Perception and action
• User modelling
• Intention
• Action
• User adaptation
• Short term memory, Perception

13
The Data Layer Requirements

• Distributed network oriented
• Application independence
• Cross-platform
• Cross-language
• Cross-application
• Shareable and reusable content
• Built-in security/privacy model
• Change attribution and auditing
• Legacy document and application support

14
The Data Layer Requirements

• Flexible communication model
• Synchronous
• Asynchronous
• Hyperlinking
• To/from any context
• Stability through changes
• Rich search facilities

15
NODAL The Data Layer

• Network-Oriented Data Abstraction Language
• Web-based filesystem and database
• Granularity below file level
• General, extensible data model
• Fully addressable content
• Inherent hyperlinking and reuse
• Radical simplicity and generality

16
NODAL Filesystem

• Traditional filesystem model
• Hierarchical directories
• Documents
• Accessible by HTTP/S
• Documents all typed (MIME)
• Encoder/decoder
• Type of root node of document data model

17
NODAL Type System

• Atomic types
• boolean, integer, character, octet, float, double
• Node types (collections)
• Struct like C struct
• Sequence indexible sequence
• Map extensible dictionary
• Language for extending/composing types

18
NODAL Data Model

• Document is graph of nodes
• Accessible from root node
• Nodes reusable
• Within and between documents
• String is
• ltsequence itemTypecharactergt

19
NODAL Navigation

• Every data item addressable by URI
• Paths
• Document URI specifies root node
• Path operators for moving through node graph
• Every path expressible as URI
• Hyperlinks (dependencies)
• Any set of URIs can be linked
• Embedded and external links managed

20
NODAL Metadata

• Node is basic unit of granularity
• Extend filesystem principles inside documents
• Every node has
• Unique ID
• Version and transaction history
• Permission record

21
NODAL Communication

• Local memory store (cache)
• Storage interface (transactional)
• Wire protocol(s)
• XML-RPC
• SOAP
• OAA
• SQL binding (PostgreSQL, Oracle, etc.)
• Lightweight databases (Berkeley DB)

22
NODAL Search

• Simple search language
• SQL as model
• Structure and content
• Usable as content filter (views)
• Always up-to-date
• Possible integration with P2P search
• Gnutella
• NEVRLATE

23
NODAL Application Model

• Design data model
• Adopt/adapt document format
• MIME type encoding plugin
• Build application on top of NODAL API
• Object model gt data model
• A shared, collaborative application!

24
NODAL Legacy Applications

• Filesystem interface via WebDAV
• Traditional distributed filesystem over HTTP
• COM/CORBA applications expose controls
• Windows applications (Office) augmentable
• Linux (GNOME/KDE) expose CORBA controls

25
New capabilities

• All documents integrated into reusable knowledge
  base
• Live collaboration recorded and integrated into
  knowledge base
• Dependencies easily tracked and notifications
  processed
• Automatic processes integrated with manual

26
Why Open Source?

• Ubiquity
• Create an industry, not an application
• Web model
• Simple, open standards
• Open source implementations
• Free, end-user applications

27
Open Source Funding?

• Government funding
• Agencies becoming more aware of the advantages of
  open source efforts
• Foundations
• Mostly educational
• Enlightened corporations
• Sun, IBM, HP, MSDW,
• Venture capital?

28
Open Source Funding?

• Service model
• Companies pay for service and development of open
  source technologies they depend on
• Client has much greater control of development
  process than with proprietary software
• In-house developers can easily work with others
• Given enough eyeballs, all bugs are shallow.

29
Competition

• None really.
• Alternatives all smaller scope
• Groove
• Successor to Lotus Notes
• Proprietary, Windows-specific
• Integration of DB and UI
• Must build new applications (SDK)

30
Competition

• Subversion
• Version control for software development
• Purely asynchronous model
• Incremental path from CVS
• No direct access to document structure