Community Information Management
- A Grand Challenge in Distributed Systems

Doug Terry
Microsoft Research Silicon Valley
terry@microsoft.com

Problem
While the Internet has allowed unprecedented sharing of information via
electronic mail and the Web, little support exists for loosely
structured communities with common information needs.  In this context,
I define a "community" as a collection of people with minimal
organizational structure that wish to share electronic documents.
Consider, for example, the parents of a youth soccer team that need to
share team rosters, phone numbers, game and practice schedules, action
photos, party plans, and other information that maybe contributed and
updated by various community members.  Changes to a shared document
should be readily visible to all of the relevant community members.  A
given person may belong to many distinct communities, which change over
time with the person's changing activities, interests, location, etc.
Documents should remain available for as long as they are useful but no
longer.  Ideally, documents and their associated communities should be
managed in a seamless way while maintaining accessibility, consistency,
and privacy.  This is naturally a distributed systems problem since,
with the emergence of large capacity portable storage devices, users
will want to carry much of their daily information with them, e.g. on
their terabyte key chains, and yet share select information freely
within their communities, probably wirelessly.

Challenges
Solving the problem of community information management requires
fundamental research in a number of areas that go beyond storage and
communication including:
- Identity: How are communities established and named?  How do users
refer to shared information?
- Replication: Where does information reside?  Is information replicated
to facilitate sharing, to avoid centralized servers, to increase
availability, to provide reliable backup, or all of the above?  How are
updates tracked and propagated between loosely connected personal and
public storage devices?
- Trust: Who is trusted to update the info?  How is protection enforced?
How does a community member recover if their trust is misplaced?
- Versioning: How does one access old versions?  Can updates be easily
undone?
- Searching: How is information retrieved and integrated from
decentralized repositories of various types: public (e.g. the Web),
private (e.g. one's desktop), and community storage.
- Lifetime: How does one avoid seeing obsolete or irrelevant info?
- Context: How does one's current situation (place, time, meetings,
deadlines, companions, etc.) affect their view of and operations on
shared info?
- Usability: How can users readily form communities and create shared
documents without being aware of the underlying distributed nature of
the system or its inherent unreliability and vulnerabilities?  

Approach
Expanding on the notion of personal information management (PIM)
requires a fresh, clean-sheet, long-term look at information sharing
within loosely structured, semitrustful communities.  The goal should be
to produce an architecture for community information sharing and
demonstrate it through a prototype implementation.  This research can be
conducted within the context of the current Internet and wireless
communication networks.  It can build upon the Web, e-mail, and modern
storage systems (such as the WinFS platform from Microsoft).