Engineering meeting on sharing

Attending: AndiVajda, HeikkiToivonen, JeffreyHarris, TedLeung, StuartParmenter, LisaDusseault, KatieCappsParlante

Context/notes

• We are exploring the proposal to go with a server model, that is the stake in the ground and we are evaluating whether or not this can work. The server is to be used for syncronization, we expect clients to work with a local copy complete with rich connections.
• We expect to also have the p2p option down the road
• A few folks are disappointed to see us not address the p2p model out of the gate -- have we lost the coolness factor/chandler vision? In particular: trusting some central authority with your data.
• The repository sharing that Andi has been working on will be used in service of replication features. The repository-level sharing is semantic free -- we're going to be doing semantic rich/standards based sharing above the repository.

Issues identified

• Access control
  • how acls work, what is responsibility of client, what is responsibility of server, certs?
  • what are the requirements for the repository and how does that fit with current implementation?
  • what is the design model, what permissions, what about username/password, relationship to users, etc.

• Item clouds
  • relationship to version history
  • relationship to access control

• Versioning
  • conflict resolution
  • ui supporting conflict resolution

• CHANDAV design
  • roundtrip/idempotency issue

• Staged implementation plan

• Server plan that meets design requirements

• Generally, want designs to be flexible enough for future features
  • p2p model
  • federated servers (the need is perhaps addressed w/xmpp -- shared notion of identity)

Next Actions

• Lisa to meet with Mitch/Jurgen
• Lisa to write down an access control proposal
• Lisa to work on a staging plan
• Ted to drive Item Clouds as data model feature
• No next action yet:
  • CHANDAV (revisit in 2 weeks)
  • Versioning (revisit if we need to)

-- KatieCappsParlante - 15 Jun 2004