WebDAV conference calls

June 9, 2000: 11am – 12:00 pm

Attendees:

Barry Lind – Xythos

Eric Sedlar – Oracle (taking notes)

Geoff Clemm – Rational

Jim Whitehead—UCIrvine

Andrew Fuque—???

Resolved:

• We need to add into the spec the point that to update an ACL on a resource, you must go to the URL specified by the source-resource tag, LOCK that resource, and PROPPATCH.
• We agreed to use PROPFIND/PROPPATCH for all ACL information (rights discovery, ACL updates, etc.). This should address most of Jim Davis's concerns.
• We agreed on the need to enumerate the various schemes for ACE interactions via an OPTIONS-style method.
1. We need to produce a list of known ACE interaction schemes and DAV: tags for them
2. ACE interaction schemes will be identified by an XML namespace-qualified tag, but no other information is available
3. An IANA registration method will be needed for vendors to add different interaction schemes to the list.
• We need a REPORT to list all of the resources using some or all of an ACL (Action Item for Geoff to write up)
• We need a way to find the URL of the root collection for where servers will make principal resources available. (Action Item for Geoff to write up)
• Eric to bug Anne about the next draft ;-)

Open Issues:

No new issues raised!

Discussion:

• We agreed to keep the current "replace all of the ACEs" procedure for modifying ACLs rather than go with Geoff's incremental update proposal due to the complexity of dealing with ordering/reordering for systems for which that is important.
• Discussion of Anne's concern about the cost of computing the source-resource information. The conclusion was that you only need to get an ACL at all when doing authoring, since the "ls –la" style listing of access control information for a set of resources in a collection will probably display the <rights-list> rather than an entire ACL. So ACLs are expected to be retrieved with low frequently and in small quantities, so this cost shouldn't be too bad. Only if the client requests <allprops/> will this cost be an issue, and everyone seems to agree that people should expect something in <allprops/> to be slow. Servers should optimize to make the <rights-list> come back quickly, where possible.
• Jim Davis proposed combining the <principal-type> and <resourcetype> information into the <resourcetype> tag by allowing multiple tags there. The consensus on the call was that there wasn't a clear benefit in this, and forced more complex parse logic, so we would keep this the same.