- From: Julian Reschke <julian.reschke@gmx.de>
- Date: Mon, 17 Nov 2003 10:26:49 +0100
- To: <w3c-dist-auth@w3.org>
Here's the content of the minutes document (for archival on the mailing list): -- Notes for WebDAV WG meeting at IETF 58. Thanks to Larry Greenfield for notes/chat transcript. TOPIC: Agenda bashing - no objections TOPIC: ACL status - Security ADs did not approve and did not block. - Now in the RFC editor queue. - major problem was the unbounded number of rights. - security ADs says nothing could be done to make this spec a good spec, but weren't willing to make the WG restart. - hopefully "application acl workshops" will be held. - for future revs, think about "intersection" instead of "union" solutions. TOPIC: Quota/size properties Brian Korver - draft to let things interoperate with quota implementations. - -02 document aligns better with NFS. - quota_avail added back in to deal with low disk space condition - the thought is that a MAY for including disk space is ok. - we could remove from spec and a future application can give disk limits. Some additional questions on Quota Q (Reschke): why are disk limits treated as quota when (a) NFS behaves differently and (b) making a difference could enhance client error handling (such as mapping to error codes in file system drivers)? A: some memory that WG members had asked for disk limits to be treated as quota Proposal: marshall as separate properties, define distinct condition codes, so that clients can distinguish both conditions (just like in file systems) Q: (Randy presuhn) Why do you need to know this number? A: to not attempt operations that would almost assurdly fail. Suggestion (Reschke): For instance, we may just define a generic mapping of RFC3530 properies to DAV properties, then apply that mapping to the two quota and the two disk space properties defined there. TOPIC: Redirect draft. - issue: Replace MKRESOURCE with MKREDIRECT? - issue: allow authoirng 301-type redirects as well as 302-type? - issue: How to handle requests that address mutliple resources, including redirects - no comments from attendees. TOPIC: Interop Report. - 4 clients, 6 servers. this is only about 30% of implementors. - possible to participate remotely - also ongoing interop testing using "always online" test servers - Jim Luther @ Apple is maintaing a list of public test accounts TOPIC: RFC 2518bis issues Do namespace prefix tags need to be preserved - issue is with XML vocabularies that use prefixes in text or attribute content, - such as XSLT or XML Schema - "prefix rewriting") would break fragments of these vocabularies appearing in properties - Requirements for ETag support - question for ted: when going to draft standard, how hard do you try to keep existing implementations compliant? - Ted: It's certainly possible that one or more implementations will become uncompliant. - if it changes behavior, then usually you recycle at propose. - other things are big changes, too. - you can remove features but you cant change features or add features without recycling at propose. - Ted: the standard level isn't a big deal, so don't worry about recycling at proposed. make the spec as good as possible. More commentary: Suggestion (Reschke): It would be really good if some of the mod_dav maintainers could state whether they are planning to actually meet the stronger etag requirements. Q (OGeisser): are there any plans for "ptags" = "etags for properties"? A (Lisa): Yes, or ctags = "ctags for collections": would rather not add major new design to webdav, just make stronger compliance, simplify features, etc. New features that would be added in new extensions is the thought here. (reschke): I certainly dislike the idea of having an RFC2518++ that isn't supported by Apache. (Ted): send the source to apache to fix the issue, and that's done. Issue: Must resources be in collections - Consensus is that resources can be standalone. Issue: Some interest in new writable modification-date property - for instance, a backup program could upload stuff to a server and set the modification-date property from the original. - Jim Luther on the filesystem team here at Apple wanted this property. - Changing getlastmodified considered dangerous - Note that setting DAV:getlastmodified *back* implies changing the MIME header last-modified as well, and that may be a bad idea. - it could screw up things for intermediaries and caches, unless we divorced the property value from the header value which might create more confusion. - We may instead want to define a new property that can track modifications to properties as well (that would resemble more closely a filesystems last mod date) TOPIC: PATCH proposal http://www.sharemation.com/~milele/public/dav/draft-dusseault-http-patch-00. txt - WebDAV use case: would make it easier to make small changes to very large files. - DeltaV use case: small changes to many source files. - SIMPLE use case : changing my buddy list or presence document on HTTP server. - Netconf use case: Making changes to large configuration documents. Q (OGeisser) does the PATCH proposal allow a partial change of a property value, too (sorry - have not read the proposal) A: No Suggestion (Reschke): an alternative would be to model parts of a resource as indivually addressable sub-resources (with their own URIs) TOPIC: BIND Status is at <http://www.webdav.org/bind/bind-issues-list.htm> Reschke: My understanding is that we want to last-call the spec when the open issue is resolved. Lisa has other issues which she will raise on list.
Received on Monday, 17 November 2003 04:27:03 UTC