- From: Jim Amsden/Raleigh/IBM <jamsden@us.ibm.com>
- Date: Wed, 11 Oct 2000 17:02:28 -0400
- To: ietf-dav-versioning@w3.org
I believe that these issues should be discussed on the mailing list during the last call period, and not deferred to a discussion at the Dec working group meeting. Certainly, any issues that cannot be resolved on the mailing list should be discussed at the Dec working group meeting. <jra>The mailing list is a great place to capture, discuss, and resolve issues. But the working group meetings are often attended by a broader community. To get a little broader coverage, some issues should be discussed in this community, especially when there are conflicting views presented on the mailing list.</jra> We explicitly decided to not discuss leveling over a year ago, since it didn't make sense to discuss what functionality belonged in what level until we had actually settled on what functionality we would supporting. Over the last year, the design effort has been primarily focused on firming up the details of advanced versioning support, which are primarily of interest to configuration management system providers. <jra>We have never really given up on leveling. We just renamed to to core versioning and advanced options because there appeared to be little agreement on what the levels should be. This is fine. My only point is that its interesting to note that optional labels never came up (that I can remember).</jra> Now that the dust has settled, and we have received a very thorough review from a document management provider, I believe we need to take that review very seriously. The fact that configuration management providers are happy to provide labeling support is to be expected, and is therefore neither very surprising nor especially interesting. If we want document management system providers to implement the versioning protocol (and I certainly do), we need to set the "functionality bar" at a level appropriate for document management. <jra>There has been DMS representation in the DeltaV working group and on the design team. Of course, not every DMS was represented. But we have included many concepts and compromises to accommodate DMS systems. Perhaps optional label support is another. But I think its a little like saying we want optional user-specified URLs for resources. I'm willing to consider making labels optional, but I would like to challenge the assumption that including them in core is setting the bar too high. I have designed WebDAV interfaces to two versioning management systems that didn't support labeling as specified by the protocol. Both designs supported labels. I just hid them in my own server-defined properties.</jra> The combination of standard client defined properties like DAV:comment and DAV:creator-displayname, custom client defined properties, and standard server defined properties like DAV:version-name and DAV:getmodificationdate, are sufficient to name and locate versions of interest, and this is demonstrated by the document management systems that do so (without the use of labels). <jra>This does not sound like either an adequate, optimal, or interoperable way of distinguishing revisions of the same versioned resource. DAV:verion-name might not be meaningful to a user, might not be used on more than one versioned resource, and can't be moved. DAV:creator-displayname is not unique to a revision. Relying on DAV:comment to distinguish revisions would introduce the possibility of parsing comment strings to identify a revision. This doesn't sound convenient or interoperable. DAV:getmodificationdate has no logical meaning although is is sometimes useful to identify revisions by when they changed. Custom client defined properties would not result in an interoperable solution. Lisa, of the versioning repositories you surveyed, which ones are planning on implementating a WebDAV interface? Of those, how many would like to have label support if there was some convenient way to provide it? Could labels be persisted in a server-defined property? Would it be OK for the WebDAV server implementation to do the labeling semantics, and just use the underlying DMS repository for persistence?</jra>
Received on Wednesday, 11 October 2000 17:14:36 UTC