From: jamsden@us.ibm.com To: ietf-dav-versioning@w3.org Message-ID: <8525676D.003C8806.00@d54mta03.raleigh.ibm.com> Date: Mon, 10 May 1999 06:59:47 -0400 Subject: Re: comparing configurations and baselines (I have trimmed the items I didn't have any comments on) "Geoffrey M. Clemm" <gclemm@tantalum.atria.com> on 05/07/99 05:19:15 PM To: Jim Amsden/Raleigh/IBM@IBMUS cc: Subject: Re: comparing configurations and baselines From: jamsden@us.ibm.com > A configuration is a versionable resource (not necessarily versioned). > A baseline is a revision (a deep revision of a baselined-collection). It seems strange to talk about a revision of something that isn't a versionable resource. This doesn't feel like consistent semantics. In any case, a configuration could be a revision too. A revision is not a versionable resource. A baseline is a special kind of revision (in particular, a deep-revision). How is this any more strange than saying that "a revision is not a versionable resource"? <jra> I wasn't saying a revision is a versionable resource, but that a baseline is a revision that was not created by checking in an unversioned resource. </jra> If you have a mutable thing that has this "transitively add all members" property, then you get all sorts of confusing and error prone edge cases. Let's say you explicitly add a resource. Then you add a resource that contains it. Then you delete the second resource. Does it take the first resource out with it, or does the first resource stay around? On the other hand, if you just are allowed to take a snapshot, then there is no question of what to do. <jra> This case couldn't happen. If you add a resource, and then attempt to add a resource that contains it, this would generate an error because a configuration, like a collection, can only contain a member once. It looks like most of your concerns are based on implementation considerations for efficiency. Is there any way we could add restrictions or usage scenarios to configurations to achieve what you want? </jra> Cheers, Geoff