- From: Mike Little <little@bellcore.com>
- Date: Thu, 10 Apr 1997 16:25:24 -0400
- To: Judith Slein <slein@wrc.xerox.com>
- cc: w3c-dist-auth@w3.org, russell@wrc.xerox.com, sauvain@wrc.xerox.com
Judy, Thanks for the list of requirements issues, it's much appreciated. I'd like to ask for some clarification on a couple of the issues. >7. Structured Documents: Not currently mentioned in the requirements, but a >proposal for structured documents is under discussion. Should structured >documents be added to the requirements? Not coming from a document management viewpoint, could you clarify what is meant by a structured document? The baseline position is we have structure as defined by reference links. Other structuring has been implied through meta-data adjuncts. Could you elaborate on what structure (and/or other semantics) is being implied by structured documents? >8. Reservations and Versioning: Should reservations be discussed separately >from versioning? Do we want to support reservations on resources that are >not version tree handles? I view reservation as a locking mechanism and versioning as a mechanism for distinction. Not having the benefit of group discussion to give me clues otherwise, does the group in general consider additional semantics to versioning (perhaps akin to those associated with a versioning system)? While I have the editor running, I have the following comments concerning the changes asked for by Jim and Yaron: >3. Do not say which functionality is mandatory or optional. If this is to expedite discussion and keep the work moving along where it has otherwise led to unresolved discussion, I support this. Otherwise, I can't see the point of continuing the working group if the decision is just to publish a document of good suggestions of things that can be done concerning web distributed authoring and versioning. I can see ironing out the "must" and "should" near the end of the effort, but I can't see dropping it completely. >5. Change "relationship" to "link" throughout. Change "relationship" to "reference" throughout. >10. Combine the discussion of reservations with the discussion of locking. >Treat reservations as advisory locks. Also talk about shared vs. exclusive >write locks. I fully support merging discussions on reservations with those on locking (or vice-a-versa). Call them all locking mechanisms if it will help. This should help with the technical approaches. User semantics can be decided upon later. >12. Get rid of the requirements that copy and move leave audit trails. Are audit trails being dropped altogehter? If one wants to worry about authenticating users then one should be worrying about audit trails in a versioning system (at least, in the presence of versioning semantics such as "give me the newest version" or any time there are multiple participants potentially involved in the same activity). Lest I sound too harsh, are audits being kept for version establishment, but not for intermediate actions such as move or delete? That would be reasonable. A comment on one of the issues: >9. Diff / Merge: Do we want these or not? To me a versioning system deals in distinctions. As such a basic function would be to assign a distinction. If the attributes of distinction are not arbitrary, and in particular if they are derived, it would seem most appropriate to allow comparison operations. For example: find all versions with such and such attribute (ah, attribute of distinction, just in case the attributes alluded to by attribute search involve other types of attributes). Another comparison operation would be to ask what is the attribute-based difference between two (or more) versions. This leads into a discussion as to what defines a version (i.e. what is considered in distinguishing one version from another)? If it is defined soley to be some handle such as a version number/string, then either such comparisons would be pretty lame or we admit the handle is just a short hand representation of the attributes that really are considered in distinguishing one version from another. In contrast, the version/number string could be the only attribute considered, and the basis for its derivation indeterminate (i.e. arbitrary). In this case, if we still are interested in a Difference operation, we need to come to agreement on what can be compared - *independent* of the concept of versioning. The only thing versioning would be helpful for here would be to state "what" is being compared, i.e. version this will be compared against version that. I can see arguing against this latter instance as being outside the scope of the working group and an area for producet distinction. I support difference (and other) operations on attributes of distinction and feel this is part of the scope of the working group. -Mike
Received on Thursday, 10 April 1997 16:29:12 UTC