- From: Yaron Goland <yarong@microsoft.com>
- Date: Tue, 20 May 1997 18:17:23 -0700
- To: "'Sankar Virdhagriswaran'" <sv@hunchuen.crystaliz.com>, "'Judith Slein'" <slein@wrc.xerox.com>, w3c-dist-auth@w3.org
Versioned Stores Vs. Non-Versioned Stores: We are doing both. The issue was only if a server had to support versioning in order to be compliant with the standard. I was explaining that we had always agreed that there would be different levels of DAV compliance. One level would not support versioning, another would. Email - I believe we should take e-mail needs into consideration. However I also believe that we should not constrain our decisions based on e-mail's needs. Yaron > -----Original Message----- > From: Sankar Virdhagriswaran [SMTP:sv@hunchuen.crystaliz.com] > Sent: Tuesday, May 20, 1997 6:09 PM > To: Yaron Goland; 'Judith Slein'; w3c-dist-auth@w3.org > Subject: RE: Requirements Changes for Your Review > > >> 3. Locking Model: Must servers support the WebDAV versioning model > in > >> its > >> complete generality? (5.9.1.3) > >> > >> [Yaron Goland] I am not familiar with anyone who has stated that a > >> base level DAV server must support versioning. It has always been > the > >> case that we would provide different levels of functionality, at > >> minimum between versioning and non-versioning systems. > > Yaron's point does not make sense to me. Maybe I am confused? > Versioned > stores vs. non versioned stores has a fundamental impact on naming and > identity of objects. Therefore, I would think this is a fundamental > decision. Either we are going to do versioned stores or not. Perhaps I > am > missing something. > > >> So language such as "advisory lock" should disappear. > > While I agree with Yaron's sentiment about where implementation > specifications should appear, I do not want to see the notion of > "locks > that are advisory" disappear. We think of all types of locks as > advisory > only. When a lock is violated, what happens is a policy notion and can > be > implemented in different ways given the state of the doucment-object > and > state of transaction that is trying to break the lock. Therefore, I do > want > to see advisory locks in place at a different place in the doucment > perhaps. > > >> types of objects to the Web: links, collections, version graphs, > etc. > > I am trying to find where the etc. is defined. Can you point to a > document > segment that describes the new object model. I am sure we will have > something to say about it. > > >> [Yaron Goland] If we need to do something that breaks e-mail, so > be > >> it. We are not about distributed authoring over e-mail. We are > about > >> distributed authoring over HTTP. It is very nice to keep in > >> consideration e-mail issues but we should not randomize the group > by > >> requiring ourselves to maintain interoperability. The last > requirement > >> should be removed. > > I agree with this sentiment. However, a requirement a store and > forward > transport scheme such as email system may be used to address is > consistency > maintenance in replicated versioned stores. We would like to see this > requirement be considered. > > Sankar Virdhagriswaran p. no: 508 371 0404
Received on Tuesday, 20 May 1997 21:17:26 UTC