From: jamsden@us.ibm.com To: ietf-dav-versioning@w3.org Message-ID: <8525690E.005538E1.00@d54mta02.raleigh.ibm.com> Date: Fri, 30 Jun 2000 11:31:58 -0400 Subject: Branching, repositories, and activities <geoff> We spent some time on the CHECKOUT/CHECKIN proposal. There are several motivations for this proposal. One of them is support simple branching models such as that of RCS, where a versioned resource has simple activities called "branches", but a branch is always relative to a particular versioned resource (i.e. you create a branch for a particular versioned resource). Another motivation is that if a server supports multiple repositories, and an activity in one repository can only be used for checkouts and checkins in that repository, then a client needs to let the server know somehow which repository the activity should be created in. Another motivation is that even if "multi-repository" activity membership is supported, that it is more efficient if the activity is located in the same repository as "most" of its revisions. </geoff> <jra> The repository will be identified by the URL the user gives for the activity being created. That is, the activity resource will be created in some collection which resides in a repository. What, above activities, needs to be added to support these simple branches? Or are you proposing that branching be supported directly as "simple activities" for servers that don't want to support full activity semantics? I think branching is an implementation mechanism supporting parallel development or multiple development lines-of-descent while activities are a more logical way of expressing the same problems. I don't think we need two ways of expressing these problems, and servers are free to implement activities using branching. </jra>