Revision identifier and revisions label namespace sharing

Geoffrey M. Clemm (gclemm@tantalum.atria.com)
Tue, 5 Oct 1999 13:03:36 -0400 

Date: Tue, 5 Oct 1999 13:03:36 -0400
Message-Id: <9910051703.AA14121@tantalum>
From: "Geoffrey M. Clemm" <gclemm@tantalum.atria.com>
To: ietf-dav-versioning@w3.org
In-Reply-To: <1999Oct05.123200.1250.1342211@otismtp.ott.oti.com>
Subject: Revision identifier and revisions label namespace sharing


<gmc/> Some of the earlier drafts of the protocol did require that
revision identifiers and revision labels share the same namespace.
Since the revision identifiers are server-defined, this provides a
significant interoperability barrier, since a client would have to
know the revision identifier naming conventions of all servers it
might work against in order to avoid a name clash between a label it
wants to create and all server identifiers that it might clash with.

<gmc/> So I believe the current protocol is correct in not sharing
namespaces between revision identifiers and revision labels.  This
should be made clear in the protocol specification.

   From: Jeff_McAffer@oti.com (Jeff McAffer OTT)

   <tpe>
   I can easily imagine revision selectors being Integers -- these would
   collide with labels that are equivalent String form of Integers;
   i.e.
     revisionid: 3
     label: 3
   </tpe>

   <jra>
   The revision could never been labeled "3" as this conflicts
   with an existing revision name.
   </jra>

   <jm>
   A couple things to note.
   1) The spec currently does not state this.  revision-ids are described as   
   unique amongst the set of other revision-ids on a resource but labels are   
   much more weakly defined.

<gmc/> Labels are also guaranteed to be unique amongst the set of other
labels on a resource (since it is a revision name, and revision names
are defined to have this property).

   2) If names are in fact unique, then why do we have rsr-label and   
   rsr-revision-id elements?  There should just be an rsr-name element as a   
   names are labels or revision-ids.
   </jm>  

<gmc/> These elements were defined under the assumption that they are
separate namespaces.  If we changed that assumption, I agree that we
should just have an rsr-name element.

Cheers,
Geoff