Re: Revision names

Chris Kaler (ckaler@Exchange.Microsoft.com)
Tue, 12 Oct 1999 16:52:57 -0700


Message-ID: <FD7A762E588AD211A7BC00805FFEA54B041DD98F@HYDRANT>
From: "Chris Kaler (Exchange)" <ckaler@Exchange.Microsoft.com>
To: "'infonuovo@email.com'" <infonuovo@email.com>, jamsden@us.ibm.com,
Date: Tue, 12 Oct 1999 16:52:57 -0700
Subject: RE: Revision names


I don't think the problem is one of revision-id and revision-label being
from the same namespace so much as the requirement that they partition the
namespace.  I don't see any way for a client to know how that works without
an out-of-band agreement among all sources of revision-id and revision-label
assignments for a given DeltaV server.
<ck/> I think the problem is simple.  Given 'X', how do you know if it is
      a label or a revision id?

Is it not the case that the revision-id is exclusively assignable by the
server, so that it is always possible to avoid duplication of a revision-id
assignment?  (Requiring that a revision-id never duplicate an assigned
revision-label strikes me as raising the cost way beyond marginal utility,
but it is certainly possible to honor that if DeltaV were to require it.)
<ck/> We do not currently require revision ID to be unique on the server,
      only for a resource.

I think I am missing something about the requirements, or user models, that
are behind the different stances on this topic.
<ck/> I think I'm not communicating well :-(.  My point is simply that
      when I pass 'X' to the server, I need a way to indicate if 'X' is
      a revision id or a label because the server may need to handle the
      request very differently.