Re: configurations vs. deep revisions (aka baselines)
Bradley Sergeant (Bradley.Sergeant@merant.com)
Mon, 4 Oct 1999 11:36:33 -0700
Message-ID: <F3B2A0DB2FC1D211A511006097FFDDF534386A@BEAVMAIL>
From: Bradley Sergeant <Bradley.Sergeant@merant.com>
To: "'Geoffrey M. Clemm'" <gclemm@tantalum.atria.com>,
Date: Mon, 4 Oct 1999 11:36:33 -0700
Subject: RE: configurations vs. deep revisions (aka baselines)
Perhaps we could then call each revision of a versioned configuration a
baseline.
-----Original Message-----
From: Geoffrey M. Clemm [mailto:gclemm@tantalum.atria.com]
Sent: Friday, October 01, 1999 4:38 AM
To: ietf-dav-versioning@w3.org
Subject: configurations vs. deep revisions (aka
baselines)
One of the issues that we have faced is a clash between the
protocol
for manipulating configurations and the protocol for
manipulating
deep revisions of collections (currently called "baselines"
in the
protocol). In particular, we have been saying that a deep
revision
is a special kind of configuration, but while a deep
revision is just
an immutable object that can be created and used in revision
selection
rules, a configuration is a mutable object with members that
can be
added and deleted.
I propose that we change the specification to remove any
implied
"inheritance" relationship between these two kinds of
resources.
So a deep revision is not a configuration, but rather is
just
a special kind of revision.
I also propose that we replace the term "baseline" with the
term that
Jeff has always preferred, namely "deep revision". Although
the
former term is intuitive to many/most advanced CM users, I
don't find
that it is meaningful to the wider audience targetted by
DeltaV.
In contrast, deep revision emphasizes the key aspects of the
concept,
namely that it is immutable (a revision) and it refers to
all members
of a collection (deep).
In this context, I am willing/happy to remove my earlier
objection
to saying that a configuration is a collection (I can hear
Jim Amsden
chuckling in the background, and probably Brad Sergeant as
well :-).
The members of collection are revisions, and the binding
names of
a member is the GUID of the history resource that contains
that revision
(as Brad proposed on Monday). Then you just use normal
BIND/DELETE
requests to add and delete revisions to a configuration.
If there are no violent objections, I'll make a pass through
the protocol
making these changes so we can see what this would look
like.
Jeff,Brad: Could you mail me the current state of your
changes to
the protocol, so I can minimize the "merge" that I'll need
to do
to combine your changes with the ones I propose to make?
Thanks!
Cheers,
Geoff