W3C home > Mailing lists > Public > ietf-dav-versioning@w3.org > January to March 2001

Re: disjoint baselines

From: Greg Stein <gstein@lyra.org>
Date: Sun, 18 Feb 2001 18:09:29 -0800
To: ietf-dav-versioning@w3.org
Message-ID: <20010218180929.K29904@lyra.org>
On Sun, Feb 18, 2001 at 02:39:49PM +0000, Tim_Ellison@uk.ibm.com wrote:
> > Consider that I have two resources:
> >
> >     /non-vcr-1/vcr-a/foo
> >     /non-vcr-2/vcr-b/bar
> Assuming foo and bar are VCRs themselves.

Of course :-)

> > I put both of these resources under baseline control and
> > create a baseline from them. What does the DAV:baseline-collection
> > collection look like? Does it include "filler" collections to
> > reach the VCRs that are under baseline control?
> >
> > Concretely, would it look something like:
> >
> >     /BCs/b72/non-vcr-1/vcr-a/foo
> >     /BCs/b72/non-vcr-2/vcr-b/bar
> Yes, this is required for the baseline to capture the namespace of the
> configuration.

Euh. Baselines capture a set of versions. Who said they capture the layout
of the namespace in which the VCRs appear?

Given my current understanding of baselines, I don't believe your statement
necessarily follows.

Oops. Reset. Somebody changed the definition on me in the past couple months
:-) ... I just read the intro to the baseline section. "... and their names
relative to the collection, ..."


> > And would non-vcr-* contain *only* the children necessary to reach the
> baselined values?
> I'm planning on 'yes', otherwise these would be empty, filler collections
> -- not very useful.


> > Same applies to "/" -- does it contain just non-vcr-*, or would
> > /BCs/b72/ contain other bindings?
> only those that were required to reach VCRs.


> > [ this doesn't apply in the Subversion scenario, so I'm not
> > personally worried, but it appears to be a "hole" in the draft. ]
> >
> > Note that my property proposal would resolve the situation. Each
> > VCR under baseline control would have a path to its corresponding
> > VCR in the BC.
> Agreed.  Though I've not figured out yet how the proposed property works
> when the VCR is a member of multiple baseline-controlled collections.

Right. I'm thinking: toss the property proposal, depend upon locate-history.

> It is an important property of baselines that they capture the namespace of
> the configuration, so collapsing the namespace this way into the 'disjoint
> sets' would be a Bad Thing.

Seeing the current definition of "baseline" ... I agree.

Okay... so maybe I'll go write a FAQ entry -- non-versioned collections can
appear in a BC as a way to reach the VCRs.


Greg Stein, http://www.lyra.org/
Received on Sunday, 18 February 2001 21:07:00 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Tuesday, 8 January 2008 13:57:40 GMT