Re: Labels

From: Geoffrey M. Clemm (geoffrey.clemm@rational.com)
Date: Thu, May 04 2000

  • Next message: Geoffrey M. Clemm: "Re: Locking" 

    Date: Thu, 4 May 2000 22:58:53 -0400 (EDT)
Message-Id: <200005050258.WAA15957@tantalum.atria.com>
From: "Geoffrey M. Clemm" <geoffrey.clemm@rational.com>
To: ietf-dav-versioning@w3.org
Subject: Re: Labels


   From: "Tim Ellison/OTT/OTI" <Tim_Ellison@oti.com>

   <geoff>
      I believe this is a better model than adding and removing
      a label from a revision, because it makes explicit the two
      key semantic properties of a label:

      - when you put a label on a revision, it removes it from the
	revision that currently has that label (if any)

   <tim>
   I infered that if I did a SET-TARGET and added an existing label to a new
   revision it would be *im*plicitly removed from the other revision.  There
   is no indication that I must explicitly remove it first.
   </tim>

Good thing, because you *don't* explicitly remove it first (:-).
Your inference that it would be implicitly removed from the
other revision was the correct one.

      - you use a label in a Target-Selector header to select a
	particular revision.

   <tim>
   The srewed observer would also notice the keyword "label" in the
   Target-Selector ;-)
   </tim>

Since the definition of SET-TARGET is:

"A SET-TARGET request can be applied to a versioned resource whose
target is a revision to specify that revision to be the target
when the specified label is used in a Target-Selector header",

I don't think you'd have to be that shrewd to expect "label"
to be one of the choices for a Target-Selector header (:-).

   <tim>
   My claim is only that it is not inferable from the SET-TARGET method
   either, but at least LABEL is the obvious place to look in the spec. when
   trying to understand set/get/remove labels.  (I'm not proposing a change in
   semantics.)
   </tim>

My claim is that the underlying model shouldn't be set/get/remove label,
but rather "set-target rev-xx" and "set-target none".  This provides a
consistent way to set a "label" target and set the "default" target.

But as you say, we agree on the semantics, and are only debating how
those semantics should be marshalled.

Cheers,
Geoff