RE: Prelim. DAV spec.

I say we take the weasel way out and support both. The draft already allows for this.
			Yaron

-----Original Message-----
From:	Larry Masinter [SMTP:masinter@parc.xerox.com]
Sent:	Sunday, October 27, 1996 10:48 AM
To:	connolly@beach.w3.org
Cc:	ejw@rome.ics.uci.edu; w3c-dist-auth@w3.org
Subject:	Re: Prelim. DAV spec.

I think there's a design choice people are facing in trying to
reconcile versioning and content negotiation, and I want to argue for
one choice over another:

Under content negotiation in HTTP, a single URL
     http://myhost.dom/somestuff

connects to a single resource (the 'somestuff') but a resource can
have multiple representations; one of those representations is chosen
using content negotiation.

When we add versioning, you can say that 
   a) versions apply to resources
or
   b) versions apply to representations

and there is some temptation, for generality, to try to go down the
road (b), but I want to argue that choice (a) is the choice you should
make.

Content negotiation ALLOWS (but does not REQUIRE) that each
representation might also have its own URL, that is, a representation
of one resource might be another resource (presumably with only a
single representation).

I think if you want to version separate representations, it should
only apply in the situation where there is a separate URL for each of
those separate representations. So that 
      http://myhost.dom/somestuff
might have both an english and a french representation, but if you
want to talk about the VERSIONS of those representations, you should
have separate URLs for the resources
	http://myhost.dom/somestuff.en
	http://myhost.dom/somestuff.fr

and talk about the versions of those resources in terms of versioning.

Content negotiated objects are, in fact, a kind of compound object. If
you have a compound (a book consisting of separate chapters) then the
versioning of the subparts is not completely independent of the
versioning of the whole. Similarly, the versioning of a resource with
multiple representations must of course be linked to the versioning of
those representations, if the representations are allowed to be
versioned independently.

While we're at it, we might want to supply LINK relationships for
variants in content negotiation so that we can tell when a resource
being returned is to be considered just one variant, but that's
probably a discussion for a different mailing list.

Larry

Received on Thursday, 31 October 1996 02:22:46 UTC