RE: RE: Collections

Hum.... my "oy yevah" alert just went off. In a world of dynamically
generated namespaces can we make requirements regarding the structure of
the namespace?

I think the answer is yes but has to be in the context of particular
methods. One of the things I did with the language in the new namespace
draft is to put in requirements, on particular methods, that state "The
namespace MUST be consistent after the execution of this method." Where
consistent means that if there is an /a/b/c then there is an /a and an
/a/b.

So if a server does not want to support consistent namespaces then it
MUST NOT support methods like COPY, MOVE, or most importantly, INDEX.

			Yaron

> -----Original Message-----
> From:	Del Jensen [SMTP:dcjensen@novell.com]
> Sent:	Friday, September 19, 1997 1:44 PM
> To:	johnt@cgocable.net; ejw@ics.uci.edu
> Cc:	w3c-dist-auth@w3.org
> Subject:	Re: RE: Collections
> 
> What kind of language ("MUST" or "SHOULD") most effectively promotes
> interoperability in this case?  Why use the protocol if you are going
> to ignore the object model?
> 
>     Del
> 
> >>> Jim Whitehead <ejw@ics.uci.edu> 09/19/97 11:54AM >>>
> 
> John Turner writes:
> > If the the whole hierarchy is not treated as a collection, you leave
> open
> > the possibility of some user confusion.  The namespaces of the URLs
> and 
> the
> > collections overlap, but are not identical.  For example, if you
> have
> > collection /A you could have /A/B where B is an internal reference
> and 
> /A/C
> > where C is not part of the collection, simply part of the URL space.
> 
>  This
> > might be useful, but it will certainly be confusing.
> 
> My assumption has always been that any collection-like object in a DAV
> 
> server's namespace should be modeled using a WebDAV collection -- an 
> assumption so deeply ingrained none of the authors ever thought it was
> 
> worth writing down. (Bzzt! :-)
> 
> I agree a requirement should go into the specification stating that 
> collection-like objects SHOULD be WebDAV collections.  My question is 
> whether this should be a MUST or a SHOULD requirement.
> 
> I'm leaning towards SHOULD because there are cases where part of a
> server's 
> namespace is computed, and hence can be potentially infinite.  For
> example, 
> if a server has a part of its namespace which acts as input to a cgi
> script 
> (e.g., http://www.foo.org/finger-script/cgi-bin/user@host) what should
> 
> INDEX return for the /cgi-bin/ collection if there is a MUST
> requirement?
> 
> On the other hand, if it was a MUST requirement, there could be
> language 
> added which states that a server is not required to expose every
> member of 
> a collection where the membership depends on the current request.
> Although 
> I'm unsure what the best way would be to express this, or whether this
> 
> causes more problems than it solves.
> 
> - Jim
>

Received on Sunday, 21 September 1997 14:55:14 UTC