RE: collection with ordered members

At 01:32 PM 10/23/97 PDT, Yaron Goland wrote:
>What happens if you DELETE a compound document on a server which models
>compound documents as a collection but does not support TREE methods?
>That is why modeling compound documents as a collection is just wrong.
>
You are right that collections can't represent compound documents
effectively unless the TREE methods are implemented.  In my heart of
hearts, I think those methods should be mandatory.  Anyhow, if they are
optional, that makes support for compound documents optional -- possible
only on servers that have the TREE methods.

>Simply PUT the document to the server. If the server understands the
>structure of the compound document then INDEX will work on the document.
>You can even PUT into its namespace to add members. At that point you
>add headers to specify ordering information. This works on ALL DAV
>servers, does not require support of TREE methods, fully supports
>ordering, and doesn't require we change a line of the DAV spec.

I agree that it's possible to represent compound documents, given the
infrastructure WebDAV provides.  But what I want is for a way of
representing compound documents to be standardized.

I think we would have what we need if we add ordering to collections, and
define a body for MKCOL.  

Or if you prefer it, we could define a body for PUT that servers should
understand as a compound document, assigning a URL to each component as
well as to the whole, and requiring them to support INDEX for these
documents.  We could define the headers to be used for ordering.  We could
define how MOVE, COPY, and DELETE work for such compound documents.  It
just seems more straightforward to work collections, since we've got all
these basic behaviors defined for them.

--Judy

Received on Friday, 24 October 1997 09:45:52 UTC