Thomas Broyer wrote:
On Sun, Nov 9, 2008 at 11:32 PM, Danny Ayers wrote:I vaguely remember early on in discussions some criticism of parts of WebDAV as being not-quite-RESTful. Could someone please remind me/point me to relevant discussion - what was wrong with WebDAV and how was it fixed in atompub?"WebDAV isn't very RESTful […], because it relies on folder operations instead of hypertext" — Roy T. Fielding, http://roy.gbiv.com/untangled/2008/no-rest-in-cmis
Hmm.I understand (and agree) that PROPFIND is problematic, as it makes the stuff behind PROPFIND (collection containment and properties) unavailable for GET. A proposed fix for that is described in <http://greenbytes.de/tech/webdav/draft-reschke-http-get-location-01.html>.
But besides that, I fail too see why an Atom feed document would be hypertext, but a PROPFIND/multistatus response wouldn't...
The "fix" in atompub is the use of POST on a collection to create/add a member, and no assumption that http://example.net/aaa/bbb is a "member resource" of http://example.net/aaa/ which would then be a "collection".
That's a separate issue. WebDAV has a containment model different from AtomPub, but that's not a bug, but by design. It's supposed to give the client control over the URL.
This has also been proposed as an extension to WebDAV by Julian Reschke in his "POST to add to WebDAV Collections" Internet Draft <http://greenbytes.de/tech/webdav/draft-reschke-webdav-post-latest.html>.
Actually, that proposal doesn't change WebDAV's containment model, it just offers a way for the server to pick the member name -- in that it's slightly different from AtomPub's POST (I may have to clarify the draft on that).
BR, Julian