RE: DELETE Semantics

> From: "Yaron Goland (Exchange)" <yarong@Exchange.Microsoft.com>
> 
> DELETE nukes the resource.

Another point of view is that DELETE
on a URL ensures that the next GET on that URL returns the appropriate
error status.  Furthermore, unless you want downlevel clients to trash
the history of a versioned resource, DELETE *must* have UNBIND semantics
rather than DESTROY semantics.

> If the resource gets nuked so does its bindings
> since they are associated with the resource. Hence DELETE is DESTROY.

I don't see how this follows.
Embedded HREF's in xml documents are also associated
with the resource they reference, but they are unaffected by any
operation on the resource.

> BTW, I personally believe that MOVE SHOULD allow the lock to be moved. The
> reason we didn't do this had to do with supporting the majority of existing
> systems. 

In the past, I've only heard this said for Windows95.  I have not been
able to find anything in Windows95 resembling WebDAV locking behavior.
Is this some internal implementation thing not exposed to end users?
Or is this a reference to how it was implemented in Office-2000?  In either
case, although it should influence locking semantics, I don't think we should
let it determine locking semantics, if it results in a complex or confusing
protocol.  For example, with multiple bindings to a resource (say
/a/x.html and /b/y.html), if you issue a LOCK on /a/x.html, can you move
/b/y.html?

Cheers,
Geoff

Received on Friday, 24 September 1999 16:23:18 UTC