2006/5/10, Mark Baker <[EMAIL PROTECTED]>:

I can live with "SHOULD" on a response body being returned.  But why
MUST it be an Atom entry?  When you were trying to force a return of
the (possibly modified) requested entry on the response, that made
some sense since the request was an Atom entry.  But without that it
just seems to limit extensibility for no good reason, AFAICT... as
I've described before.

+1
This implies conflicts with HTML-form-driven POST's or 2-Way-RSS when
you want to support them at the same URI.

To be good HTTP clients, they should be checking the Content-Type
header anyway, right?  You wouldn't want them assuming it was
application/atom+xml just because a non-HTTP spec said so.

So I could live with "SHOULD be an Atom entry".

-1

I'd however be +1 to suggesting this as an optimization trick for
server implementers, in the sense that clients won't have to issue a
GET just after the POST/PUT response.

In any case, there must be something like "clients MUST NOT assume a
returned Atom Entry Document representes the newly created resource
unless there is a Content-Location whose value is identical
character-by-character (after relative-URI resolution) to the Location
header value".
I could live without the "after relative-URI resolution" part if this
is made clear that clients are allowed to expect only absolute-URIs
for comparison (i.e. if this is a relative URI, it will never be
identical character-by-character to the Location header, as this one
must have an absolute-URI value).

--
Thomas Broyer

Reply via email to