Re: Optional SOAP response; the transfer binding view

On 1/11/06, noah_mendelsohn@us.ibm.com <noah_mendelsohn@us.ibm.com> wrote:
> I do note one detail not covered in Dave's current draft, that I expect
> should be in the future.  RFC 2616 seems to imply that an entity is
> required on POST [1].

FYI, though it doesn't seem to matter to this issue, there's an errata in 9.5;

http://skrb.org/ietf/http_errata.html#post

FWIW though, I don't think an entity should be required, because;

1) if a server/resource wants to interpret a message with no entity in
some particular way, it should be allowed to ("The actual function
performed by the POST method is determined by the server")
2) an HTTP message is self-descriptive with respect to message size,
so there should be no issue with actually sending POST messages with
no entity

It's a similar argument to the HTTP FAQ about why HTTP doesn't
preclude GET requests from including an entity.

> Someone who knows more of the details will have to remind me how this
> squares with application/x-www-form-urlencoded.

Not quite sure what you mean by that.

>  In any case, if we're
> shifting to Dave's proposal that the Request envelope be optional in the
> base MEP, I think we'll need to explain in the binding what HTTP does in
> the case that WebMethod is POST and no outbound message is provided.  This
> is already covered by the binding for Response-only, but we're potentially
> openning a new path, which is MEPs that are not the one we call
> "Response-only", but which in fact have no request envelope.

I don't see any problem with that optional request envelopes.

Cheers,

Mark.

Received on Wednesday, 11 January 2006 18:52:58 UTC