Re: Another try at HTTP binding

Hi Jean-Jacques.

Please find a few comments below.

* Jean-Jacques Moreau <jean-jacques.moreau@crf.canon.fr> [2004-03-03 09:37+0100]
> Thanks to the plenary, I have now been able to read your updated HTTP binding.
> 
> This really looks great! It's a big improvement to the current HTTP binding,
> both in terms of functionality and description. The new structure and much
> clearer and modular.
> 
> So, my proposal would be to incorporate your revised HTTP as is, and add your
> issues into the issues list.

* Jean-Jacques Moreau <jean-jacques.moreau@crf.canon.fr> [2004-03-09 13:01+0100]
> Your suggested changes are now in Part 3, for everyone to review 
> (changes are in red [text deleted], green [new text] and orange [text 
> slightly modified]).

It seems that the URIPath issue went away during your merge. It should
be left in the spec as it is asking for user and reader feedback.

Regarding issue 153, I think that it is more natural to have the URI
resolution be relative to the base URI property of the Infoset. XML
Base could be used to declare a set of endpoint locations relative to
the service URI. More generally speaking, it would be useful to have a
section in Part 1 about relative URI resolution mentioning XML Base,
and have Part 3 refer to it.

I notice that you left out most of 1.1 Origin and Goals. It provides
some useful background information about the HTTP binding. It also
discusses safe interactions and the HTTP GET method which is useful
IMO.

Talking about the safety property, it would be interesting to link it
with the method chosen for the HTTP binding. Right now, the spec says:

  It is a fatal error if there is no method attribute information item
  or methodDefault attribute information item.

We could default the method to get if the operation bound is safe, and
uses the URI style.

Regards,

Hugo

-- 
Hugo Haas - W3C
mailto:hugo@w3.org - http://www.w3.org/People/Hugo/