Re: WS-Addressing and Web architecture

Hi again,

On Wed, Aug 11, 2004 at 01:22:19PM +0200, Hugo Haas wrote:
> I think that your comments apply more to the message information
> headers, endpoint references being a tool for them.

Well, arguably I suppose, but see below...

FWIW, I do see some value with some of the headers, but consider some
of the others to be harmful, in particular Action and To.

> As mentioned in the Team comment and as an example, the current
> version of WS-Addressing does not talk about serializing certain
> information outside of the SOAP envelope. The action parameter and the
> request URI in the SOAP 1.2 HTTP binding are given as an example.

Right, and that's a very important concern which I share, but not the
only concern I was raising.

> I recognize that there may be cases where one message is sent to
> certain service which acts as a Web services gateway, dispatching
> based on addressing headers, which may be the ones you are disputing.

Yes, that's exactly right.

> However, the [destination] property is a URI, so the resource
> receiving the message is still identified by a URI.

Well, that part of my comment related to the possibility of "dispatch"
occurring via the optional WSDL data such as portType included in the
EPR.  If there is any information there, then as I understand it, it's
part of the EPR, and therefore is identifying information that isn't
part of the URI.

> I believe that the concerns that you bring up here echo discussions
> that we have had in the Web Services Architecture Working Group,
> resulting in the SOA section[1] of the Web Services Architecture
> document about the different uses of Web services.

Correct.

> > I'm not sure what action to take at this point.  I'd like to get your
> > feedback first though.
> 
> My recommendation would be, if and when W3C creates a Working Group to
> address this area, to use the usual ways to send comments to the
> Working Group, that I trust you are familiar with.

Well, I'd prefer to either a) see the WG not chartered in the first
place, or b) have my comments considered as part of the chartering.  In
particular I consider it long overdue that a Web services group charter
include a requirement that work products be consistent with Web
architecture.  The "a priori" blurb used to date, which appears to have
been intended as a surrogate for such a mandate, has demonstrated its
inadequacy in this regard on several occasions.

Cheers,

Mark.
-- 
Mark Baker.   Ottawa, Ontario, CANADA.        http://www.markbaker.ca

Received on Wednesday, 11 August 2004 12:17:31 UTC