RE: TAG discussion of WS visibility issue

> -----Original Message-----
> From: Mark Baker [mailto:distobj@acm.org]
> Sent: Monday, July 14, 2003 11:03 PM
> To: www-ws-arch@w3.org
> Subject: Re: TAG discussion of WS visibility issue
> 
> 
> 
> On Mon, Jul 14, 2003 at 09:39:59PM -0400, Champion, Mike wrote:
> > I personally think that Roy Fielding summed it up quite well:
> > 
> >   Things that are universal standards are inherently more
> >   "visible" than object-specific semantics, because you don't
> >   have to go look up the non-standard semantics. It is a design
> >   trade-off. There is no point in convincing Web Services to use
> >   a uniform interface, since the whole point of WSA is to develop
> >   programmable interfaces
> 
> Exactly.  I find it curious that you chose this quote of his, because
> IMO, it's the most damning of the approach currently being taken, of
> anything in the minutes.  You realize that Roy believes that
> object-specific semantics are a mistake, right?

[link to a March 1992 post by Fielding used to support this ... BTW, I don't
think it's useful to quote Fielding's rants against SOAP from before the
XMLP WG accepted the TAG's advice and added support for the web method
feature.  SOAP 1.2 and the current WSA draft has a very different
conceptualization of Web services than the SOAP 1.1 Web-RPC idea that he was
demolishing back in the Spring of 1992.]

If one thinks of SOAP as an "object access protocol" then your point is well
taken.  If one thinks of SOAP as an XML header/extension format with a well
defined processing model that helps reduce the impedance mismatches between
different object systems, protocols, programming languages, etc., then I
think Fielding is agreeing that SOAP and XML are the "universal standards"
that can provide visibility.  

The main point I was trying to make in reproducing the quote is the idea
that this is all about engineering tradeoffs and not incommensurable
paradigms. If one is storing and retrieving documents over the Web with
HTTP, using object-specific semantics *is* generally a mistake.  But if one
is developing programmable interfaces that must work across protocols, or in
situations where the model of a resource with a known identity doesn't work
well, or in situations where the semantics of the operations are a lot more
complex and subtle than GET/PUT/DELETE, then WSA is saying, and the TAG is
apparently not disagreeing, that forcing the interaction into the REST
constraints is equally a mistake.  I see the TAG as agreeing with our
consensus: understand the tradeoffs and use the approach that is opimal in a
given situation.  Sometimes the optimum will be RESTful, sometimes it won't.


I know you disagree, but the TAG has spoken and I really, really, REALLY
think it's time to move on.  That's one trout back from the taxidermist and
up on the wall, only a few dozen left!

Received on Tuesday, 15 July 2003 01:34:39 UTC