W3C home > Mailing lists > Public > www-ws-desc@w3.org > October 2003

RE: RE: RPC Style Issues (3)

From: Jonathan Marsh <jmarsh@microsoft.com>
Date: Fri, 31 Oct 2003 12:07:39 -0800
Message-ID: <DF1BAFBC28DF694A823C9A8400E71EA201A02CDA@RED-MSG-30.redmond.corp.microsoft.com>
To: "Sanjiva Weerawarana" <sanjiva@watson.ibm.com>, <UMIT.YALCINALP@ORACLE.COM>, "Jeffrey Schlimmer" <jeffsch@windows.microsoft.com>, <jacek.kopecky@systinet.com>
Cc: <tomj@macromedia.com>, <www-ws-desc@w3.org>

I think it's even weaker than that.  If a binding is wrong, you probably
can't communicate effectively with the Web service.  That sounds like a
fatal error to most WSDL processors.

But if the messages don't match the style the author has asserted, this
does not imply that you can't use the Web service just fine.  Your tool
might warn you that "the author of this WSDL lied when he asserted the
style blah, so the stubs I'm generating may be in different forms than
the author intended."  That doesn't sound fatal to me.

It should not be a conformance violation to build a tool that only
accepts WSDLs marked and verified to be RPC style, though marketing such
a tool to customers who want to consume arbitrary WSDL will be
challenging.  It should not be a conformance violation to build a tool
that ignores RPC style and generates its stubs through some other
heuristic, though marketing such a tool to customers who want the
authors intensions followed in their function signatures will be
challenging.  It should not be a conformance violation to build a tool
that recognizes RPC style, but doesn't "honor" all the programming hints
(e.g. munges operation names to match a particular coding convention),
though marketing such a tool to customers would entail convincing them
that the heuristics are better than following the hints to the letter.
None of these approaches impede interop at the message level. 

> -----Original Message-----
> From: www-ws-desc-request@w3.org [mailto:www-ws-desc-request@w3.org]
On
> Behalf Of Sanjiva Weerawarana
> Sent: Friday, October 31, 2003 11:45 AM
> To: UMIT.YALCINALP@ORACLE.COM; Jeffrey Schlimmer;
> jacek.kopecky@systinet.com
> Cc: tomj@macromedia.com; www-ws-desc@w3.org
> Subject: Re: RE: RPC Style Issues (3)
> 
> 
> Yes, we're *only* talking about message conformance. However,
> the way do that is by *first* defining a schema and *then* saying
> "oh BTW, I followed certain rules when I defined that schema, if
> you care to know." So, interop is guaranteed purely by whether you
> conform to the schema or not, period.
> 
> *If* you do care to notice the fact that the author followed the RPC
> rules in defining the schema, *then* you have an obligation to
> verify that the guy didn't screw it up. This does lead to a kind of
> interop problem, but the problem is that the WSDL is simply wrong
> (for people who wish to pay attention to the style=RPC part), and
> thereby non-interoperable ;-).
> 
> I agree with Jeff and others that this is analogous to the model we
> have for bindings. *If* you care to use the HTTP binding in a WSDL
> then you may complain that the binding has some flaws. However,
> if you only care to use the SOAP binding, then you will not examine
> the HTTP binding nor complain about its flaws. Similarly, if you do
> not care to pay attention to the fact that the author claims certain
> rules were followed in defining schemas, then the you are not
obligated
> to verify that the author is not a two-faced liar.
> 
> Sanjiva.
> 
Received on Friday, 31 October 2003 15:07:18 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Monday, 7 December 2009 10:58:27 GMT