Re: Processor conformance: fault on non-conformant input

Here is a use case I have. What if the binding was both not used and 
illegally formed because the user made typing errors and the user 
expects that the tool (the WSDL processor) would parse and flag such 
conditions so that the user can fix it. On the other hand the user 
downloaded a WSDL from a registry and would like to work with it and 
does not care about certain parts of the WSDL that meet the above 
criteria and would like the processor not to fault on the unused binding 
parts.  Here IMO we should not require that the processor faults but 
allow it to fault. This behavior should be driven by the user preference 
rather than from a flat ruling from the spec, IMO. That is, don't put a 
MUST or MUST NOT fail :)

I know this conflicts with a clean conformance spec on the processor 
goal but, usability has more importance here IMO.

I think I am agreeing with Jacek here ..

Regards, Prasad

-------- Original Message --------
Subject: 	Re: Processor conformance: fault on non-conformant input
Resent-Date: 	Fri, 19 Mar 2004 11:34:13 -0500 (EST)
Resent-From: 	www-ws-desc@w3.org
Date: 	Fri, 19 Mar 2004 11:34:06 -0500
From: 	David Booth <dbooth@w3.org>
Reply-To: 	David Booth <dbooth@w3.org>
To: 	Sanjiva Weerawarana <sanjiva@watson.ibm.com>, <www-ws-desc@w3.org>
CC: 	Jacek Kopecky <jacek.kopecky@systinet.com>
References: 	<5.1.0.14.2.20040318120010.03a40bd8@localhost> 
<5.1.0.14.2.20040318145802.03bd3438@localhost>


Sanjiva,

Oh!  I assumed we'd want to be consistent with our treatment of 
unrecognized required extensions, but I guess we should ask the rest of the WG.

BACKGROUND
In the case of required extensions, we do NOT currently require a 
conformant WSDL processor to fault if it encounters an unrecognized 
required extension that appears in a part of the document that the 
processor doesn't need (for example, in a different binding).

THE QUESTION
If a *part* of a WSDL document is not conformant with the spec, but the 
WSDL processor doesn't need or care about that part (for example, it may be 
in a different binding that the one being used), should a conformant 
processor be required to fault?

What do others think?


At 08:21 AM 3/19/2004 +0600, Sanjiva Weerawarana wrote:
>. . .
> > and then change the newly added bullet item to:
> > [[
> > A conformant WSDL processor MUST fault if a portion of a WSDL document is
> > illegal according to this specification and the WSDL processor attempts to
> > process that portion.
> > ]]
>
>I don't agree with the text - if a part of a WSDL document is *illegal*
>then the whole thing should fail. If there are parts that are not
>understood we already have ways of dealing with it (effectively by
>invalidating the parent wsdl namespace'd component) but if the doc
>is illegal (e.g., a broken QName reference exists) then I don't think
>any processor has any business processing such a broken beast.
>. . . .

-- 
David Booth
W3C Fellow / Hewlett-Packard
Telephone: +1.617.253.1273