RE: [w3c sml] sml:ref and interoperability

I'm inclined to agree with the concerns that Ginny lays out below. I suppose we all knew interop wouldn't be easy, but at times it does seem to be slipping away. That said, I would like to make sure I am clear on the following:

 *
The proposal in comment #2 seems to me to be more about ref identification, not validation. So to me this primarily affects the interpretation of the model rather than the validation.
 *
Comment #6 bullet 1 focuses more on consistent validation between two different implementations and bullet 2 points out the same issues as in comment #2.

I suppose both aspects are important to interop, I tend to focus more on the interpretation of the models, i,.e. identification of the references, so I would be inclined to say we make this the goal of interop (regarding this particular issue).

Jim


________________________________
From: public-sml-request@w3.org [mailto:public-sml-request@w3.org] On Behalf Of Smith, Virginia (HP Software)
Sent: Wednesday, May 21, 2008 10:11 PM
To: public-sml@w3.org
Subject: [w3c sml] sml:ref and interoperability


I've looked at http://www.w3.org/Bugs/Public/show_bug.cgi?id=5541 and, after considering comment #6, bullets 1 and 2, I have serious reservations.



As you all know, I believe that interoperability is a very important part of SML and, to some degree, its "raison d'être". Leaving it up to each SML validator whether or not to recognize SML references based on XML infoset or PSVI destroys interoperability. I am ok with the interoperability issues as detailed in section 4.5 (http://dev.w3.org/cvsweb/~checkout~/2007/xml/sml/build/sml-if.html?content-type=text/html;%20charset=utf-8#Interoperability) because those issues are under the control of the SML-IF producer. However, this new twist means that interoperability is now a crapshoot and there is no alternative available to the SML-IF producer to guarantee interoperability.



I would appreciate hearing any thoughts on why this (allowing either/or) is a good idea. I understand the issue of allowing sml:ref to be specified in the schema - that is a valid issue and I think my preference is to allow that option. However, interoperability is key and I don't think we gave that sufficient consideration.



--

ginny