[Bug 2822] RQ-144 Define processor-conformance profiles (WhichPSVIPropertiesReqd)

http://www.w3.org/Bugs/Public/show_bug.cgi?id=2822

           Summary: RQ-144 Define processor-conformance profiles
                    (WhichPSVIPropertiesReqd)
           Product: XML Schema
           Version: 1.1 only
          Platform: Other
        OS/Version: All
            Status: NEW
          Keywords: needsDrafting
          Severity: normal
          Priority: P2
         Component: Structures: XSD Part 1
        AssignedTo: ht@w3.org
        ReportedBy: cmsmcq@w3.org
         QAContact: www-xml-schema-comments@w3.org


This issue was originally reported by Noah Mendelsohn.


[N.B. this candidate requirement was originally posed in the

terms given by the email from Noah Mendelsohn cited below; it was

amended on 25 March 2004 to read as follows:]


   "[Definition:] Minimally conforming processors must completely

   and correctly implement the Schema Component Constraints, Validation

   Rules, and Schema Information Set Contributions contained in this

   specification."


This appears to require that conforming processors report the PSVI.

This would rule out, for example, processors that implement a simple

validation check function such as:


     boolean IsValid(schema, instance) 


But such a function is clearly useful in query languages,

spreadsheets, and other systems that manage or access instance

documents.  Other interesting if somewhat less common reports might be

to give only the type assignments of each element or attribute, etc.


Even for validity, some applications will want details of validity

down the entire tree, while others will want only the net result at

the root.  Interestingly, many applications will want the

"value" from a simple type value space, which for some

reason we have declined to include.


The range of useful processor APIs goes well beyond providing the full

PSVI and only that.  We should modify the spec in the light of that

fact.


* There should be a clean separation between descriptions of the

language(s) and descriptions of processors.


* We should make a good attempt to call out (in a separate chapter) a

particular specification of processor classes that we hope will be

useful and widely deployed.


* We should make clear that other processor class descriptions are OK;

they are not to be thought of as second-class compared to ours.


See mail: from Noah Mendelsohn

(http://lists.w3.org/Archives/Member/w3c-xml-schema-ig/2003Oct/0106.html)


In January 2005 it was agreed to change the title of this requirement

from *Which PSVI properties must processors report?* to *Defined

processor conformance profiles* or something similar.


This item was discussed in the meeting of 2004-03-11

(http://lists.w3.org/Archives/Member/w3c-xml-schema-ig/2004Mar/0026.html).


This item was discussed and amended in the meeting of 2004-03-25

(http://lists.w3.org/Archives/Member/w3c-xml-schema-ig/2004Mar/0133.html).

The amended requirement was classified as Req.


This item was discussed in the meeting of 2004-04-01

(http://lists.w3.org/Archives/Member/w3c-xml-schema-ig/2004Apr/0003.html).

We identified two points (originally formulated as part of RQ-142 but

moved here by WG decision):


Any specification of a class of processors (including ours) can

    require specific additional information not in the PSVI, though

    should note that interoperability is better if applications depend

    only on the properties present in the PSVI as we define it.


In our own specification of processor classes, we should be

    explicit that processors may provide additional information.

    (Or alternatively be explicit that they must not -- but the

    chair believes the WG consensus was to allow it.)


Wording to resolve this issue has been drafted and was discussed by the

Working Group at the Technical Plenary in 2005; the proposal is to be

revised by the editors in light of that discusion.

Received on Saturday, 11 February 2006 00:33:47 UTC