Re: [regrep] [Meeting Minutes] XML Schema Spec review special meeting1/26/2005

[Farrukh Najmi <Farrukh.Najmi@Sun.COM>]

TC members search for <important> in this message where your opinion is 
being explicitly asked. Thanks.

Richard Martell wrote:

>
>
> Sorry I missed this telecon. A few more comments regarding rim.xsd
> (v 1.18 2005/01/03 01:21:57). Most of these are less-than-minor
> advisories; the first two items are more significant.
>
Thanks Richard. Please see comments inline below...

>
> * [MAJOR] minOccurs="1" for xsd:any in QueryExpressionType prohibits
>   direct use of simple query strings.
>
>   minOccurs should be "0" to allow direct use of query strings (e.g.
>   XPath, XQuery--we also use these in our implementation). The mixed
>   content model will allow such character content.
>
>   <any namespace="##other" minOccurs="0" maxOccurs="1" />
>
>   Also note that @processContents is dropped--the default value is
>   "strict" to require that an element declaration be available to the
>   processor for validation purposes (even if validation is not actually
>   performed in all cases).
>
>   <documentation xml:lang="en">
>   A query specified using any supported query language, as identified
>   by the queryLanguage attribute. The mixed content model also
>   accommodates queries expressed using character data.
>   </documentation>

I will need to discuss this over the phone with you as I do not 
understand the issue yet.
Lets try and sync over IM or phone on Monday. Thanks.

>
> * [MINOR]: Empty id in RegistryObjectType. A zero-length string is
>   evidently _not_ a legal URI. The XML Schema spec (Part 2, 3.2.17.1)
>   defers to RFC 2396 <http://www.ietf.org/rfc/rfc2396.txt>.
>
>   The BNF is definitive. See Appendix A (p.26), in particular:
>
>   rel_segment   = 1*( unreserved | escaped |
>                           ";" | "@" | "&" | "=" | "+" | "$" | "," )
>
>   Hence a zero-length character sequence is not a legal URI value.
>
>   Perhaps just amend the documentation to indicate that URI values
>   (relative or not) that do not reflect the canonical URN syntax will
>   be replaced with URNs generated in the usual manner.

I will need to discuss this over the phone with you as I do not 
understand the issue yet.
Lets try and sync over IM or phone on Monday. Thanks.

>
>
> * [ADVISORY]: prune the lang attrib commented out in LocalizedStringType

+1. Fixed for draft 02.

>
> * [ADVISORY]: cardinality constraints on sequence compositor in
>   IdentifiableType appear to be unnecessary

+1. Fixed for draft 02.

>
> * [ADVISORY]: Is the comment about @code in ClassificationNodeType still
>   pertinent? The attribute remains optional, presumably to accommodate
>   taxonomies that are not coding schemes. If so, the comment can
>   probably be removed.

<important>
Thanks for reminding. This is still an open question in my mind.
What is the purspose of a ClassificationNode without a code attribute.
I truly cannot recall. Can any one else?
</important>

>
> * [ADVISORY]: Typo in documentation for AdhocQuery. Perhaps amend text
>   as follows to be more informative:
>
>   <documentation>
>   Encapsulates all registry queries. A QueryExpression element is
>   not required when invoking a stored query.
>    </documentation>

+1. Fixed for draft 02.

>
> * [ADVISORY]: @selector in SubscriptionType is a bit ambiguous. Perhaps
>   rename this to eventSelector or (queryRef).

It is too late IMO for this type of change in RIM with ripple effect 
into XML Schema, SQL Schema and implementations for the value it 
provides. I suggest no change on this.

Thanks.

-- 
Regards,
Farrukh