STC and Xerces

[Ray Plante]

Hi Arnold,

I've been looking into the errors reported by Xerces on STC that feature
the message, "The particle of the type is not a valid restriction of the
particle of the base", and I have good news and bad news.  The pattern at
issue is one of restricting an element by replacing a subelement with a 
specific member of the subelement's substitution group.

The bad news is that the pattern as you have applied it is indeed not
allowed by the standard (which Xerces faithfully enforces), even
though it was intended to be legal.  Combing through list archives,
this was apparently recently discovered (by Simon Cox of CSIRO, as
best as I can tell) and is considered by those In The Know as an error
in the XML Schema standard.  As one writer put it, "it's one of those
cases where [a parser] implementor has to choose between doing what
the spec says and doing what it means"
(http://lists.w3.org/Archives/Public/xmlschema-dev/2004Oct/0100.html).
Restriction via a substitution group replacement is only allowed when 
minOccurs and maxOccurs on the subelement are both equal to 1.  

For the record, the problem arises from a conflict between two rules that
must be applied for judging if this is a valid restriction:
http://www.w3.org/TR/2004/REC-xmlschema-1-20041028/#cos-particle-restrict
(2.1) applied to the restrictions base type and 
http://www.w3.org/TR/2004/REC-xmlschema-1-20041028/#rcase-RecurseAsIfGroup
applied to the restricted type.  (Xerces' reporting of
rcase-RecurseLax.2 and derivation-ok-restriction.5.4.2 is somewhat
misleading.)  A good discussion of the problem can be found on w3c's
xmlschema-dev list in a thread that starts with
http://lists.w3.org/Archives/Public/xmlschema-dev/2004Oct/0097.html.
I believe the first mention of the problem was by Simon Cox in 
http://www.stylusstudio.com/SSDN/default.asp?action=9&read=2948&fid=23&FirstTopic=1&LastTopic=9999
in the context of the Geography Markup Language (GML) which employs 
the same pattern.

The good news is that there is a work around.  This was worked out for
GML and is described in 
http://lists.w3.org/Archives/Public/xmlschema-dev/2004Oct/0103.html.

BTW, I think there are some other compliance issues that need being
looked at.  More on that later.  

cheers,
Ray