Re: a:attributeType

[James Clark <jjc@jclark.com>]

> I think that the user should not be forced to use the XML Schema datatype
> library, which  is buggy and complicated, only to use ID/IDREF/IDREFS.

I agree.  How about the following?

We have two specs

1. A DTD Compatibility spec that defines
   1.1 Compatibility Annotations (namespace 
http://relaxng.org/ns/compatibility/annotations), and
   1.2 Compatibility Datatypes (namespace 
http://relaxng.org/ns/compatibility/datatypes)

2. A "Guidelines for using XML Schema Part 2 Datatypes" spec

The compatibility datatypes provides at least "ID", "IDREF" and "IDREFS" 
and, if we so choose, others as well.  Conformance wrt to the compatibility 
datatypes would involve checking ID-soundness.

Our guidelines spec can then say that a compatibility processor (at least 
one that is integrated with a RELAX NG processor that supports XML Schema 
Part 2 datatypes) should recognize the XML Schema Part 2 datatypes as 
aliases for the corresponding compatibility datatypes. (NOTATION would be 
an exception since the XML Schema Part 2 NOTATION is distinct from the XML 
1.0 DTD NOTATION.)

Users who just need the compatibility datatypes in their schema can use the 
compatibility datatype library. Users who also need other datatypes from 
XML Schema Part 2 can just use the XML Schema Part 2 library, without the 
inconvenience of mixing datatype libraries.

One reason I suggest this is that it makes it easy to describe crisply what 
the compatibility spec is doing: "defining annotations and a datatype 
library" rather than "defining annotations and also defining additional 
semantics for some of the datatypes from XML Schema Part 2 when used in 
conjunction RELAX NG".  I think it also keeps things nicely modular.

James