Re: [SpecGL-impl] xml:id Last Call

I just re-skimmed the xml:id spec and I lean toward Lofton's view about 
Class of Product. Even Chapter 7 (Conformance) doesn't really nail down 
what I expect: what is the test subject to which a conformance test would 
be applied? I think that xml:id is actually an optional module for an XML 
parser that happens to have its own spec. Parsers may be validating 
parsers or not, but xml:id attempts to provide a validation outcome even 
for non-validating parsers. If SpecGL had been in force 8 years ago, XML 
Core documents (with add-ons such as Xinclude, xml:base, xml:id) would 
have nicely arranged everything into modules and (maybe) levels, and it 
would be a lot easier to classify an XML parser by which modules it 
implements. We don't have that clarity from the past, so we have to start 
now to clarify the accreted specs.

I don't think xml:id imposes new conformance rules on XML documents. 
Earlier comments in this thread suggest that it would, which shows that 
there is some confusion about the conformance requirements of xml:id. The 
XML Schema WG might want to amend their documents to require that an 
attribute named xml:id must have type ID, but then the conformance 
requirements of Schema would be affected. Perhaps the XML Coordination 
Group did not exert a strong enough influence to align all the affected 
specs, but the reader of xml:id should not be required to investigate that 
broadly.
...............David Marston

Received on Wednesday, 5 January 2005 15:29:33 UTC