Re: Meaning of normative references [was: Update on namespaces]

Dan is of course perfectly correct here:
in the terminology of most standards and specifications, a normative
reference to another publication means that everything in that publication
applies to this standard, and every implementation has to conform to
that publication.  The implication is that the referencing standard is
incomplete, and you have to read the other application.

We already have, I think, a normative reference to ISO 10646 or Unicode.

We should not have a normative reference to ISO 8879.

Instead, we should have a note that says that every valid XML document
will ipso facto conform to ISO 8879, and possibly refer the interested
reader to that document.

I think, though, that it's just in the way that the terminology is
being used.  I don't expect that Len wants all XML parser writers to
read and understand SGML before starting out on their heroic 
One Week Journey of Implementation...  I recall James Clark at SGML 96
mentioning that it took him a few months to get to the point where he
understood the SGML standards well enough to begin coding.  I doubt
that this is atypical, except that few people can devote themselves
to such study full-time...

So this fails the "sanity test"....  any reference to SGML in the
XML spec must not require that people read SGML.  Hence, SGML conformance
cannot be a normative requirement for XML conformance.  Rather, the
XML spec must stand alone, but the resulting valid XML documents must
always be valid SGML documents.  So any reference to SGML in the XML
spec must be for interest, not one that forces people to read ISO 8879:1986.

I am saying the same thing in multiple ways in the hope that everyone
can agree with at least one of them easily, and then see that they're
all the same.

Lee

Received on Tuesday, 24 June 1997 22:02:05 UTC