coming clean with the SGML crowd (was re: namespaces)

Richard L. Goerwitz III richard at
Fri Sep 11 00:32:23 BST 1998

> Q:  "Why not associate namespaces optionally with DTDs (not necessarily
> via the name-space URI)?"
> A:	This was discussed extensively during the design, and rejected for
> two reasons: First, it was not the minimal necessary to enable namespaces.

You pay now or you pay later.  A minimal namespace spec _now_ means that
DTDs won't know about namespaces - and that, in order to use DTDs, we will
have to follow a silly, kludgy process of expanding all our namespaces,
and then editing the expanded elements back into our DTDs.  In other words,
we pay later.

Note that in this scheme, the SGML compatibility crowd pays more than the
rest of us.  The more namespaces are actually used (and the more they make
the process of writing DTDs that much less direct and useful), the more
these silly kludges become necessary, and the more XML loses its connec-
tion with SGML.

It's unbelievable, but people are already talking about legacy XML and
about the abandonment of SGML.  Is it any wonder that vendors like Micro-
soft seem to be sitting on the fence?  We have some people who declare
firmly that DTDs are the only accepted schema/validation mechanism.  And
we have other louder, stronger voices who are already declaring them dead.
For a while it seemed the loud voices were simply disenfranchised.  But
namespaces (among other things) have shown them to be legitimate.  They're
dealing with the problem.

Maybe it's time to admit to the poor SGML crowd what's really going on,
by the way.

> Second, there are many possible resources that could be associated with a
> namespace.  DTDs are one, but various forms of schemas, style sheets,
> documentation, etc. are also likely.

I don't get your point.  If there's no DTD associated with a namespace,
then we're back where we started.  You either kludge the DTD by tipping in
the namespace elements, or you give up validation.  It's exactly where we
are with the current spec.

So make the DTD optional, and invent a way to tell the processor that it
is a DTD.  There's no reason that we couldn't associate other schemas with
namespaces too.  Each may use its own validation method.  Leaves room for
growth.  And if there is no DTD associated with a given namespace, then you
either kludge your main DTD or live without validation.

Yes, this breaks SGML compatibility in the sense that an SGML-based val-
idator won't pick up the namespace DTDs.  But SGML compatibility is be-
coming more and more superficial in XML anyway - namespaces and their in-
teraction with DTDs being a case in point.

I hear chanting:  "Reverse course; where are architectural forms?"  I hear
raving:  "The whole thing is a mess; log live HTML."  I hear musing:  "Ar-
gue about this long enough and nobody will be using old browsers, and we
won't have any more reason to shun the old PI solution for namespaces."

xml-dev: A list for W3C XML Developers. To post, mailto:xml-dev at
Archived as:
To (un)subscribe, mailto:majordomo at the following message;
(un)subscribe xml-dev
To subscribe to the digests, mailto:majordomo at the following message;
subscribe xml-dev-digest
List coordinator, Henry Rzepa (mailto:rzepa at

More information about the Xml-dev mailing list