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

Paul Prescod (papresco@technologist.com)
Fri, 11 Sep 1998 00:42:57 -0500


Tim Bray wrote:
>
> I repeat: all this noise about the difficulty of validation is
> completely missing the real point, namely we have neither a theoretical
> basis nor industry experience to guide us in constructing and using
> compound schemas (DTDs or any other sort), and doing partial validation.

I believe that we have both industry experience AND a formal model.

The industry experience is that people can build compound schemas by
combining schema parts through *parameter* entities. I highlight the word
*parameter* because it is the hint to how a more robust solution must
work: it must parameterize content models and schema fragments. The
problem with parameter entities is that they are way too flexible because
they work at a textual level. We have the right model, but the wrong
mechanism. So to move beyond that, we must implement parameterization in a
more structured sense.

--
The so-called "module proposal" did the same thing with SGML DTD syntax.
(it uses a not-sufficient-constrained variety of parameter entities,
however).

http://www.ornl.gov/sgml/wg8/document/1987.htm http://www.ornl.gov/sgml/wg8/document/1982.htm

--

The formal model is the forest automata theory. See:

http://lists.w3.org/Archives/Public/www-html/1998Mar/0017.html

This post shows how to do parameterization and composition.

Paul Prescod - http://itrc.uwaterloo.ca/~papresco

The past is inaccurate. Whoever lives long enough knows how much what he had seen with his own eyes becomes overgrown with rumor, legend a magnifying or belittling hearsay. "It was not like that at all!" -- he would like to exclaim, but will not, for they would have seen only his moving lips without hearing his voice. - Czeslaw Milosz (translated)