Re: Request to relax PCDATA content model restriction
Eve L. Maler wrote:
> I've been corresponding with Terry Allen, who's doing an XMLish version of
> DocBook. DocBook has a lot of content models like the following:
> where %xxx.char.mix; resolves to (#PCDATA|a|b|c...). The current XML Lang
> spec doesn't allow for this; production 45 is very specific about how many
> pairs of parens can appear, rather than going for the generalized model
> group solution of productions 40-44.
> Of course, the outermost parens aren't really necessary; you can achieve
> the desired effect with (%xxx.char.mix;)* instead of ((%xxx.char.mix;)*).
> But where content models are heavily parameterized and you can't easily see
> what kind of model you've got, this makes the DTD writer split hairs.
> It should always be safe to throw another couple of paren pairs around a
> model, and currently in XML you can't do this. Can we consider loosening
> this restriction?
I like the way it is handled now. However, I would be interested
in the way you would describe Production 45 in EBNF then. You are not
very specific about the alternative way you suggest.
I can see your point, from the perspective of a DTD designer. However,
we should try to keep things as simple as possible. (Maybe) XML
will require a slightly different approach to DTD design anyway.
Norbert H. Mikula
= SGML, DSSSL, Intra- & Internet, AI, Java