On Fri, 23 Oct 1998, Jose Abilio Oliveira Matos wrote:
> On 23-Oct-98 Jean-Marc Lasgouttes wrote:
> >>>>>> "Jose" == Jose Abilio Oliveira Matos <[EMAIL PROTECTED]>
> >>>>>> writes:
> >
> > Jose> My proposal regards the problem that we have deciding if some
> > Jose> structure (inset) allows other insets as childs or not? Do you
> > Jose> understand, now? (It was my fault if I was not clear before :)
> >
> > Sorry, I only replied to one part of youer post. I really do not know
> > much about dtd:s. So I cannot really tell whether they are
> > easy/convenient to implement/use... In the same vein, we could replace
> > textclasses with CSS. But this would require way too much work.
> >
> > JMarc
>
> The comparison is fair, but as you notice our textclass already have a css
> style, so why to change? :)
>
> We still miss the information about the struture, here you must agree with
> me.
Can we pinch anything from Mozilla? or that against their Licence?
> Regarding how difficult it would be to construct the dtd parser, I have
> asked that to Cees de Groot, SGML maintainer, his answer follows:
>
> ----------------------------------- begin answer -----------------------------
[...]
> The long answer: use lex and yacc (or something else) to build a DTD parser.
> The parsed form would probably naturally fit in the format of a tree, and
> then it's a matter of bookkeeping (keep note about where you are in the tree,
> and make sure the user only selects tags from children of the current node).
>
> --------------------------------- end answer --------------------------------
We could perhaps build a library of preparsed dtd's then we wouldn't
actually need a dtd parser in LyX we instead link to an appropriate
library. The dtd's don't change often (not any more than article.cls
would -- almost never) so this scheme should be okay for long term use.
We'd have reasonable speed then also.
That just leaves the .layout files.
Allan (ARRae)
