Re: DocBook -> Texinfo (via DSSSL) Proof-of-Concept Demo

Ulrich Drepper <> writes:

> Jim Pick <> writes:
> > I didn't try to do any stripping out of whitespace.  But couldn't that
> > be done just using a function written in DSSSL?  (heh, it's Scheme,
> > mostly)
> This completely ignores the purpose of DSSSL :-)
> If you look at the DSSSL specs you'll find in the later half (chapter
> 11?) the desription of the "output language".  The DSSSL itself should
> more or less be output format independent.  I.e., one should be able
> to use Norm's DSSSLs.  The work to be done is to get a Jade backend
> which maps the lowlevel stuff of the backend description to reasonable
> texinfo.  Forcing a string out really is only the last escape.

You mean the "DSSSL Transformation Language".

Unfortunately, Jade does not support the DSSSL Transformation language.

Norm's stylesheets don't use it either for HTML (well, he can't).

So, it's not really a "pure" approach.  But I don't think a "pure"
approach really makes all that much sense in this case anyways.  The
stylesheets just use one less level of abstraction than if the
transformation language existed.  That's not so bad, because I don't
really see what having the addition level of abstraction would buy
> Since you've started writing it this way and I cannot provide anything
> similar I don't suggest to not use your code.  Rather I'd propose to
> not spend to much time on this one approach since it is a dead end.

I don't agree that it is a dead end (not yet anyway).  DSSSL seems
suited to the task.  A robust DocBook -> Texinfo stylesheet probably
doesn't have to be very long.  I didn't test my stylesheet on anything
longer than the demo document I used, but it will probably work on
longer documents.  (the demo doesn't handle many entity types, so it
would very likely have ugly output)

To me, it just seems a lot cleaner and more generalized to write a
single stylesheet (maybe split across several files, as Norm has done)
than to try to stitch together a bunch of different tools (ie. nsgmls
parser -> perl -> output) and writing "real code" to do the same job.
Plus we get to share code with the output methods for HTML and


 - Jim

[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]