Re: DOCBOOK: Element for literal "passthrough"?

["M. Wroth" <mark at astrid dot upland dot ca dot us>]

At 09:39 AM 4/3/01 -0400, Norman Walsh wrote:
> / "M. Wroth"
> <mark@astrid.upland.ca.us> was heard to say:
> | The project is, BTW, a literate programming extension to DocBook,
> and
> | this is part of getting the "tangle" (code output) branch to
> work.  So
>
> Interesting. Have you considered using XSLT instead of DSSSL? I
> believe
> that you'd get the "right" answer if your "tangle"
> transformation used
> the "text" output method.

I considered it, and chose to go with DSSSL for several reasons:

• The "weave" branch needs to go to printed output, and I haven't figured out a good way to do that
• I want the solution to work with both SGML and XML based documents, rather than just XML
• I know DSSSL (more or less) and would have to learn XSLT (this may be the biggest reason :-)

An XSLT solution may be in the future; one of the attractive things
about doing this project in SGML/DSSSL is that it should be
straightforward to process the resulting documents using various tools,
rather than a major code effort.

> In DSSSL, I'd be tempted to examine
> each character in the output and
> do the right thing without the <literalchar> hack. It'd be
> slower,
> probably, but so much cleaner :-)

That would require more DSSSL expertise than I have, I think.  But
the main objection I have is that the code text can contain markup --
xref at a minimum, but potentially all of the markup currently in
programlisting plus some additional markup aimed at
"prettyprinting" the code, and entity references (taking the
place of macros in some senses).  So doing the "right
thing" is pretty close to parsing general SGML.  It's easier to
define a few entities and use them in appropriate places in the code, I
think (although if I wind up using this for SGML code, I may change my
mind :-).

Mark B. Wroth
<mark@astrid.upland.ca.us>