DocBook to proprietary DTD: basic advice sought

["Christian Roth" <roth@infinity-loop.de>]

Hello,

up-front: I am one of the developers of upCast and downCast, RTF<->XML
conversion tools, and my questions are business-related. If this is not
acceptable for this list, please let me know.

Currently, I am investigating ways of converting DocBook to our
proprietary DTD ("upCast DTD"). This DTD is essentially a mix of markup
for the native structural possibilities and few layout features of RTF
(think: "Word"). The difference to WordML is that the upCast DTD tries
to be small, reprensting the logical document structure and non-verbose
(as far as possible), and factors styling out to CSS. Basic WordML seems
to be more of a re-formulation of RTF in XML syntax and holds formatting
information inline as markup.

I have collected the following points that influence the way a
conversion could be done:

1. The DocBook XSLs are a de-facto standard, so I am facing the
challenge that the output our conversion should create will be compared
to/measured against the former's FO output.

2. Existing customization layers should probably be usable with
reasonable adaptation effort.

3. Basing the conversion on the FO output, however, seems to lose many
details which need to be put to good use in the target DTD, like: the
'role' attribute of paragraphs, <indexterm> element information, <title>
status etc. These seem to no longer be present in the FO output (at
least in the basic configuration). Therefore, this route does not seem viable.

4. The FO holds the styling information inline, whereas ideally, it
should be factored out into a CSS stylesheet for our target DTD.

5. (X)HTML output seems to be much "nearer" to my target DTD, but still
some "back-conversions" must be performed like recreating a <footnote>
element from the <span>ned link and the contained reference to the
content. However, role (=class) attributes are preserved nicely.

6. Being able to roundtrip the DocBook source document through Word
(i.e., spooling out a DocBook document to Word for review/editing
purposes, then converting it back to DocBook without structural changes)
should happen whenever possible. (I am aware that there are situations
where this will not be feasible.)

My question is: Is there a (recommended?) way to leverage the basic
calculations and the full-fledged element support of the DocBook XSLs
for converting DocBook to one's own DTD? Is there some architectural
layer in one of the output format implementations I could hook into? Or
is trying to "plug in somewhere" probably going to be much more
difficult than simply starting from scratch and trying to recreate FO
output's layout in my own stylesheet? I guess that requirement 6 above
is actually what will make the latter almost necessary.

So I am interested in any experience from those of you who have
targetted their own DTD from DocBook and those who are much more
intimate with the internal structure of the DocBook XSLs than I am to
hear whether plugging into one of it is possible or worthwhile.


Kind regards
Christian Roth

Link to basic documentation on the target DTD (="upCast DTD"):
<http://www.infinity-loop.de/support/documentation/dtd/>

Actual DTD (concise namespaced version):
<http://www.infinity-loop.de/DTD/upcast/5.0/upcastNS.dtd>

--
Christian Roth (CTO)
Phone: +49 89 890432-95
Web  : http://www.infinity-loop.de
infinity-loop GmbH * Neideckstr. 25 * 81249 München * Germany