Ideas for generic document processing tools

Ian and I have been thinkng about how to make tools that we can use for
generating specs.

In making the Authoring Tool Accessibility Guidelines we do a bunch of little
tricks with different scripts which have been heavily adapted from the
scripts used to generate the CSS2 spec, and we were looking at changing some
of them. I thought I'd throw this open, and therefore exlpain a bit about how
we do what we do, and what that is.

The scripts that we now use can be found at
/afs/w3.org/pub/WWW/WAI/AU/Group/wai-authtools-src/

Some background:

There is a single document that forms the source of the Guidelines, the
Techniques document, and is used to generate some of the conformance
evaluations, the checklist of checkpoints. The conformance evaluations and
checklists have blank templates into which we add the content - the
techniques and guidelines documents are generated completely from the source.

We do some useful tricks: 

references are in a DL at the end of the source documet, and anything in a
body that looks like [[something]] is replaced by an appropriate reference
(at the moment there are no inline direct links, only to to the reference
section, which is something I would likeot change). The references that are
used in a given document are kept there and the rest are dropped.

Definitions are handled similarly - anything that links to something called
def-anything means that the definition item whose DT is an A with the name
def-anything gets kept, and a link rel="glossary" is inserted. It would make
more sense to me if we started by picking things that were linked
rel="glossary" - then I could style it.

There are sections for the techniques or guidelines only. These have a class
applied, and I use a user stylesheet and Amaya to figure out which is what.
Similarly there are classes for checkpoints of various conformance
evaluations, since the easiest way to do one is against the guidelines. I
would like to be able to separate this out and do it from a checklist - I
suspect that isn't actually too difficult.

Now we are moving from using sed-type substitutions to remove things to using
IFDEF sections in comments, since then we can deal with things that are
nested more easily. In fact I would like to use an XSLT or some DOM
programming to do this, since having the ifdef means our source document
becomes invalid (eg it has two titles), although we may be able to solve that
one, and more seriously that I can't use classes or style to see different
pieces, and have to go back to editing the source (at which point it's easier
to do it in something other than HTML)

Chaals


--
Charles McCathieNevile    mailto:charles@w3.org    phone: +61 409 134 136
W3C Web Accessibility Initiative                    http://www.w3.org/WAI
21 Mitchell Street, Footscray, VIC 3011,  Australia (I've moved!)

Received on Sunday, 2 January 2000 10:00:34 UTC