Re "Move to Divide the Question"

[keshlam@xxxxxxxxxx]

1) For the CSS versus XSL debate, see W3C's webpage for positioning of the
two tools. Vendors I've heard have been saying "We like some parts of both
of these, and will use the one that fits a particular problem while waiting
for both of them to evolve and -- possibly, eventually, maybe -- converge."

My own best understanding is that CSS is good for back-annotating an
existing document without changing its basic structure, and has the
advantage of clearly defined "cascading" behavior, while XSL is much better
able to handle changes that _do_ involve substantially rearranging and
restructuring the information. Which is best thus depends on what you're
trying to do to your document. The overlap may increase over time, but for
now either use the one that suits your particular task, or provide feedback
to the designers about how it could better do so.

2) Re the need for formatting objects: This is the abstract markup debate
in yet another guise. HTML really is both overly specific and not specific
enough as a markup language for directing rendering, as far as I'm
concerned... in large part because it started out as a minimal abstract
markup and then had lots of concrete detail dumped into it on an ad-hoc
basis. I would Really Like to see a richer language for expressing document
formatting at the conceptual level -- an abstraction of formatting, which
the data-level abstractions could be transformed into on their way to print
-- and that's the niche that FO's are supposed to fit. Of course they then
have to be further elaborated/transformed to produce the final rendering;
they're not intended to be the final form, but directives to a processor
which will in turn produce the final form. This is a useful reference back
toward our GML "document-compilation" roots.

(Yes, you could transform directly into directives describing precisely how
to lay out the page. But an intermediate layer -- one which states
formatting intent but can be rearranged by the formatter to fit the media
available -- strikes me as a good thing. Of course, if you don't need it
for your problem, you'd simply ignore it.)

3) Re dividing the question into a transformation language and a formatting
language: I'm _almost_ in favor. My main hesitation is that I think FOs
could be one good "use case" for making sure that the transforms really are
rich enough. There's something to be said for having a testcase actually
built into the design process.

______________________________________
Joe Kesselman  / IBM Research
Unless stated otherwise, all opinions are solely those of the author.



 XSL-List info and archive:  http://www.mulberrytech.com/xsl/xsl-list