[
Lists Home |
Date Index |
Thread Index
]
J.Pietschmann wrote:
> What's the advantage of using a three sttage approach
> source XML - >(XSLT) -> XSLFO+CSS -> (CSS processor) -> XSLFO
> if you can transform the source XML into the target XSLFO
> directly? What's the advantage of expressing the style properties
> in a language with Yet Another Syntax which can provide difficulties
> when it has to be processed in the XSL transformation? (except of
> showing that one still knows the CSS syntax, of course)
The number of steps is of no importance as long as they are performed
automatically and with reasonable performance. That's just an
implementation detail.
The advantage is that you specify the properties in a compact
notation that has been around for a few years and used in other
similar situations like HTML and SVG. Having (X)HTML + CSS
and SVG + CSS, then just XSL-FO looks like Yet Another Way Of
Doing It to me.
This isn't really about which syntax is universally better. It depends a
lot on your environment, background, personal preferences etc.
I think it's about consistency. David Carlisle tried to justify the
inconsistency by arguing that XSL-FO is a completely different
animal. I just don't buy that.
Now, ideally, I think that any modern specification language should
either be defined with an XML syntax or have a normative mapping
to an XML syntax. The primary reason would be to facilitate processing
by XML tools. For some languages like XSLT, the XML syntax
is also a natural syntax to actually author in. For others like RELAX NG
it's probably the non-XML syntax that will be the first choice.
I read in a post you made a couple of messages down this thread,
that you suggested that the semantics of XSL-FO should be defined in
a syntax independent way. Essentially I agree with that and
theoretically it looks like an good idea. But in practice, I think that a
specification becomes easier to understand if you define it in terms
of a concrete syntax (ideally XML), and then provide normative
mappings to other syntaxes if feasible. I think RELAX NG is an
excellent example of this.
Cheers,
David
---------------------------------------------
David Rosenborg
Pantor Engineering AB
|
