Re: Using an XSL Formatter as an XSL-FO Web Browser

(This message was mistakenly sent as a private mail instead of the list - I was
fooled by the "Reply-to" thing. Sorry Arved for receiving it twice.)

> Taken at face value I think it makes some good points. I like the suggestion
> that general XSL "things" not be called "stylesheets", because they aren't;
> so I've been pushing that idea rather strongly myself. I like the idea that
> FO documents should contain a link back to their source...also, even if
> processing XSLT+XSL-FO together cannot be rigidly enforced it can still be
> strongly encouraged. In general I agree that delivering FO to Web clients
> (or really, any disconnected client, say FO via SMTP or JMS) isn't a great
> idea.

IMO, delivering FO is as good as delivering PDF  - or XHTML with those great
rich CSS3 styles :-). It should be avoided but still may be useful under certain
circumstances.

Speaking in general, I don't like anathematizing XSL-FO as the worst enemy of
the Semantic Web just because they can express fine-tuned layout in a
self-contained document. Following this logic, CSS2 is equally harmful because
it lets you style a document consisting entirely of  <div>s and <span>s :-).
(I feel I am repeating things abundantly said in 1999, so I stop :-)).

> FWIW this is not an attack on XSL WG folks...I'm
> guessing you guys maybe are forced to do some things that maybe don't feel
> quite right just because you're part of W3C (no need to answer :-))

I share your feeling. It looks like W3C has a firm intent to promote CSS: I
cannot otherwise explain why shorthands and messy inheritance penetrated into
XSL. My impression is that the WG has a "political" constraint to be loyal to
CSS; and this prevents XSL from becoming better than CSS (alas!).

In the current version of XEP, XSL FO tree normalization (shorthand expansion,
inheritance calculations, etc) takes about 60% of the processing time, and
requires more memory than the rest of formatting. When I think that the
formatter could be twice as efficient if there were no CSS2 legacy, I'm getting
sad :-(.

Regards,

Nikolai Grigoriev
RenderX

P.S. Happy XSLT processor writers - they don't have to parse CSS selectors
(I wonder how they escaped :-)).

Received on Sunday, 11 February 2001 10:25:47 UTC