xml-dev - RE: xslv updated

["Gregg Reynolds" <greynolds@greynolds.com>]

(Caveat: I recently took a new job, so I'm no longer at Datalogics and I'm
no longer on the XSL WG.)

> -----Original Message-----
> From: owner-xml-dev@xml.org [mailto:owner-xml-dev@xml.org]On Behalf Of
> Rick JELLIFFE
> Sent: Monday, June 19, 2000 5:34 AM
> Cc: xml-dev@xml.org
> Subject: Re: xslv updated
>
...
>
> built under XSLFO. From vague memory, it seems that the biggest design
> decision is whether to follow FOSI (an old DOD stylesheet language
> written using instance syntax which allowed expressions to query the
> current page position: this required tight coupling with the formatter
> so only ArborText really implemented it)

Correction:  Datalogics also implemented it (I used to work on their
engine).  So far as I know, those are the only two.  Naturally, they are
incompatible (although I've never used the Arbortext product so I can't say
for certain.)  FOSI is aging, at least in net time, but still drives the
typesetting of lots and lots of very large and complex SGML documentation,
and not only for the DOD, believe it or not.  The fate of FOSI makes for an
interesting case study in standards e/de-volution.  Customers come up with
new requirements not covered by the standards, so the implmentor customizes.

I wasn't there, but my understanding is that DSSSL was an attempt to improve
upon FOSI.

The real design decision regarding any style language, I think, is whether
to try to cover everything with a predefined abstractions (i.e. FOs or Flow
Objects or whatever you want to call them), or to design a minimal language
out of which such higher level abstractions can  be built - Knuth's
approach.  DSSSL, CSS, and XSLFO take the former approach.  I'm unaware of
any language that takes the latter approach while also maintaining a strict
separation content and style characteristic of SGML-centered style
languages.  But then my knowledge of the field is hardly encyclopedic; does
anybody know of such a language?

> the breaking routines/paginator; sacrificing the ability to
> test current
> page position forces more declarative nature {e.g."keep-togethers"},
> reduces expressiveness, sharply increases retargetability.)

I strongly disagree with the assertion that this reduces expressiveness.  On
the contrary, I think declarative languages tend to be more expressive than
procedural languages.

>
> (As for no quality typesetting being done with DSSSL (or near
> dialects),
> I am sure that Uniscope in Tokyo had good material to disprove that. I
> have seen it.)

What's their market share?  ;)

> vary for the material used too: for typesetting yellow page telephone
> books (i.e. with display ads), according to one contact it is
> difficult
> to get a good result without 4 page lookahead for determining which
> pages floats should go on.  I would not expect XSL-FO to
> provide someway
> of setting the lookahead for optimum page-floatation: that
> should be an
> implementations decision for vendors to diferentiate their products.

In general I don't think you can do high-end typesetting without the logical
equivalent of multiple passes, especially since you can't constraint the
organization of coordinated text flows in the source doc.  Maybe it's
possible for most jobs, but not for all.  E.g. footnotes can be anywhere in
the source, but they have to end up in the right place after composition.  I
don't think you can do that reliably without composing and recomposing the
whole thing.

-gregg


***************************************************************************
This is xml-dev, the mailing list for XML developers.
To unsubscribe, mailto:majordomo@xml.org&BODY=unsubscribe%20xml-dev
List archives are available at http://xml.org/archives/xml-dev/
***************************************************************************