Google Groups no longer supports new Usenet posts or subscriptions. Historical content remains viewable.
Dismiss

XSLT for webpages

13 views
Skip to first unread message

Jeff Rubard

unread,
Feb 13, 2004, 6:41:11 PM2/13/04
to
I am curious to know whether anyone has experience using XSLT for
web XML (non-XHTML) styling, either with CSS or standalone. I
myself have engaged in rather unsuccessful experiments with the
DocBook chunk stylesheet, but it seems to me that retaining the
structural markup of XML formats (i.e., the marking out of salient
properties of a text in "anticipation" of RDF "semantics") demands
something more powerful than CSS as it is known to me; and would
be curious to hear opinions on either the viability of XSLT sans
Formatting Objects, or structure-preserving CSS tricks.

Jeff Rubard

Andy Dingley

unread,
Feb 13, 2004, 8:54:33 PM2/13/04
to
On Fri, 13 Feb 2004 23:41:11 GMT, Jeff Rubard
<jeffr...@fusemail.com> wrote:

>I am curious to know whether anyone has experience using XSLT for
>web XML (non-XHTML) styling,

This has been commonplace since around '99 or so.

>it seems to me that retaining the
>structural markup of XML formats (i.e., the marking out of salient
>properties of a text in "anticipation" of RDF "semantics") demands
>something more powerful than CSS as it is known to me;

Yes, CSS doesn't cut it.

XSLT is a powerful tool for most transforms you can require and can
express in a "structural" manner - i.e. you can describe what's needed
in terms of XML-Infoset, both for input, output and control. An
experienced XSLT developer should be able to build these, to such a
level that your biggest problem becomes managing the XSLT code (which
is a nightmare) rather than implementing the transform.

XSLT is good at "non-semantic" transforms, and it fails at "semantic"
processing. Trying to process an RDF/XML document is impractical, just
because it may have a model structure that's more than XML-Infoset can
express (usually rdf:resource is the culprit), even before you start
to worry about making complex decisions based on semantic content.

XSLT is a good final styling step for web content, expanding a DocBook
core document into a "site" with nav menus on the page and links to
related pages. But it's not a content assembler to select DocBook
fragments that match some semantic query ("animal" and "has wings" and
"not bird"), nor is it a search engine that can efficiently query a
huge phone directory and return a single record.

--
Smert' spamionam

Jeff Rubard

unread,
Feb 13, 2004, 10:37:17 PM2/13/04
to
On Sat, 14 Feb 2004 01:54:33 +0000, Andy Dingley wrote:

> On Fri, 13 Feb 2004 23:41:11 GMT, Jeff Rubard
> <jeffr...@fusemail.com> wrote:
>
>>I am curious to know whether anyone has experience using XSLT for
>>web XML (non-XHTML) styling,
>
> This has been commonplace since around '99 or so.

Two questions: 1) What formats are commonly styled for web
presentation using XSLT? 2) I wasn't aware that browser support for XSL
was incorporated until quite recently. Are you referring to something
else?

>>it seems to me that retaining the
>>structural markup of XML formats (i.e., the marking out of salient
>>properties of a text in "anticipation" of RDF "semantics") demands
>>something more powerful than CSS as it is known to me;
>
> Yes, CSS doesn't cut it.

Well, as far as I can tell XHTML is designed to have no more structural
markup than CSS can style, which means that other formats require
something more powerful basically *by design* (i.e., that CSS should be
"deprecated" in the minds of those who think that formats with more
extensive tags should be widely used, as I do). But XSL-FO doesn't seem
to have been designed with any kind of presentation other than PDF
(*perhaps* good enough for print, although it wouldn't have been in the
"camera-ready" era). I dunno, maybe .dvi can be brought back as a
rendering standard since PostScript appears to be off limits

> XSLT is a powerful tool for most transforms you can require and can
> express in a "structural" manner - i.e. you can describe what's needed
> in terms of XML-Infoset, both for input, output and control. An
> experienced XSLT developer should be able to build these, to such a
> level that your biggest problem becomes managing the XSLT code (which
> is a nightmare) rather than implementing the transform.
>
> XSLT is good at "non-semantic" transforms, and it fails at "semantic"
> processing. Trying to process an RDF/XML document is impractical, just
> because it may have a model structure that's more than XML-Infoset can
> express (usually rdf:resource is the culprit), even before you start
> to worry about making complex decisions based on semantic content.

Well, the thing is the "semantic model" for mark-up documents is,
*logically* speaking, a fantasy: formal systems extensive enough to
encode all the material we'd want simply aren't tractable enough to
allow machine processing on the CERN XML model; i.e., they really came a
cropper with RDF.

> XSLT is a good final styling step for web content, expanding a DocBook
> core document into a "site" with nav menus on the page and links to
> related pages. But it's not a content assembler to select DocBook
> fragments that match some semantic query ("animal" and "has wings" and
> "not bird"), nor is it a search engine that can efficiently query a
> huge phone directory and return a single record.

This is what I'd like to know more about: as I've said, I've looked at
Walsh's stylesheets and guide for DocBook XSL but would be curious to
know more about an implemented TEI standard or something along those lines.

Andy Dingley

unread,
Feb 14, 2004, 8:18:40 PM2/14/04
to
On Sat, 14 Feb 2004 03:37:17 GMT, Jeff Rubard
<jeffr...@fusemail.com> wrote:

>Two questions: 1) What formats are commonly styled for web
>presentation using XSLT?

What does "styled for web presentation using XSLT" mean ?

If I use XSLT on the server, I can use it to serve content to
everything from Telex machines and NS4 upwards.

Do you mean "XML to the client, styled with client-side XSLT" alone ?

>) I wasn't aware that browser support for XSL
>was incorporated until quite recently. Are you referring to something
>else?

Yes - although at least one browser offered XSL in '99, it was a
rarity. It's still impractical to offer XSLT-alone content over the
web - most users are using IE 6, but there are plenty who aren't.
Only if you're some form of intranet or extranet where you can
influence browser choice can you really do this.


>>>it seems to me that retaining the
>>>structural markup of XML formats (i.e., the marking out of salient
>>>properties of a text in "anticipation" of RDF "semantics") demands
>>>something more powerful than CSS as it is known to me;

What does CSS have to do with anything ? You can express (some)
semantics with HTML 3.2 and liberal use of class and id attributes,
although this isn't enough to really allow more than simple end-to-end
communication between systems that are already hand-built to
understand each other.

CSS can be used to apply styling to a classified HTML markup, and it's
not hard to arrange the semantic description so it's also available to
drive the CSS. However the CSS selectors have no comprehension of
semantics, they're just using the markup coincidentally.

The next generation is as much about communicable standards for
semantics, as they are about semantics themselves. I've been using RDF
since '99 for the internals of systems, but it's only recently that
DAML or OWL have made these semantics communicable to other systems
that haven't this tight and hand-coded link set up beforehand.
Incidentally, talking of RDF, then have a look at this week's new
document release.

>Well, as far as I can tell XHTML is designed to have no more structural
>markup than CSS can style,

What does this mean ? XHTML certainly _wasn't_ designed with sole
reference to CSS (I did once have this very discussion with Dave
Raggett - should we ditch the lot in favour of <div> and <span> and a
bunch of CSS). XHTML exists in isolation. If it has simplistic
semantics, then that's for its own reasons - not because of CSS.


>which means that other formats require
>something more powerful basically *by design* (i.e., that CSS should be
>"deprecated" in the minds of those who think that formats with more
>extensive tags should be widely used, as I do).

What's your obsession with CSS ? I'm beginning to wonder if you
have even a glimmer of a clue here. CSS lives at a presentation layer
- if you care about semantics, you need to fix this at a deeper level.
if you're to have any hope.


>But XSL-FO doesn't seem
>to have been designed with any kind of presentation other than PDF

This is untrue. If PDF is predominant, that's just because PDF is a
popular target. It's not an XSL:FO limit.

>Well, the thing is the "semantic model" for mark-up documents is,
>*logically* speaking, a fantasy:

Why ?

>formal systems extensive enough to
>encode all the material we'd want simply aren't tractable enough

That's more because demand expands to fill available technology. What
are you specifically thinking of here, are are you just re-arranging
words you've overheard ?

>allow machine processing on the CERN XML model; i.e., they really came a
>cropper with RDF.

Why ?

And what does CERN have to do with things today ?


>This is what I'd like to know more about: as I've said, I've looked at
>Walsh's stylesheets and guide for DocBook XSL but would be curious to
>know more about an implemented TEI standard or something along those lines.

Look at the huge amount of work going on right now with RDF and OWL.
Go to the Protege workshop in a few months. But no, you;ve already
decided that RDF "came a cropper"

--
Socialism: Eric, not Tony

Jeff Rubard

unread,
Feb 16, 2004, 4:28:18 PM2/16/04
to
Andy Dingley <din...@codesmiths.com> wrote in message news:<2rdt20t3dvuq0ig8i...@4ax.com>...

> On Sat, 14 Feb 2004 03:37:17 GMT, Jeff Rubard
> <jeffr...@fusemail.com> wrote:
>
> >Two questions: 1) What formats are commonly styled for web
> >presentation using XSLT?
>
> What does "styled for web presentation using XSLT" mean ?

An XML file prepared for presentation through a web browser
(specifically, cutting out the stage of HTML conversion).

> If I use XSLT on the server, I can use it to serve content to
> everything from Telex machines and NS4 upwards.

But in what formats can the content be served? If the answer is
any format, as you suggest, then the contemporary content model
is misprised -- really, we seem to be dealing with an immanent
FO level within XSLT (i.e., the contemporary standards for Formatting
Objects are something like "cripple-ware" relative to rendering
technologies of the none-too-recent past -- why not Ghostscript FO?)



> Do you mean "XML to the client, styled with client-side XSLT" alone ?

More or less (see above).

> >) I wasn't aware that browser support for XSL
> >was incorporated until quite recently. Are you referring to something
> >else?
>
> Yes - although at least one browser offered XSL in '99, it was a
> rarity. It's still impractical to offer XSLT-alone content over the
> web - most users are using IE 6, but there are plenty who aren't.
> Only if you're some form of intranet or extranet where you can
> influence browser choice can you really do this.
>

Well, obviously anyone can afford Mozilla and it has an XSLT engine.
I.e., the thought is obviously a little blue-sky but presentation is
going in certain directions and there are not necessarily overall
reasons determining this.

> >>>it seems to me that retaining the
> >>>structural markup of XML formats (i.e., the marking out of salient
> >>>properties of a text in "anticipation" of RDF "semantics") demands
> >>>something more powerful than CSS as it is known to me;
>
> What does CSS have to do with anything ? You can express (some)
> semantics with HTML 3.2 and liberal use of class and id attributes,
> although this isn't enough to really allow more than simple end-to-end
> communication between systems that are already hand-built to
> understand each other.

Well, on my understanding <p> is a semantic marker: semantic structure
is
non-presentation structure.

> CSS can be used to apply styling to a classified HTML markup, and it's
> not hard to arrange the semantic description so it's also available to
> drive the CSS. However the CSS selectors have no comprehension of
> semantics, they're just using the markup coincidentally.

Well, CSS has to do with "latent" content as understood in traditional
layout. Have you ever had to do paste-up? You get the idea that what
you are sending "camera-ready" to the printers is not really that
arbitrary relative to established typographical standards.

>
> The next generation is as much about communicable standards for
> semantics, as they are about semantics themselves. I've been using RDF
> since '99 for the internals of systems, but it's only recently that
> DAML or OWL have made these semantics communicable to other systems
> that haven't this tight and hand-coded link set up beforehand.
> Incidentally, talking of RDF, then have a look at this week's new
> document release.
>

Well, what is so fabulous about having a set-theoretic language (OWL)
handling textual content? The traditional understanding (deriving
primarily from Tarski) is, really nothing: if you were actually to do
semantically-based reasoning using a "calculus of inclusion" you'd run
into paradoxes, and a different kind of structuration is required to
do less stringent "reasoning" such as automated theorem-provers are
capable of (i.e., working recursions upon sets).



> >Well, as far as I can tell XHTML is designed to have no more structural
> >markup than CSS can style,
>
> What does this mean ? XHTML certainly _wasn't_ designed with sole
> reference to CSS (I did once have this very discussion with Dave
> Raggett - should we ditch the lot in favour of <div> and <span> and a
> bunch of CSS). XHTML exists in isolation. If it has simplistic
> semantics, then that's for its own reasons - not because of CSS.
>

Well, you seem to be really locking with the W3 teleology here: XHTML
is a objectively defined standard with certain structural properties,
and the assertion was that one of these properties is tight
integration with CSS (i.e., there would be no need to do XHTML styling
using a more powerful styling language).



> >which means that other formats require
> >something more powerful basically *by design* (i.e., that CSS should be
> >"deprecated" in the minds of those who think that formats with more
> >extensive tags should be widely used, as I do).
>
> What's your obsession with CSS ? I'm beginning to wonder if you
> have even a glimmer of a clue here. CSS lives at a presentation layer
> - if you care about semantics, you need to fix this at a deeper level.
> if you're to have any hope.
>

I'm beginning to wonder whether discussion of markup languages is not
allowed to be pitched at this level of abstraction anymore (i.e.,
whether I actually even have a thought to think about SGML and
Mostowski relative to the parameters specifying W3 consortium
standards). It appears that one is to instead have a "depth" model of
semantic markup, which I am saying doesn't make a lot of good logical
sense (i.e., what-all you said isn't just what you meant).

> >But XSL-FO doesn't seem
> >to have been designed with any kind of presentation other than PDF
>
> This is untrue. If PDF is predominant, that's just because PDF is a
> popular target. It's not an XSL:FO limit.

Well, and it is too (i.e., it's designed for extremely static
presentation, *not even* ).

> >Well, the thing is the "semantic model" for mark-up documents is,
> >*logically* speaking, a fantasy:
>
> Why ?

Well, I'd need to know more about what you

> >formal systems extensive enough to
> >encode all the material we'd want simply aren't tractable enough
>
> That's more because demand expands to fill available technology. What
> are you specifically thinking of here, are are you just re-arranging
> words you've overheard ?

I've never heard anybody say anything at all like this, actually.

> >allow machine processing on the CERN XML model; i.e., they really came a
> >cropper with RDF.
>
> Why ?

Because semantic structure isn't really just "object-property"; it's
got all kinds of sides, and the minute you go to a relational level
with semantic content you have serious computational problems (it's
the *Entscheidungsproblem*, the unsolvability of which really the
fundament of computer science, in a different form). *That* is kind
of a commonplace.

> And what does CERN have to do with things today ?
>

As little as possible (as far as I can tell).

> >This is what I'd like to know more about: as I've said, I've looked at
> >Walsh's stylesheets and guide for DocBook XSL but would be curious to
> >know more about an implemented TEI standard or something along those lines.
>
> Look at the huge amount of work going on right now with RDF and OWL.
> Go to the Protege workshop in a few months. But no, you;ve already
> decided that RDF "came a cropper"

Well, I'm not really in that business. But thank you for your
observations; I'll go familiarize myself with that RDF information.

P.S. By the way, who is Eric (not Tony)?

Jeff Rubard

unread,
Feb 20, 2004, 5:14:00 PM2/20/04
to
opense...@graffiti.net (Jeff Rubard) wrote in message news:<49b5d62a.04021...@posting.google.com>...

> Andy Dingley <din...@codesmiths.com> wrote in message news:<2rdt20t3dvuq0ig8i...@4ax.com>...

> P.S. By the way, who is Eric (not Tony)?

Seriously, who is this guy? Is this the British Oskar Lafontaine?

Andy Dingley

unread,
Feb 23, 2004, 5:24:01 PM2/23/04
to
On 20 Feb 2004 14:14:00 -0800, opense...@graffiti.net (Jeff Rubard)
wrote:

>> P.S. By the way, who is Eric (not Tony)?
>
>Seriously, who is this guy? Is this the British Oskar Lafontaine?

Google for Eric+Orwell+pseudonym

Jeff Rubard

unread,
Feb 24, 2004, 5:34:11 PM2/24/04
to

Okay.

Simplicio: Questo non può essere, perché le generatzioni, mutatzioni
etc. che si facesser, verbigrazia, nella Luna, sarebber inutile e vane,
<<et natura nihil frustra facit>>.

I look not for a substantive faculty, but for the generations, mutations
and the like: what it lights upon, joyful noise, a moon of cilia, would
useless and accessible <<and nature causes nothing in vain>>.

Galileo, *Dialogues*

Jeff Rubard

unread,
Feb 27, 2004, 9:16:01 PM2/27/04
to

I guess that's not how it works at all, though.

Jeff Rubard

unread,
Mar 1, 2004, 6:58:59 PM3/1/04
to

That is, as long as you don't remember the Marne and Ferdinand Foch,
as per the joke.

Jeffrey Rubard

unread,
Jan 31, 2022, 8:27:04 PM1/31/22
to
2022 Update: Yeah, this wasn't the most.

Jeffrey Rubard

unread,
Feb 2, 2022, 3:08:55 AM2/2/22
to
Not everybody can work in IT, either.
0 new messages