Lists Home |
Date Index |
From: Lars Marius Garshol <email@example.com>
Date: 31 Jan 2002 10:13:33 +0100
> I would rather say horizontal vocabularies, I think. One place where
> XML really does need something like AFs is in XLink. People don't like
> having to give their linking elements particular names, and yet they
> would like to have the benefit of standardized linking semantics. AFs
> provide a way to do this (this is how HyTime works) that is simple and
> easy to use.
The short answer to this is "yes." The long answer, or at least part of the
explanation for why more effort wasn't put into making this happen, involves
a large chunk of SGML/XML history. Here's a summary:
In the fall of 1995, I gave presentation on document management at Fujitsu,
Ltd., including a discussion of the ISO SGML, HyTime, and DSSSL standards.
The upshot of this presentation was the development of the HyBrick browser
by Fujitsu Laboratories.
A little over a year later, FJ Labs sent me the first version, just in time
for the 1996 ISO meeting (which immediately preceded the SGML 1996
conference, where work on XML was announced publicly for the first time).
This meeting opened with Charles Goldfarb commenting on the need to address
DSSSL/HyTime integration. The DSSSL and HyTime work groups, warring factions
for years, had arrived at a synthesis based around property sets for SGML
and HyTime. The problem, as I soon discovered, was that this "synthesis" was
at the lowest level only. The larger, architectural, question of how DSSSL
and HyTime were to relate to each other had not been addressed.
I had been introduced to the subject of linking from the HyTime direction.
Consequently, when I made HyBrick available for distribution, I was
surprised to find that those with commercial requirements requested that
linking behavior be controlled through DSSSL stylesheets and not through
HyTime. HyTime includes AFs for various types of links. DSSSL, intended for
print-oriented stylesheets, contains no such behavior-control definitions.
What I discovered was a chasm between those who saw linking control in
HyTime-terms and those who saw it as part of the formatting "spectrum."
Later, this "discussion" would become downright acrimonious.
There was also another, even bigger surprise. The TechnoTeacher folks, whose
HyMinder engine Fujitsu had licensed and with whom I had written the "HyTime
Application Development Guide", weren't interested in demonstrating their
GroveMinder server-side product in concert with HyBrick as a front-end
Possibly the strangest part of their reaction was that the then existing
draft of XLink had been implemented in HyBrick in terms of the HyTime
property set. Look on the Working Group members and Acknowledgements list
for XLink and you'll find the name Masatomo Goto, Fujitsu Laboratories.
Goto-san had visited TechnoTeacher and worked out the expression of XLink in
HyTime terms with them. Nonetheless, the TechnoTeacher folks weren't
interested. According to them, "HyBrick didn't add anything."
How did this affect the development of XLink? I preceded Goto-san as
Fujitsu's representative to the XLink Working Group. Faced with the
indifference of TechnoTeacher, the prime developers of HyTime, there wasn't
much point in fighting a battle to see XLink expressed in terms of HyTime
AFs. My last message to that group was, however, a suggestion that XLink be
expressed in inheritable terms. That would have made it possible to define,
for example, links that were bi-directional, as defined in HyTime and
supported by HyBrick, but acted like an HTML A tag if that's all the
receiving browser would support.
Another historical footnote: HyBrick was used to give the first
demonstration of XML "as it was meant to be" - or as the Abstract to the W3C
Recommendation says: "served, received, and processed on the Web in the way
that is now possible with HTML." The demonstration was given on XML
Developer's Day in the Spring of 1997, Jon Bosak presiding. The demo was
arranged by Jon himself, after I put him in touch with the HyBrick
developers at Fujitsu Labs. in Japan and the developers of a Web-based
publishing system at Fujitsu, ICL in Ireland. HyBrick was used to download
and render the files from the site in Ireland. Unfortunately, the demo was
poorly put together, used a choice of material that was unrelated to the
point being made, and consequently had no influence.
My work on HyBrick was terminated at the end of July, 1999. The work at
Fujitsu Labs, Japan ended sometime later.