Markup Technologies '98

By Bob DuCharme

The two-day Markup Technologies '98 conference was held in Chicago directly after the three-day XML '98 conference in the third week of November. By splitting what had formerly been one big annual conference, the Graphic Communications Association hoped to make it easier for attendees to focus on the issues and level of discussion most appropriate for them. I think they succeeded. While the first conference offered a great deal to XML neophytes, the second skipped the simplistic introductory topics and focused on system design and implementation issues of interest to more advanced XML people (in other words, programmers and SGML people).

Amid the ongoing debate of XMLvs. SGML, the conference's title of Markup Technologies '98 showed that it wasn't necessarily about SGML or XML. The opening keynote speech by former Stanford professor Brian Reid, inventor of Scribe, made this clear. His talk, titled 20 Years of Abstract Markup: Any Progress? discussed issues of planning, implementing, and using abstract markup that applied to all past, present, and future markup languages. (Scribe was one of the first--if not the first--widely-used markup languages to separate content from presentation in order to make documents more portable; time has shown the importance of its innovations such as nested elements, high-level nonprocedural systems, device-independent documents, and tree structure of "entities.") To show the relevance of the same issues he faced twenty years ago, most of Reid's slides were from a talk he gave in Lausanne, Switzerland in 1981.

After Reid's talk put the present in the context of the past, the keynote by Michael Sperberg-McQueen (coeditor of the XML specification and an important figure in SGML's development) discussed issues of meaning and markup language design for the future. In practice, this doesn't mean some future language that offers a third choice beyond SGML and XML, but instead the planned applications of XML that will provide richer data modeling capabilities. (For example, better data typing--in Sperberg-McQueen's words, giving us a way to make sure that "something tagged as a date really is a date and something tagged as a monetary amount really is a monetary amount"--an important need for the growing number of nonpublishing XML applications.) Sperberg-McQueen made it clear that he wasn't speaking in his capacity as chairman of the W3CXML Schema Working Group, but it was good to hear about the larger issues that they're keeping in mind as various groups petition to have their immediate short-term wish lists incorporated into the XML Schema Working Group's agenda.

Outside of the keynote speeches, the conference offered two separate tracks of presentations: a theory track and a practice track. The latter offered more case studies of existing systems. For example, Jorge Leal Portela of Belgium's SGML Technologies Group described their GRASP system, which tracks stolen and found art objects across the European Union. Its combination of SGML, XML, artificial intelligence, and distributed objects technology is a fascinating combination of technologies one rarely sees in an active, production system. Other case studies presented included a DTD maintenance system and rapid reporting tools developed at the legal publisher Matthew Bender and experimental efforts to share medical records among multiple institutions by Oceania, a Palo Alto firm that develops systems for accessing and analyzing patient information.

The theory track's presentations and papers were more general, and probably more challenging to those with no background in markup languages or computer science. Timothy Arnold-Moore's Models for Structured Document Database Systems and Liam Quin's Links in XML: Detection, Representation and Presentation both provided detailed surveys of work in these fields that will provide reference points for much future work. Other work presented in the theory track included case studies, but perhaps because of their academic origins (for example, Automated DTD Subsetting Using Architectural Processing by the Summer Institute of Linguistics' Gary Simons and In Defense of Invalid SGML by the University of Pittsburgh's David Birnbaum) the presentation of the lessons learned made them a better fit for the theory track than the practice track.

Arnold-Moore's and Quin's papers were not the only ones that will prove valuable to future work. The Markup Technologies Conference's Proceedings , while slimmer than the big fat books provided with the longer conferences of past years, had a lot more genuinely useful material. (It's the first SGML/ XML conference proceedings I've seen with no printouts of Powerpoint slides masquerading as papers.) You'll be seeing some of these papers cited by future papers in conference proceedings and the brand new Markup Languages journal recently added to MIT's stable of academic journals, which also received much attention at the conference.

An interesting theme of the conference both on and off the podiums was the role of abstraction in systems using SGML and XML. Many presenters, when describing the advantages and disadvantages of adding layers of abstraction to a system, included "more confusing to novices" as a disadvantage of using abstract data models in their development. (In his closing keynote, Tim Bray quoted Microsoft general manager Adam Bosworth as saying "Every time we add a layer of abstraction we lose 90% of our audience.") For example, the use of groves ( Graph Representation of Property Values --see http://www.oasis-open.org/cover/topics.html#groves for more information) outside of the HyTime and DSSSL arenas where they originated were considered too much complication by some and a useful tool for planning and implementing a system by others. Paul Prescod's hastily convened tutorial on groves convinced me of their value; I hope members of the relevant W3C committees keep them in mind.

Other "high level" issues that came up often were many hardcore computer science concepts that SGML people formerly considered to be too, well, "computer sciencey" to be valuable in the publishing world. The group of tech writers who took enough interest in publishing systems to get involved in SGML's early days often felt that traditional talk about parsers and data structures didn't apply to their work. Ironically, their suspicion about tools and techniques that were too academic and difficult for them was coupled with an attitude about the technical difficulties of SGML that often intimidated non-initiates, squeezing themselves into a middle ground that had a difficult time expanding.

There are plenty of indications that XML has helped the world of structured markup expand in the direction of the people formerly intimidated by SGML. The Markup Languages conference showed how far it has grown in the other direction. As I headed toward the closing keynote talks, I discussed the growing comfort level between SGML people and computer science people with Steve DeRose of INSO, one of the important early figures in SGML to hold a PhD in a field related to computer science. He described how computer science academics of his acquaintance such as Brown University's Stanley Zdonik, an important contributor to object-oriented database research, had warmed to XML.

I didn't realize that a keynote talk were walking toward was by Dick Grune of Vrije University in the Netherlands ( http://www.cs.vu.nl/~dick/mt98.ps ), a key figure in the field of parsing computer languages, who titled his presentation Parsing-Who Needs It? More to the point, the subtitle was SGML/ XML and Computer Science , which Grune admitted to changing from SGML/ XML vs. Computer Science. (Examples he gave of XML's better fit with mainstream computer science and hence the classical approaches to software design and development were its avoidance of the ampersands, inclusions, and exclusions allowed by SGML element declarations' content models.) This change in his subtitle, as a symbol of some important progress in markup language technology, was a nice way to close the conference. <end/>