Getting Topical

What are XML Topic Maps?

Topic Maps provide a set of structures developers can populate with information about information -- metadata, to put it more concisely. Armed with such a map, both humans and computers can locate information by navigation among labels with well-understood relationships, rather than hoping for a successful keyword search. Topic Maps are similar to hypertext (and XLink) but operate at a level of abstraction above documents. Instead of searching for information by navigating links between documents, Topic Map users navigate links in a map, which then links to the information in documents.

Topic Maps provide flexible navigation among information items. Unlike classic indexing sites like Yahoo! or approaches like Gopher, Topic Maps uses a more hypertext-like approach: topics can have various relationships with many other topics, without any necessary conception of a hierarchical tree structure. It is possible to build topic trees using Topic Maps, but it isn't required. Instead, a "multidimensional topic space" is the model, allowing topics to connect among themselves and to information resources freely.

While Topic Maps can seem arcane, for the most part they represent a model most people are familiar with in reference works, though substantially enriched. While ordinary hypertext is like reading an encyclopedia and moving from article to article when an interesting keyword appears in the text, working with a topic map is like browsing a thesaurus and moving along the related words. When you've found the right word, you can connect to a resource that tells you much more about that word. You might have found the same information browsing an encyclopedia, but a thesaurus is often a more efficient way to navigate relationships, providing structured descriptions rather than large quantities of textual context.

Work on creating such tools in markup has been proceeding for a long time as part of HyTime SGML activity, culminating in the publication of ISO 13250:2000, Topic Maps: Information Technology -- Document Description and Markup Languages. A core group of developers, including several of the editors of the ISO specification, wanted to create an XML version of Topic Maps to bring this technology to the rapidly-growing XML community.

The Process

XML Topic Maps (XTM) launched from a meeting at last year's XML '99 conference in Philadelphia, attracting around 30 attendees after most conference-goers had moved from conference activity to dinner conversation. Those in attendance agreed that an XML version of Topic Maps was a desirable thing and decided to move toward the creation of an organization to support that development.

TopicMaps.org became the new home for XML Topic Maps development, "an independent consortium of parties interested in developing the applicability of the Topic Maps Paradigm to the World Wide Web, by leveraging the XML family of specifications as required." IDEAlliance acts as the host organization, continuing a role it has played since 1993, but participants do not have to be IDEAlliance or GCA members.

Work on the Topic Maps specifications, per its charter, has taken place in public view, on the XTM-WG mailing list, hosted by eGroups. A file repository holds additional information, like use cases, meeting agendas, and minutes of face-to-face meetings. The public has read access to working documents, while members of the authoring group form the core group building the specification. Joining the authoring group requires a two-thirds vote of its current membership.

Along the way, the scope of the conversion from ISO 13250:2000 to XTM grew. "It did not happen at all the way we planned -- which is normal, because it's always like that," said Michel Biezunski of Infoloom, one of the Editors of the XTM 1.0 specification. "The thing that we did was we have the same model. Instead of just simplifying it, we discovered that many things in the model were implicit, and we had to make that clear and explicit, which was a lot of work."

The work proved to be immense, with the group dividing work among a Use Case subgroup, a Conceptual Model subgroup, and an Interchange Syntax subgroup. The need for a processing model emerged as well. Steven Newcomb, Editor of the XTM 1.0 specification, described the trial and the excitement of separating this work and uniting it into a specification: "Conceptual model group and syntax groups worked separately. The sangfroid of this group must really be admired, because we didn't know if these would fit together until the Dallas meeting."

Newcomb acknowledged the stress of the process: "Everyone worked well beyond rated capacity, and there was considerable personal sacrifice and hardship involved in getting the thing published, even in its present, only-partially-finished condition." The results, however, seem to justify that cost, as Newcomb notes:"Our interchange syntax is tuned to an explicit conceptual model, and we also provided a processing model which expresses exactly how you take a syntactic instance and turn into something that's ready to roll, ready to use."

Biezunski also noted "a lot of passion... we have had an extraordinary group -- an unexpected number of bright people, all having their personal ideas about how things should be done. We have had some very animated discussions." Murray Altheim of Sun Microsystems, an Associate Editor of the XTM 1.0 spec, described "14 hours of teleconferences in three weeks -- 20 hour a day work sometimes. A lot of time and energy. It's been an amazing process."