|
Michael Priestley, IBM DITA Architect, is based in Toronto, and is our local luminary. He recently took the time to talk to SiberLogic about DITA.
It's no secret that IBM was the birthplace of DITA - and that Michael was a catalyst. As an information architect and XML expert (with experience in topic-oriented writing, task-oriented information design, and single sourcing) in IBM Toronto's Software Development Laboratory, he was involved with documenting software products using both HTML and the SGML-based DTD, IBMIDDoc, back in the late 90s. This documentation system had been in operation for many years, and Michael and his team started thinking about a more flexible approach - something that would allow the separation of content from its presentation and support the concept of centralized control of reusable assets.
We spoke about some of the "Eureka" moments - those defining points or ideas that constituted a quantum leap towards the DITA of today. Here's how Michael saw them:
The IBM documentation team was already typing its information, and had defined concept, task, and reference topic types. IBMIDDoc had good support for more general document types, and Michael's team needed both: strong typing, plus a flexible base. Michael had previously worked on the Open Class Library documentation (a common code base to develop native applications for multiple platforms), and the idea of putting the base topic type as an ancestor of the more specific concept, task, and reference types led (Eureka!) to the class mechanism, which reconciled the two conflicting requirements.
Michael was in favor of strong mapping that limited how developers could extend the information model, in order to enable a future-proof view of information structures. By combining an OOD approach (with which he was already familiar) with XML architecture, specialization was conceived. But inheritance had to be carefully controlled for it to work in XML, and the team figured out (Eureka!) the specific restriction rules (i.e. what kinds of changes a specializing type can make without breaking the inheritance model) that allow DITA constructs to be extended to embrace an industry-specific, or even organization-specific vocabulary, whilst maintaining common processing behavior.
Once it became clear that significant reuse would be possible, it became a matter of how the information could be assembled for delivery. Michael and his team considered various technologies, including RDF and XLink, before coming up with the concept of (Eureka!) maps that used the document structures that info architects use to plan or design information sets, as the instructions that the build process could use to assemble and integrate information sets, So, instead of generating a report document that shows how your Website links together, you would author a sitemap document
that directly determines how your Website links together.
According to Michael, the final Eureka moment was that of developing the map content model around relationships, rather than components - i.e. rather than creating a report or design structure that simply mirrored the topics one-to-one (i.e. for each topic, define what it relates to), they moved to a relationship-centric view (i.e. for each relationship, define what topics are involved). Instead of one topic having multiple relationships, it became possible to have one relationship involving multiple topics.
DITA was developed to support best practices, and now it is helping to drive those best practices (topic-based authoring, information typing, architecture-driven development). Moving forward, Michael believes this process will continue to cycle, with improving practices driving improving technology driving practices again. For example, the question used to be "What have I got that I can reuse?", and now it is changing to "I need these topics; where are they". In other words, documentation teams can develop to the deltas, rather than starting over and rewriting - significantly reducing the cost of innovation.
Michael says he is much gratified that the initial concept developed by his team was so scalable. But on behalf of techdoc teams everywhere, we'd like to add one more Eureka moment - the one when IBM decided to share DITA with the community...!
Our thanks to Michael Priestley for sharing his time and enthusiasm! Note that Michael is not affiliated with SiberLogic, and does not endorse our products (or any, apart from IBM's!) He is co-author of the DITA spec, co-moderator of the DITA-users mailing list, and sits on the OASIS DITA Technical Committee, which is driving development of the DITA standard. To see Michael in person, sign up for his highly-recommended full-day pre-conference workshop at DITA 2006, or visit the Toronto STC Spring Conference (March 27-29) where he will be giving a half-day workshop, including a live demo of how DITA can be used for information architecture and content development.
|
