DITA 1.2 includes the new subject scheme feature. I gave a presentation about the subject scheme feature on the 12th STC India conference in New Delhi 13th of November and how it can be used to manage and define the topic granularity. My experience is that many DITA users have problems in defining the size of a topic. How big is a topic? Is it two sentences or 10 pages?

In brief the subject scheme allows you to (among many things):

• Build a taxonomy of subject your content is treating or shall be classified into and
• Classify topics according to the subjects in the taxonomy.

To classify topics means that it becomes possible to retrieve, filter etc topics based on the classification. The metadata that is used to classify topics does no longer have to be hardcoded into any vendor specific CMS, nor do we have to hardcode the classifications into the topic prolog. Great!

The subject scheme also allows DITA users in defining the topics needed and to determine the size of a topic. The suggested approach means that the DITA user shall first identify the subjects that the content shall treat and then identify the topics needed. Topics can then be created and finally topics are organized in a deliverable map. This means that a definition of what topics are needed and how they shall be organized in deliverables are two separate steps.

The classification of topics also helps technical writers to create the short description. Since a topic can be classified to multiple subjects we can derive the short description from the different subjects the topic is classified into.

Also faceted search environments can be built using the classification. I wish CMS vendors could implement faceted search navigation into the CMS. A facet is then the same thing as a subject.

Many CMS currently available on the market are not subject oriented. The core of a subject oriented CMS is then the subject taxonomy. You register a subject in the CMS and then classify topics to a subject. In many CMS today a subject is scattered across the database as folder names, system metadata, condition variables and also DITA XML metadata. What happened to the single source of metadata?