[Fwd: [dita-users] DTD practices for DITA]

I posted this on dita-users but did not get a response. Any opinions
from the TC? My concern here is that the prescribed specialization
procedure for DTDs is unnecessarily limiting. On the "specialization or
attributes?" thread, Barbara Douma of Compuware had remarks that seems
related:

About 8 years ago we moved to a topic-based SGML
CMS in which we had very heavily specialized topics, including an
Introduction, Concept, Main process, Process, Task-group, Task, and
innumerable reference topic types. The end result was inflexible. This
was also due to the structuring DTDs, which only allowed certain topic
types in certain contexts.

Chris

-------- Original Message --------

Hi. I would like to know what current practice is out there for using 

the DITA DTDs. My initial thought was that the typical DOCTYPE 

declaration would use the infotype-specific shell DTD, so a concept's 

DOCTYPE would look like:

    <!DOCTYPE concept PUBLIC "-//IBM//DTD DITA Concept//EN"
"concept.dtd">

A colleague pointed out its limitations. This shell DTD contains no 

information about other info types. For instance, a topic level element

can embed other topic level elements. But if you use only concept.dtd, 

you can embed only concepts. If, by contrast, you use this:

    <!DOCTYPE concept PUBLIC "-//IBM//DTD DITA Composite//EN" 

"ditabase.dtd">

Then the concept element can embed all the standard DITA topic types: 

topic, task, reference and concept, as the language reference says it 

should. My earlier assumption was that ditabase.dtd was for documents 

with <dita> as the root element, but it looks like it's the only
way to 

handle multiple topic types in DTD-land no matter what root element you

work with. We have been using ditabase.dtd for all our documents.

This seems to complicate specialization. The specialization exercises
in 

IBM's Deep Dive presentation prescribes a topic-specific shell DTD,
such 

as wiztask.dtd. But it seems to me that for interoperability with other

DITA topic types, you really need a superset of ditabase.dtd. This 

ditabase-superset.dtd would need to incorporate each new infotype that 

you create through specialization.

Chris