Assuming that the spec allows DTDs to be configured so that nested topics won't be present, I have found a design flaw in the DTD implementation pattern for *-info-types. Consider this example from topic.mod:
<!ENTITY % topic.content "((%title;), (%titlealts;)?, (%shortdesc; | %abstract;)?, (%prolog;)?, (%body;)?, (%related-links;)?, (%topic-info-types;)*)">
The problem with this construct is that it will not allow %topic-info-types; to be set to empty. The presences of the comma, the parenthesis, and the * occurrence-operator mean that a valid element name must be furnished by the %topic-info-types; parameter entity or the DTD will fail to parse. To correct this, the definition for %topic-info-types; must include the comma, the parenthesis, and the occurrence operator. For example:
<!ENTITY % topic.content
"((%title;),
(%titlealts;)?,
(%shortdesc; |
%abstract;)?,
(%prolog;)?,
(%body;)?,
(%related-links;)?
%topic-info-types;)"
>
The parameter entity definition in topic.mod becomes necessarily ugly:
<!ENTITY % topic-info-types
", (%info-types;)*"
>
As also does the one in topic.dtd:
<!ENTITY % topic-info-types
", (topic)*"
>
This flaw has existed throughout DITA 1.2 and lives on in DITA 1.3. It also suggests that few, if any, have tried to turn off topic nesting.
I discovered this while researching George Bina's question about the "topic" element being available at the end an embedded task topic in a troubleshooting topic. That flaw was caused by the fact that I had neglected to override the the task-info-types parameter entity in troubleshooting.dtd. Specifically, the following parameter entity needs to be added to the TOPIC NESTING OVERRIDE section in troubleshooting.dtd:
<!ENTITY % task-info-types
"task"
>
Without that override, task.mod defaults the task-info-types value to %info-types; allowing topic to appear as nested topic at the end of task. Of course, my preference would be to set task-info-types to empty in troubleshooting.dtd. And, that's how I discovered that the DTDs won't parse with empty *-info-types.
At any rate, the fix for troubleshooting.dtd and the *-info-types problem need to go on the TC agenda soon.