OASIS Darwin Information Typing Architecture (DITA) TC

Additional Details on Intent or Implications of Topicrefs in Relcolspec

  • 1.  Additional Details on Intent or Implications of Topicrefs in Relcolspec

    Posted 02-18-2015 23:33
    I'm writing a lesson on relationship tables and thus doing a fresh close reading of the reltable language reference entries. The <relcolspec> element can contain zero or more topicrefs. The 1.2 spec language says: "Beginning with DITA 1.2, you also can add <topicref> elements to the <relcolspec> element; this defines a relationship between the topics that are referenced in the <recolspec> element and the topics that are referenced in the column of the relationship table." It also defines rules for using any navigation title or referenced topic title as the title for column (if no <title> element is present). My question: what was intended by "a relationship" in the paragraph above? Was there any specific type of relationship that was intended beyond simply providing a column label? I don't recall any of the discussion that must have gone into this addition in 1.2 (either because I wasn't involved or because it was so long ago). My guess is that the intent was that linked topics are in a "type of" relationship with the relcell entries for the column, e.g., if I have a topicref to the topic title "Subtask" that explains what a subtask is, then it is establishing that topics referenced from that column are instances of the type "subtask" in the context of the relationship table. But the 1.2 language certainly doesn't limit you to that interpretation. For example, there could be an implication for some navigation relationship or something. I don't think we need to limit the type of relationship that can be implied in this case but I think we need to add some commentary that either reflects the original intent or provides some typical use cases and what a presentation or behavior effect might be beyond simply setting the column label. The example in the 1.2 spec shows using a topicref and says there's a relationship, but doesn't say what the relationship means or what presentation or behavior implications it might have beyond providing the title. Also, the language as written seems to assume that there is at most one topicref in the relcolspec, but relcolspec allows zero or more. The obvious rule is that the first topicref provides the title if there's no <title> element, but the spec doesn't say that now. (I'm actually surprised that Jarno hasn't noticed that--maybe he hasn't had to touch the relationship table processing recently.) In the first edition of my book I just use topicheads in relcell, this emphasizes the title-setting aspect. But looking at it now, it would make more sense to just use <title> in that case. Cheers, E. ————— Eliot Kimber, Owner Contrext, LLC http://contrext.com