Cascading attribute metadata that merges (or not)

[Robert D Anderson <robander@us.ibm.com>]

Hi,

I'm looking at comments in the "metadata cascade" topic, and realized that the definition of the new @cascade attribute left at least one case unspecified:
https://ditaweb.com/oasis-dita/#/00028678-DB.00028004-DB.Cascading%20of%20attributes%20and%20metadata%20in%20a%20DITA%20map

Specifically, what is the evaluated value for @platform on the second topicref in this scenario?
<map platform="A" cascade="merge">
  <topicref platform="B">
    <topicref cascade="nomerge">

When the cascade is evaluated, the @platform value for <map> cascades to and is merged with @platform="B" on the first topicref, resulting in an implied value of platform="A B".

For the second topicref - does this already merged value "A B" cascade? Or does the "nomerge" token mean that only the most recent (unmerged) value "B" cascades? One of these must be true (@cascade does not turn off the cascade entirely, it only controls whether cascading values merge or not).

I think that the already-merged value cascades to this topicref. Essentially, the meaning of the cascade is that the first topicref applies to both platform A and platform B. It is equivalent to having an explicit setting of platform="A B" on that element. Thus, I think that once the cascading is calculated, the sample above is equivalent to the following:
<map platform="A" cascade="merge">
  <topicref platform="A B">
    <topicref platform="A B" cascade="nomerge">

In other words, the algorithm here is essentially:

• Establish @platform for <map> -- the value "A" is explicit.
• Establish @platform for the first <topicref> -- using the rules we've laid out, the value is "A B" (cascading "A" from parent merges with explicit "B")
• Establish @platform for the second <topicref> -- cascading "A B" from parent is used because there is no @platform value specified here. We don't ignore the already-established "A B" value from the parent.

My initial thought was just to update the topic with this example, but I am running it by the TC as a sanity check because I'm sure the example was not discussed previously.

Thoughts?

Robert D Anderson
IBM Authoring Tools Development
Chief Architect, DITA Open Toolkit (http://dita-ot.sourceforge.net/)