Conref Push Via Keys and Implications of Normal-Mode Key Definitions

[Eliot Kimber <ekimber@reallysi.com>]

I've started trying to capture in a document more formal thinking about this
issue based on the initial conversation Michael and I had. I'm still working
out the details of my writeup but I think I can present my general
conclusions as follows:

1. By general principle that the details of addressing should not affect
semantics, it follows that whether you use @conref or @conkeyref and whether
or not you use @conkeyref with keys defined on resource-only topicrefs or
normal-mode topicrefs, the processing result should be the same. This argues
for Michael's general position, which was that all uses of a push-target
topic should reflect all pushes irrespective of the nature of the keys
involved.

2. However, because keys defined on normal-mode topicrefs do unambiguously
establish distinct use contexts for topics, it means that processors *could*
clearly distinguish different pushes to the same topic in different use
contexts. The question raised is should they and, if they should, under what
circumstances?

My initial reaction and the statement that started this conversation was the
realization that being able to get different pushes to the same topic in
different use contexts could be really powerful, because it would allow you
to unilaterally modify a single topic in different use contexts where
otherwise you would have to physically copy the topic or use some really
ugly conditional processing to get the same effect. Or use a feature we
haven't invented yet.

However, in thinking it through more carefully and trying to enumerate the
use cases, I am convinced that Michael's position is the only reliable one
given the current DITA feature set. In particular, I identified two use
cases that would be very challenging:

1. Two authors happen to use different normal-mode keys to construct conref
pushes but intended their pushes to apply in all cases.

2. Some pushes to a given topic are via resource-only keys and some are via
normal-mode keys that then use the resource-only keys. In this case it would
be ambiguous at best what the intended resolved result would be since there
could be an effective layering of pushes, which is way beyond anything we
can reasonably impose on the 1.2 semantics just through implication.

So my conclusion is that for DITA 1.2 conref push must *always* apply to all
uses of a topic regardless of the addressing details. That is, I have come
to agree with Michael's position.

However, I think there is a general requirement for the ability to impose
use-context-specific content onto topics. However, I'm not sure conref push
the right mechanism, even if we could think of way to make authorial intent
clear.

I've been thinking about this requirement generally and hope to be able to
make a more complete and concrete proposal for new features for DITA 1.3
that should address the requirement.

Cheers,

Eliot

-- 
Eliot Kimber
Senior Solutions Architect
"Bringing Strategy, Content, and Technology Together"
Main: 512.554.9368
www.reallysi.com
www.rsuitecms.com