RE: [dita] indexterm page ranges have come apart
I don't have a problem with
filling in gaps in the approved proposals. What I am uncomfortable
about is that we are also reopening and possibly reversing major
architectural decisions made along the way.
-
The question of whether
see/see also elements should have textual content or an explicit
linking/href scheme was a fairly long-winded fight that settled on the
former. Our last discussion has reopened the question again, with the
same arguments being traded. This is not new.
-
The question of what to do
with index-sort-as expressions that differ was also discussed long ago,
and the compromise was to have universal sort ordering set at the map
level and individual sort ordering at the topic level. So the issue of
"conflicting" sort-as elements is nothing new (sort orders cannot be
"inconsistent" unless they duplicate the entry). This question too has
been reopened.
-
The question of whether to use
linking vs textual content for page range start/end was also discussed
and settled before. Again, this is not a new question. Again, the old
arguments (precision of reference vs ease of use) are being traded.
So that is what I meant by
everything being reopened. Again, I am happy with discussions that
clarify gaps in the proposals. What I am not quite happy about is that
this is not all we are doing.
Chris
It was inevitable that all
issues be reopened because the design did not address questions like
what happens if index-see appears on a primary index item with nested
secondaries). We all resisted discussing elements other than
index-range until Bruce convinced us (correctly) that those issues had
to be addressed. Similarly, we have spent a lot of time discussing what
happens if you have conflicting sort-as because the answer wasn't in
the original design (it discussed "inconsistent" sort orders but not
duplicated entries). I don't see what choice we have in our process but
to re-open a design that isn't complete yet.
Unfortunately, we have a habit
in the TC of hoping that everyone else has thought through the
specification carefully. When you combine this with vacations which may
take a stakeholder out of the loop, you end up with the situation we
are in now.
The range issue is similar. Paul
Grosso started asking: "What happens if you do X (where X is nesting,
processing instruction, comment, Unicode normalization)" and it became
clear that the proposal had some undiscussed issues. Going all of the
way back to DITA 1.0: even for relatively simple cases we lack a
normalization algorithm that will answer the question of whether the
index items "a b" and "a b " and "a <!-...-> b" are the same
index entry. (this problem becomes much more acute when combined with
the DITA 1.1 features)
It's hard to know whether
indexing is the only issue with these problems or whether deeper poking
around in other specs would turn up similar ones. e.g. bookmap makes me
nervous...how many people are confident with the thoroughness of their
review of bookmap?
I agree that we should not
re-open the issue of whether DITA will have ranges. We voted, we have
no new information since the vote, no new issues have arisen other than
detailed implementation issues. Let's fix the details and move on.
Contrary to your subject title, it is not only
the page ranges issue that has "come apart". At this point in the
subgroup's discussion, ALL indexing proposals -- see/see also, sort
order, page ranges -- are back on the table. The old points of conflict
are being revisited, and we are back to square one, questioning every
proposal.
Indexing is a contentious issue. One of the original architects of
DocBook once told me she was glad she did not have my task: it was just
too ugly. The existing indexing proposals are a mix of hard-fought
compromises that nobody involved was fully satisfied with. Now that the
issue is reopened, all hell has broken loose again. That's natural for
an issue with so many unsatisfied opinions.
It's probably not so much one issue that has "come unglued", but
perhaps this TC's process. The votes to approve these proposals were
unanimous. By contrast, we never voted to reopen any of indexing issues
-- let alone ALL of them -- and never agreed on any kind of charter or
limits. Yet right now, anything goes. That is what bothers me: nothing
is binding, anything can be reopened at any time by anyone without a
vote.
Some people here will be happy drop page ranges from the standard.
That's easy enough to suggest. But I'd point out that we have not voted
on that idea either.
Chris
-----Original Message-----
From: Dana Spradley [mailto:dana.spradley@oracle.com]
Sent: Thu 8/10/2006 1:09 PM
To: dita@lists.oasis-open.org
Subject: [dita] indexterm page ranges have come apart
I'm sorry, but it's been summer, and I haven't been paying close enough
attention to the indexterm page range debate that started to heat up
just before the 4th of July.
I've now reviewed that debate, and I can understand Chris's frustration:
this issue seems to have come seriously unglued.
Worse, I find myself in disagreement with several TC consensuses that
were made in mid July, I think in my absence - if not, then I was
dozing.
While much of the debate has been technical, for me it comes down to the
fundamental issue of whether support for page-ranged indexterms is
compatible with two fundamental principles of DITA:
* DITA is topic based
* DITA encourages a minimalist approach to documentation
Both of these principles speak against supporting page-ranged indexterms
at all. Topics should be short (minimalism) and independent
(topic-based). How far do you need to read from a point-based indexterm,
except to the end of the topic? And if you're ready to try using the
device/software before you reach the end of the topic - well, more power
to you. You can always come back and read some more.
Chris's references to the Chicago Manual of Style are symptomatic to me
of what's wrong with this proposal. Its history is as a style guide for
writers of academic books. Professors being longwinded and disorganized
- I should know, I was one once - page ranges are appropriate in their
indexes.
Like transitional text that would appear only in book-length output,
which we earlier rejected, page-ranged indexterms are a throwback to the
kind of books DITA was meant to move beyond.
I propose we drop them from 1.1, and keep them out of DITA for as long
as we can.
--Dana