fwiw, maybe you don't need an end at all in the
prolog, just a start - since you assume the topic end is the end of the
range, unless the next topic has the same start attribute
the start attribute is more of a canonical name for the indexterm
anyway - for arbitrary indexterms within the body, it's the appearance
of an end with the same name that turns it into span instead of a point
Dana Spradley wrote:
As I mentioned in this morning's meeting, I
really
like the proposal too, and thank Michael and Bruce for pulling it
together. It takes a creative approach to resolving all identified
issues with index ranges.
It would be a shame if we can't figure a way around Paul Grasso's
objection to the possibility that prolog-resident start/end ranges
might come apart.
Perhaps some combination/variation of my suggestion that matching
start/end tags in the prolog be required to indicate the topic is a
range, to be concatenated with the next topic if it is similarly
marked, with Gershon's (I think) suggestion that map ranges override
such prolog generated ranges, will do the trick.
--Dana
JoAnn Hackos wrote:
Hello Michael,
I am very enthusiastic about
this proposal. I believe it incorporates all the issues we have been
grappling with. Thank you for putting this together.
JoAnn
After some discussion with Bruce
Esrig, we've got a compromise proposal documented here:
http://wiki.oasis-open.org/dita/Indexing_issue_summary?action=show
This compromise is designed, among
other things, to leave the range/not range decision up to the author,
by allowing ranges to be specified on a per-topic basis as well as a
per-branch basis using the index range attributes.
See the heading "Proposal for
indexing ranges", text copied here for ease of discussion:
------------------
Index entries are interpreted as
point references. The index contains a reference to the point where the
index entry is declared. If an index entry occurs in a topic prolog,
the reference is to the start of the title of the topic.
Index ranges are structural. Most
index range declarations refer to an entire topic or set of topics. The
only exception is a range contained entirely within the body of a
topic.
Index range indications may occur
in
the topicmeta of a topicref at the map level. Similarly, they may occur
in the prolog of a topic. These two locations are architecturally
equivalent, so either indication may appear in either place, and a
match will still be recognized.
An index range start indication in
the meta information for a topic is interpreted to indicate the start
of the topic title. An index range end indication in the meta
information for a topic is interpreted to indicate the end of the
topic. This includes all subtrees of nested topics, including subtrees
of the start and end topics and any intervening subtrees. The subtree
of nested topics is included even when the range starts and ends in the
meta information for the same topic.
An index range may start in the
body
of a topic. Such an index range ends at a matching index range end
indication within the same body, or at the end of the body, whichever
comes first. Such an index range does not span sub-topics of the topic.
-------------------------
Michael Priestley
IBM DITA Architect and Classification Schema PDT Lead
mpriestl@ca.ibm.com
http://dita.xml.org/blog/25
Coming back to this issue today, I
agree
that a compromise is appropriate, given its relative triviality.
The only addition I'd make to Paul's previously suggested
implementation statement would be to put indesterm start/end rendering
in the hands of end users as well (my addition bolded):
"Index terms in prologs are neither ranges nor
points.
They are associated with the whole topic. DITA publishing
implementations are encouraged to let the end-user choose whether to
represent them as page ranges spanning an entire topic or individual
pages in an index. Publishing implementations
are also encouraged to let end users decide whether to render indexterm
start/end pairs as page ranges, or as point references to the start
indexterm. Another choice that publishing
implementations may wish to provide is whether to collapse multiple
continguous page references into a single page range."
--Dana
Paul Prescod wrote:
Original Message-----
From: Grosso, Paul [mailto:pgrosso@ptc.com]
Sent: Wednesday, August 16, 2006 7:01 AM
To: dita@lists.oasis-open.org
Subject: RE: [dita] Are indexterm ranges backwards incompatible?
...
Importance and ranges are orthogonal concepts.
I think that underlying this debate is a difference in styles by index
users. When I'm presented with many mentions of a particular topic,
there are three things that I look at:
1. is any bolded as being the "defining" instance of an index entry?
(hard to do in topic-oriented content!)
2. which comes first (also not necessarily informative in
topic-oriented content)
3. which is longest: likely to be a tutorial and not just a random
mention
So I understand Dana's point, but I don't (personally) think it is
crucial enough for substantial spec rewriting at this point. It is
totally true that if you have a Concept called "Cheese" then you would
want that topic to look special in the index entry for cheese. It is
also true that barring any special markup, making that mention into the
longest range is ONE way to make it stand out. Maybe we should agree for
DITA 1.2 to document other (more explicit) ways to make it stand out.
To put it another way: if a publishing tool provides the options I've
proposed then I would tend to advocate that they be set Dana's way
rather than the way others propose.
That said, I think it is acceptable to leave control of the issue in the
hands of the end-user rather than requiring it to be hard coded in the
spec. I thought that we were heading towards a compromise on those
terms.
Paul Prescod