[Date Prev] | [Thread Prev] | [Thread Next] | [Date Next] -- [Date Index] | [Thread Index] | [List Home]
Subject: external indexes (was: RFE 714764 - 'typed' indexes)
Katharina Udemadu wrote: > I am not sure that I understand what you mean by separating the > document from its indexes. Indexes are content-based, and I cannot > think of a way that would allow you to create an index without > inserting index term markup in the document. It's just an idea; no thought is given to details yet (if ever :). If I ever would need more than one index for a document (let's say it's a book), then I guess I could do it something like this: * Make no changes to the book itself. * Create two or more indexing documents. Each would probably be written in some custom XML lang (since the mechanism is not (yet?) standardized). It would assign index data to portions of the document, by pointing to them via XPointer (Dave's idea), and perhaps regexen. * Then I'd have to implement the processing :) * ... and run several passes: the book with index one, the book with index two, etc But now, when thinking of the details, it becomes clear that there's a high price in the form of (implementation and) maintenance to be paid for separation of the index data. Some of the XPointers and regexen would not have to be changed when the content they are pointing to moves or is changed, but others would have to be changed. So there probably are ways to do it, but it might be tedious. Perhaps one could tag intex terms (to eliminate the regexen), then have external sets with different additional data (type etc) ... The whole thing definitely would complicate stuff. The cost in terms of implementation and maintenance (of the indexes) would have to be justified in the specific scenario. If the book doesn't change much, and there really is a need for multiple indexes, then the cost is lower. But then the benefit of separation decreases as well ... Again: All this is not a well thought out proposal, but just an idea. Even if convenient syntax could be found and easy maintenance could be assured (perhaps in some graphical tool), it would represent an alternative to the current mechanism, not a replacement for it. > In my eyes, the proposal to add an attribute to index and indexterm in > order to classify different types of indexes is reasonable. Could very well be; I didn't comment on that. I just commented on something Michael said, which gave me the idea of external indexes. Sorry for the confusion; I changed the subject line now. Tobi
[Date Prev] | [Thread Prev] | [Thread Next] | [Date Next] -- [Date Index] | [Thread Index] | [List Home]