At 10:27 AM 2/24/00 +0100, Frederic Gobry wrote:
>Basically, it would at least involve two parts : dealing with reference in
>the text (ie inserting things like [ItsARef]) and generating the
>bibliography in itself at the end of the document.
Yep. This sounds pretty similar to the design problems we'll need to be
tackling when it comes time to implement the following sorts of features:
table of contents
index
etc.
In each case, there will be fields sprinkled through the text, plus an
automatically-generated segment of the document somewhere else.
As you may have noticed from scanning the archives of this list, we're
currently trying to figure out how to revise our fields implementation so it
can handle the first problem. AFAIK, nobody has put any serious thought
into the latter problem yet.
Given what still needs to happen between now and our 1.0 release, I doubt
that I'm personally going to be able to make much headway on this issue
anytime soon. However, that shouldn't stop you or anyone else who's
interested in getting this feature working sooner. :-)
We're always looking for more help, especially from people who know enough
about a particular feature to really think through the problem. So, here's
a quick, off-the-cuff guess to get things started.
It's our belief that fields should be thought of as a calculated span of
text, represented as a special container for (formatted) characters. The
contents of that container exhibit the following UI behaviors:
- the contents are *always* automatically generated to begin with
- the contents can be updated automatically, or via a popup menu
- the contents can be manually formatted
- in some cases, portions of the contents can be manually deleted
- otherwise, the contents are not directly editable
- the container has extra properties which control all this behavior
Reasoning by analogy, I suspect that we could think of bibliographies, TOCs,
indexes, etc. as a special type of calculated *section* (which contains one
or more paragraphs). I haven't messed with any of these features in a
while, but here again, I suspect that the entire preceding bulleted list of
desired behaviors applies. YMMV.
Presuming that this insight is correct, then once we have these two basic
mechanisms in place (for calculated fields and calculated sections), it
should be reasonably straightforward to figure out the APIs needed to
communicate with your bibliographic software.
Paul,
motto -- many hands make light work
