>> Ideas? The above does not feel good to me, and neither do PI's or
>> paired tags.
>
>What does not feel good about it? I think it's a very
>good approach.
Paired elements like the above mean that you have to maintain more
state when you are rendering a document (ie. you have to traverse the
entire document tree preceeding the current element, whereas with
container based control, you only need to traverse the ancestors
leading up to a node). The reason I don't like word offsets of for
much the same reason.
This is not much of a problem with small documents, but for optimising
the redisplay of large documents, it is a very important issue. This
could also be important for optimising redisplay when scrolling.
One fellow mentioned that we need a way to solve both the
highlighting, and the annotation problem. I think using parts from
HyTime Location Module, or at least concepts from it, combined with a
multipart MIME message might be the best way to go (and it's
applicable to other media types!!!). One way or another, specifying,
and implementing, such functionality will cause some pain.