sebastien said:
>   Now suppose that in a few years time
>   1/ this _Study of ancient Greece_ gets into PG, number 78967
>      (in the same edition John Smith used)

you've swept under the rug what could be a very important disqualifier.


>   2/ some text-crawling program detects its fuzzy quotation
>      in the Smith's essay
>   3/ some robot and/or human reworks Smith's essay to include
>      hyperlinks to PG text 78967, at the correct page number
>      (which would land at the right paragraph in the HTML)
>      and checks them one by one

why require those changes to be made after-the-fact?
why not build a program that can do all of that _on-the-fly_?
without requiring any heavy markup?  that's what i'm aiming at.
(i've got aspects of this approach working already in my stuff.)

the ability of the machine to find links like this automatically
-- i.e., without us specifying them -- is extremely important!
we want an intelligent viewer-program that knows what we want,
_without_ us having to spell out every aspect of it in great detail.


>   That would be a killer ultimate library!

well, it's one step _toward_ that killer ultimate library, anyway.
there are a lot more steps that need to be taken on that journey.

-bowerbird