jon said:
>   How is that?
>   Annotations can be linked to the text
>   using the markup as "hooks" (e.g., using XPointer.)
>   The more markup there is,
>   the more hooks to latch onto.

please show me -- and the original poster --
an implementation that actually works, now.


>   Pointing to 'id' ("fragment identifiers") is the most robust
>   and can survive various types of document edits.
>   In plain text systems, where annotations have to
>   hook to the content itself (rather than markup
>   which is separate from the content),
>   it is more difficult to prevent link breakage.

this is another case of disingenuous sleight-of-hand.

you are trying to make us believe that
the text changes and the markup doesn't.

what you've done, though, is merely specified that there is
markup which _cannot_ change (the "fragment identifiers"),
so as to assure link-permanence.  if i were to specify content
that can not change, i can guarantee link-permanence as well.

and in almost all cases, we're more likely to have text-invariance
than to have markup-invariance.  (but this is beside the point,
since it's easy enough to specify invariance of text and markup.
it is also very easy to show link breakage in cases of variance.)

-bowerbird