URI Fragment Identifiers for text/plain

re
 http://dret.net/netdret/docs/draft-wilde-text-fragment-02.html

I've been trying to nail down why I'm uneasy about this,
and I think, in the end, that it is because the fragment
identifiers proposed are not very robust against changes
to the document. Perhaps this should even be a general principal
for the design of fragment identifiers.

With HTML, the location of A/name nodes is relatively robust,
since they label the context. I don't think char= or
line= are robust, and match='s robustness depends on
the regular expression.

The document doesn't really include any motivating
examples, so I'm stuck with examples in my mind where
one might want to make a reference to a text document
and call out "section 7 of internet-draft blah"; doing
so with a regular expression -- reliably -- is hard.
Yet I have trouble imagining applications that really
need 'char=' or even 'line=' without any matching
capabilities.

See, for example, Phelps & Wilensky on 'robust hyperlinks
and locations':

http://www.dlib.org/dlib/july00/wilensky/07wilensky.html

Larry