This originally sounded like it was a proposal for a new form of URL.
Now, it looks like it is a proposal for a convention among HTTP
servers -- like CGI -- which HTTP server implementations could either
include or not, with the client completely unaware. The only issue is
whether proxy servers might be aware of the convention and deal with
overlapping byte ranges.
Except for the case of extracting individual segments out of larger
ascii files, though, I'm having trouble thinking of any real examples
of this, and, in particular, any examples where the documents aren't
text/plain.