As Gavin Nicol has remarked (and I'm sympathetic to this point of view),
once anything gets large and acquires its own internal structure, it is
crying out to become a separate, properly typed object. And since
"fragment ids" are not properly part of the URI anyway, it seems
natural to allow them to be specified as separate objects if desired.
Thus, one solution would be to add a FRAGREF attribute to the standard
set of HTML3 linking attributes. The FRAGREF attribute's value should
be an address, such that the entity representing that address will serve
as the "fragment id" for the main addressing attribute of that element.
However, this scheme would really require proper definition of cid: URIs,
and proper MIME multipart support, in order to work efficiently.
Another solution would be to provide some sort of "URI extender" tag,
at least for <FIG>. While obviously much less elegant, this solution
may be rather tempting to the Java folks because it requires less
infrastructure overhaul, compared to implementation of the full multipart
yoga.
Anyway, I think now is a good time to resolve the "long URI" issue.
---------------------------------------------------------------------
Paul Burchard <burchard@math.utah.edu>
``I'm still learning how to count backwards from infinity...''
---------------------------------------------------------------------