Two things come to mind. The first is IAFA/IIIR templates (see
internet-draft draft-ietf-iafa-publishing-01); they provide a
reasonable standard, concise, portable and extensible mechanism for
exporting information about objects.
The other is only a bit less obvious. I was thinking about GLIMPSE
and how it uses two levels of indexing, the first only narrowing down
approximate location in which words are present. Wouldn't it be great
if there were a way to do a GLIMPSE-like search, but one in which a
block corresponds not to an arbitrary portion of the information space
but a specific server to which the query can be fowarded? And what if
you could generalize this, to work not just at two levels, but in an
arbitrary hierarchy? An arbitrary heterarchy?
I actually spent at least an hour hashing out the details before
realizing I was reinventing whois++.
Both of the above, of course, are hardly finished being developed and
tinkered with, but closer than anything we could reinvent from scratch.
-- <A HREF="http://www.cs.indiana.edu/hyplan/mvanheyn.html">Marc VanHeyningen</A>