i like handling this as links. eg, HyTime.
if the search engine and the wwweb server had an explicit conception of a
link then a simple LINK, UNLINK protocol could handle this and a lot more.
this is another something ill write up a draft for one of these days.
HTTP 1.1 has a DELETE and PUT, which is the implicit link approach which,
IMHO, is weak. :-) superficially it's identical to unlink something and to
delete it. but links can be redirected which can't be done with put and
delete. i mean, i reorganize my wwweb and send a redirect request that
says that for documents with references to the old URL please replace with
the current URL. why don't i send redirects to every wwweb server in the
universe? because with LINK i know who supports linking and has made a
link into my page from what page. so i only have to send redirects to
people who have registered as linking this page. so the redirect command
could be a syntactic and semantic extension to UNLINK. so UNLINK works
like DELETE but if you want, you can use this new URL.
LINK <url source> <url target>
UNLINK <url target> [ <url optional target> ]
i think that the shape of this kind of solution implies that a wwweb search
engine, for example, isn't keeping link tables, just supporting UNLINK. i
mean, they would keep link tables for any static pages that would be ALLOWs
in a robots.txt.
i think this is the most important thing, that link tables are minimized.
of course, HTTP 1.1 doesn't have link tables because it doesn't have
links. :-))
-john