Arnt Gulbrandsen writes:
(That doesn't necessarily mean that the retrieved object must have a
certain content-type. I can imagine other ways to achieve the same
end.)
That was stupid of me. Concrete suggestions beat handwaving. Sorry.
My preferred approach would be to give archived-at the same parameter
list that MIME fields have, and define a "content-type=x/y" parameter
naming a type expected to be available. For a list that provides
archives in two forms and uses fancy HTTP to provide the right one, the
field could be like this:
Archived-At: ...; content-type=message/rfc822, content-type=text/html
For a less fancy archive providing both forms under separate URLs, two
Archived-At fields could be used:
Archived-At: ...; content-type=message/rfc822
Archived-At: ...; content-type=text/html
For today's usual sort of archive, the expected content-type would be
text/html by default:
Archived-At: ...
Arnt