Invalidation, Location and Content-Location

[mnot]

The tests show that support for invalidation on Location and Content-Location is really poor. In my experience, developers don't count on this, and in some case it's counter-productive (because a POST response might want the client to use a cached resource, rather than invalidate).

Should we drop the requirement to invalidate these URLs?

[MikeBishop]

Requirement, perhaps. I'd still keep it as a SHOULD, or minimally a MAY, though.

[mnot]

Calling it a SHOULD wouldn't exactly help interoperability. MAY makes more sense -- e.g., "Caches MAY invalidate other URLs, particularly...".