If the document has expired or if the document has no "Expires" header
field. In either case the "Last modified" date is separate.
> Isn't it better then to let the receiving cache generate
> a valid date & time from its internal representation, instead of
> keeping the original date string around?
Perhaps this is a local implementation issue rather that a protocol
issue? I agree with Gary Adams that it might be better if the
specification treated the string sent back in the "If-Modified-Since"
field as an "opaque server cookie." That way, the mechanism for cache
coherency can be independent of time representation. However, since the
receiving system may use the "Last modified date" for other purposes,
the string isn't always opaque to the receiving system.
AT&T Bell Labs