- From: Jeffrey Mogul <mogul@pa.dec.com>
- Date: Thu, 11 Apr 96 11:57:36 MDT
- To: "Henry Sanders (Exchange)" <henrysa@EXCHANGE.MICROSOFT.com>
- Cc: http-caching@pa.dec.com
The suggestion stems from my view that max-age=0 means 'must revalidate'
and that we should spec it that way. I'm not convinced that a client
that would ignore max-age=0 spec'ed as 'must revalidate' wouldn't also
ignore a 'must-revalidate' directive.
The reasoning behind this is that we've already agreed that max-age=NNNN
can be overridden but not "ignored". That is, a cache can (for
what should be good reasons, not just because it's feeling goofy
today) use a stale entry in a response, but if it does so, it MUST
mark it as stale using a Warning: header.
must-revalidate doesn't give the cache this kind of latitude. For
a non-end-user cache, it has NO latitude. For an end-user cache,
such as in a PDA, we give a tiny bit of latitude because we've been
warned that some PDA people will take liberties anyway. But we make
it clear how far they can go without making Koen's example into a
major disaster.
'Must-revalidate' strikes me a lot like those 'enforced by radar'
notices on speed limit signs - if you're going to speed these
probably won't bother you.
I'd prefer to view it as "Do not back up - Severe Tire Damage
will result". (For more information, see http://www.std.org/ .)
In other words, "we can't stop you from putting your car in
reverse, but you'll probably regret it and you can't sue us
for damages."
-Jeff
Received on Thursday, 11 April 1996 19:15:57 UTC