On Tue, 30 Sep 2008 00:36:10 +0200, Jonas Sicking <[EMAIL PROTECTED]> wrote:
Jonas Sicking wrote:
Yes, I think it would be helpful to add that information. It wasn't
clear that the credentials flag wasn't part of the key until you put it
this way (though the spec already clearly says so, just easy to miss).
I added a note that says what the primary key is now. (The credentials
flag is also part of it now.)
[...]
So in this case the credentials flag is actually part of the primary
key. I.e. the spec says to not update an existing entry if a request is
made with the credentials flag set to true, but the cache contains an
entry with the credentials flag set to false. Instead a new entry should
be created which will only differ in the value of the credentials flag
(and possibly in the value of the expiry time).
I suspect the simplest solution is to actually make the credentials flag
part of the primary key everywhere.
Indeed. It didn't seem to be worth the trouble to optimize for public non
credentialed requests for "URLs" that already have a credentialed cache
entry.
--
Anne van Kesteren
<http://annevankesteren.nl/>
<http://www.opera.com/>
