On Jul 30, 2007, at 2:07 PM, Ricardo Newbery wrote:
On Jul 30, 2007, at 1:55 PM, Henrik Nordstrom wrote:
On mån, 2007-07-30 at 13:30 -0700, Ricardo Newbery wrote:
Regarding PURGE of Vary objects. The solution is just not to
include
a Vary in the PURGE, correct? So all variants are purged.
No, currently PURGE of variants is a bit broken.. you can purge
individual variants by including the matching request headers, but
the
variant index is also purged at the same time and as a result the
other
variants gets temporarily unreachable until Squid learns the
variance of
the URL again on the next GET request.. at which point Squid suddenly
finds many of the variants again in it's cache...
It should invalidate all variants, but doesn't.. just makes them
temporarily unreachable.
Regards
Henrik
Ahh... okay.
But this seems orthogonal to the HTTP/1.0 issue. Purging instead
via HTTP/1.0 still doesn't purge all variants, correct?
Ric
Also... does this mean that purging of gzipped content (Vary: Accept-
Encoding) is also broken?
Ric
