Loggerhead setting 'Cache-Control' header for static fields

Martin Pool mbp at canonical.com
Mon May 3 02:49:40 BST 2010


On 3 May 2010 11:41, John Arbash Meinel <john at arbash-meinel.com> wrote:
>> Ian's data (ooh, data) showed that many of the hits for the window we
>> sampled were for annotation of a particular file revision, which
>> should be both very cacheable and relatively expensive.  Clearly we do
>> need different caching regimes per type of page.
>>
>
> What wasn't clear was how often a given page was requested multiple
> times. (Vs. say just requesting 100,000 different file texts across all
> file revisions in all projects.)

Sorry if it wasn't clear.  What we meant to convey is that during that
week, about 60% of hits were for one single URL.  (Obvious caveats
here that this week may not be typical etc.)  That URL should be
highly cacheable and it should be fairly easy for loggerhead to know
this, without getting into anything fancy.  At the moment it is not
cached.  To me this is the obvious place to start: get in a squid,
make loggerhead set headers, reproduce this case and check that it is
in fact getting served from squid.

-- 
Martin <http://launchpad.net/~mbp/>



More information about the bazaar mailing list