Re: Using HTTP cache control for invalidating a cache of parsed objects

On 11/7/11 5:51 AM, Henri Sivonen wrote:
> Do we have infrastructure in place that allows a cache on a higher
> layer than Necko maintain cache coherence according to HTTP rules when
> the higher layer maintains a cache of object that have been parsed
> from HTTP responses so that if the cache is fresh (from the HTTP point
> of view) the objects don't get reparsed?

Not really.  We have a bug on attaching interesting binary data to necko 
cache entries.

> Does the image cache use something like this by any chance?

The image cache has a hack where it returns possibly-outdated data, then 
does the HTTP response and sees what the HTTP level says; if new data 
comes in it sends that on.  That sorta maybe works for images, but 
wouldn't work for non-idempotent resources....

> Should a higher-level cache even want to maintain HTTP-correct cache
> coherence in situations where determining cache coherence would
> require pinging the origin server with an etag and should a
> higher-level cache work with expiry times instead and simply not
> support etags?

Seems like that would be the sanest behavior, yes...

-Boris