Re: old? heuristic in refreshIsCachable()

On Mon, 2003-04-07 at 19:30, Henrik Nordstrom wrote:
> mån 2003-04-07 klockan 10.15 skrev Robert Collins:
> > What prompted the heuristic of not caching responses that expire in
> > under 60 seconds?
>
> This only applies to objects which cannot be re-validated.

Uhm, then something is broken. I have objects that can be revalidated
getting released by this rule.

> The rule is because in a proxy the hit ratio for these objects is very
> low, close to none.
>
> > I ask because in rproxy scenarios, with surrogate-control headers, it's
> > quite common to give the surrogate (squid) a low expiry, to ensure
> > freshness to new visitors, while giving clients a high expiry, to reduce
> > surrogate load.
>
> In reverse proxies this rule can obviously be changed to 1 second.

0 works fine in my testing so far. Why 1?

Rob

-- 
GPG key available at: <http://users.bigpond.net.au/robertc/keys.txt>.