On 20 Jan 2006, at 15:14, Tony Finch wrote:
> On Fri, 20 Jan 2006, Giuliano Gavazzi wrote:
>>
>> I was going to remark that too, but is it going to cache negative
>> results? I
>> would not say so.
>
> Perhaps you should read the documentation, because it does indeed
> cache
> negative results.
well, you are right, but my "I would not say so" was based on
rational reasons.
+-----------------------------------------------------------------------
------+
|callout_negative_expire | Use: main | Type: time |
Default: 2h|
+-----------------------------------------------------------------------
------+
and it does not surprise that it is small, as it is worse to cache a
wrong negative result (in the case of a new local part) than too
cache a wrong positive result (when an address stops being valid).
A much higher value for the negative cache would be a bad thing indeed.
>
>
>> In other words, how to populate the cache?
>
> Traffic from spammers? :-)
yes, that would work if in any two hour period that the cache lasts
the spammers hit a large variety of inexisting local parts: very
unlikely. So in effect the negative cache could be non existent. The
requirement for its effectiveness is indeed that the majority of
wrong addresses hit are hit more often than every 2 hours.
It is ironical that instead the positive cache would work well if the
spammers are effective, that is if they know all valid local parts...
Giuliano
