RBL cache {Scanned by HJMS}

[Ugo Bellavance]

Hi Anthony.

  Your answer is appreciated, but I am not talking about DNS.  It is based
on the same principle, but with different data.  I do not want to cache a
name-to-IP adress mapping, I want to cache the names or IP address of
servers that has been qualified as spam source in the last, say 4 hours.
Not to prevent DNS queries, but to prevent RBL queries.

With my suggestion, the RBL process would be like that:

Let's say MailScanner has a batch of 30 messages to process.  It will first
check with its local cache if any of the source servers of these 30 messages
has been qualified as spam source in the last 4 hours.  Then, it will
perform a RBL check only for the servers that were not in the local cache.

This way if you have only one message in your batch, and this server has
been subject to a RBL query during the last 4 hours and has been identified
as a spam source, then you don't have to make any RBL query at all.

--We could also cache servers that have not been identified as spam, maybe
with a lower TTL.  With this method, the mailserver of this list would never
be RBL-queried even if I don't put it in my whitelist.

You might tell me that you never get only 1 message in a batch, but since I
run a small server with only a handfull of users, I do.  On the other hand,
people with big servers who get thousands of spam messages a day would
shorten their connection times with the RBL servers.

Since english is not my primary language, maybe I don't express myself
explicitely enough.  Sorry about that.

Thanks,