On Jul 10, 2006, at 1:26 PM, Michael Thomas wrote:
Eliot Lear wrote:
Bill,
I think what Dave could say is that in I/O intensive environments
DKIM
will have a negligible impact because minimal disk I/O is
required. If
your environment is not I/O bound then of course DKIM will have a
marginal impact. If you're using your CPUs efficiently already,
good on
you!
The point that Bill brings up, however, is a good one and I suspect
that he's
entirely right if one considers the _inbound_ which is very likely
to be very
CPU bound with anti-spam/virus/malware/etc. Still in that case,
it's not clear
that the DKIM overhead is likely to be very significant as the main
cost for
smallish messages may not even be bound to the RSA and/or SHA overhead
(seeing how RSA verifies are so much cheaper than signing).
Arguably, DKIM allows easy whitelisting of high-volume senders, allowing
the recipient to skip content-based spam filtering for that mail. That's
likely to be a significant fraction of inbound mail, so may reduce
the CPU
overhead overall.
In any case, I think he's right that in order to give advise we'd
do well to
actually state the test cases for which somebody could do their own
experimenting.
Definitely.
Cheers,
Steve
_______________________________________________
NOTE WELL: This list operates according to
http://mipassoc.org/dkim/ietf-list-rules.html