From: Daniel Quinlan
Sent: Monday, 5 July 2004 8:09 a.m.
To: Scott Truman

>Yes, but our loading is little different than any MTA-based DNS
blacklist implementation.  In contrast to
>some, we also never send a second DNS query.

>Sure ... only if (a) network tests were not I/O-bound, (b) if network
test I/O cannot be done in the
>background while CPU-bound work was being done, or (c) both.  Since (b)
is true and because the best way to improve
>performance is to send out all requests before doing the local tests
and collect the results after the local tests are
>done, the optimization you're proposing isn't an option.

I see what you are saying, but you are assuming that everybody needs
that blazing performance. All the sites I have spamassassin on would not
be hindered at all by running net checks after the local tests. I'm not
saying make it the default option, but an option none the less. As it
stands it sounds like it is too hard to implement, but if that
particular code was to be revisited for some other reason, then perhaps
it is something you might want to keep in mind. If it's a case of one
size fits all then obviously the concept has no  use; it's either net
tests on, or net tests off with nothing in between.

> The other way to put it is that by the time you know that local tests
could have done the job, it's too late.

As above - if the net tests _could_ be run sequentially after the local
tests, and the accumulated score was such that net tests didn't need to
be run, then they wouldn't be run.

Cheers
Scott

Reply via email to