David F. Skoll wrote:
John Rudd wrote:
[...] This actually far outweighs the IO bottleneck of clamd's
socket.
Hardly any data flows over clamd's socket. MIMEDefang just sends the
command:
SCAN /path/to/filename
and clamd reads the file or files to scan itself.
General comments:
1) Disk I/O: MIMEDefang will use less disk I/O than Mailscanner because
you're not queueing every mail message twice. In fact, if you have a modern
version of Sendmail with SuperSafe set to PostMilter and MIMEDefang rejects
or tempfails a message, that queue file won't be sync'd to disk.
In my experience, this wasn't really a problem. Mailscanner wants your
mqueue.in, mqueue, and scratch area to all be on one partition so that
it can just do "mv" commands to move things around. So you don't really
re-queue much if you don't want to (depending on how you hand things
off to the outgoing sendmail process).
But:
a) there's some extra complexity in running extra sendmail processes,
and only wanting to stop one piece of the puzzle but not all pieces of
the puzzle.
b) MIMEDefang has a BIG win in that its scratch area can be in memory
(via RAM disk or tmpfs), whereas MailScanner really wants/needs to be on
physical disk (due to how it stores messages as they're scanned).
Regards,
David.
Btw: thank you for not squelching this topic. When I tried to have a
similar discussion on the MailScanner list, about a year ago, the
maintainer pretty much banned the topic from discussion (even though I
was, at the time, advocating the mixed approach and not an approach of
"abandoning MailScanner"). It has left a lasting bad impression for me.
_______________________________________________
NOTE: If there is a disclaimer or other legal boilerplate in the above
message, it is NULL AND VOID. You may ignore it.
Visit http://www.mimedefang.org and http://www.roaringpenguin.com
MIMEDefang mailing list [email protected]
http://lists.roaringpenguin.com/mailman/listinfo/mimedefang
