You might want to actually try it out and see what kind of delays are
involved in firing off the MDA. Under peak loads where the bottle neck
would too many MDA's running at one time (most of the MDA code I assume
would be shared though) and under low loads when the bottle neck would
be of executing the MDA.
By the way sendmail is not the only MDA available - you would get many
MDA's with quite small footprint in case that becomes a bottleneck. What
you could also try out is use SMTP to deliver to your own mailserver and
then it can do the queueing. Ofcourse here your benchmark gets
complicated by the fact that you need to stress test under peak loads
for your webserver and mailserver both.
Mithun
PS: This has been discussed in the past and you might want to look at
the archives too.
