Re: default MTA for sarge
* Craig Sanders
> actually, the whole thing runs as root and drops priviledges or changes
> UID as needed. that's good, but it is potentially exploitable.
Indeed, it runs as root for as long time as it need to create a listening
TCP socket and setuid() to mail. From that point on, it runs as a
nonprivileged user. (Orelse 'ps' must be lying to me.) Something tells me
that Postfix, and every other daemon that intends to listen on a privileged
port must do the same.
And as for the potential exploit in setuid() -- reversing it? -- I'm sure
the kernel guys would like to know just how you intend to go about
exploiting it. The manual page says it's impossible, see.
* Craig Sanders
>> > exim is certainly not fast, and while it may be adequate for tiny mail
>> > systems with trivial loads, it doesn't scale up to large mail systems -
>> > which is an important point, debian is better off with a default MTA
>> > that can handle any load thrown at it.
* Tore Anderson
>> This is, of course, bullshit. Care to support your claims with anything
>> meaningful?
* Craig Sanders
> http://www-dt.e-technik.uni-dortmund.de/~ma/postfix/bench2.html
>
> on the same hardware, with the same test loads, postfix is 2-5 times faster
> than exim.
Congratulations, you've successfully proved that a guy with his own
Postfix fan-page and who's been hacking some on Postfix itself is capable
of setting up Postfix to run 2-5 times faster than Exim. Please, forgive
me my lack of enthusiasm.
It didn't take me long to Google up a similar comparsion whose conclusion
was the opposite of the one you found -- but it's utterly irrelevant; the
setup matters much more than the MTA itself.
I get paid to run a mail system using Exim, which does between 1.5M and
2M deliveries per day, so I do have enough first-hand experience with Exim
to see that your assertion that "[Exim] doesn't scale up to large mail
systems" is quite simply a false and ignorant one.
> yes, and postfix is simple to configure and simple to understand.
Probably very true.
> exim, however isn't.
I beg to differ -- and I'm certain there's many who agree with me. Your
impression isn't at all universal, you know.
--
Tore Anderson
Reply to: