> Nov 17 2010 3:32am from samjam @uncnsrd in Mail> to > [email protected] >Subject: Re: (no subject) > >On 16/11/10 22:29, dothebart wrote: >> >> Ok, most of the needed bugfixes were cherry-picked into stable-78x. >> >> this one i'd like to discuss first, 'caus its a little more intrusive >> then the other ones; its changes are primarily intended to get a SMTP >> client timeout: >> >> http://code.citadel.org/cgit.cgi/git.citadel.org/commit/?id=a1b7d9e895e7340116 1427363611b0160f51225f >> http://code.citadel.org/cgit.cgi/git.citadel.org/commit/?id=9148bf110ec104dd60 35c8f511d4c88afb276278 >> >> Client-sockets are in non-blocking mode now; we have timeouts after
>> which we abort the SMTP-sesison. >> >> This saves us from trapping into tarpits, or just connecting to >> SMTP-servers that won't send us a greeting within 5 seconds. >> > >I think 5 seconds is a bit low for normal slowness. Although most sites >respond more quickly, 5 seconds is still a short cutoff. > >Some sites have a latency of a few seconds, and if they need do do DNS >lookup as well over a congested link, 5 seconds will cause failure >needlessly. > >I'd favour at least 30 seconds. > >Sam > Actually, this is a value that we might want to allow configurability on... This would allow the system administrator to make a judgement call as to how long he/she feels is appropriate, and also put Citadel on par with other MTAs regarding configurability of this particular issue. - Stu
