On Sun, Jan 10, 2010 at 01:56:23PM -0600, Stan Hoeppner wrote: > RobertHoltzman put forth on 1/10/2010 1:01 AM: > > One of the Alpine (ex)devs claims it's true. If I ever get the time I'll > > see about testing it one of the distros on my desktop box. Intuitively > > it sounds right as a search would entail opening and closing many files > > as opposed to one with mbox. > > I completely agree with this position. Technically it makes sense. This is one > reason I went with mbox on my Dovecot server. I'd just like to see some recent > modern benchmarks proving so and to what degree. My gut instinct says that mbox > is faster, but probably not to such an extent that it would really make a > difference from the "human latency" standpoint. I have a list mail file with > 10,600 messages in it. Piker. One of mine hs over 52k. > The longest simple body search time I've had through > T-Bird (server side search) is about 8-10 seconds wall clock time. If I'd > chosen maildir instead of mbox, and maildir took 16-20 seconds for the same > search, that's not a huge difference in human waiting terms--unless your daily > job entails searching mail files/folders all day long. This is on a lightly > loaded server. I'd like to see data for heavily loaded mbox and maildir servers. I would love to switch to maildir. I use clamav and it has the capability of quarantining a file which test positive for malware. With mbox this would mean quarantining an entire mailbox. Definitely not desirable. With maildir only the message in question would be effected. -- Bob Holtzman GPG key ID = 8D549279 If you think you're getting free lunch check the price of the beer.
Description: Digital signature