guy keren wrote: > > On Fri, 4 Feb 2000, Omer Efraim wrote: > > yes, but the way the pop server works, it still performs these files > copies, no matter if the user eventually downloads their email or not (at > least qpopper used to make a file copy operation immediatly when the > user's mail client sent the user's password, and before returning a reply. > this caused clients to wrongly report of a 'wrong password' on timing out, > rather then reporting on a 'timeout' - that's what happened with pegasus > mail back in 95 or so... qmail-pop3d doesn't perform any file copying. Why does qpopper make a copy anyhow, in order to avoid corrupt mailfiles, in order to allow delivery while checking for mail, or...? > > you need to check out how to avoid any kind of file copying operations if > you want an fork() savings to make any sense... the POP protocol is > useless in making such kinds of checks, unless it gets extended in some > manner (e.g. by having a 'LAST_UPDATE command to notify the mail client > when was the user's incoming mailbox last modified. however, this won't > help much, unless _no_ file copy operations are performed if there is no > new mail waiting for the user, relative to the specific client checking > for mail - the user may be reading their email from different > computers...). > > now - how do you fix this up? without fixing this up - your "less forking" > system won't be any better then the current system... Like I said, I see no reason to make a copy of the file if I'm using Maildirs. And anyhow, most users have the mail removed from the server after it is downloaded, so the 'check-for-new-mail' operation is relatively painless. Just an empty LIST. > > > Well, I agree - but this is a whole other symphony and requires much more > > work, and it works on a much larger scale.I myself download mail tunneled > > through ssh so it's compressed anyhow, but I guess it would be nice to > > have clients supporting it out of the box. > > you're looking for large scale solutions - your ssh tunneling solution is > not large scale at all. I never said it is. Neither is it a good solution for the non technically savvy user. But then again, getting clients to support a new standard is a very difficult task. > > > Single file. The reasoning > > behind Maildirs was not really performance but rather robustness. > > this does not say it cannot be used for gaining in performance as well... I didn't say that either :) Obviously, you get some great benefits. I remember qpopper choking on a P120/64MB RAM when a user with a 20MB maildrop tried to retrieve mail (it throttled the whole machine for some time). > > guy > > "For world domination - press 1, > or dial 0, and please hold, for the creator." -- nob o. dy 1 0000111110000111 You system is defective, maybe get Comverse to work on it. -- /--------------- Omer Efraim, omere@tcmail.tau.ac.il ------------------\ [ Microsoft Vaccine 2000 is configuring your immune system. This may ] [ take a few minutes. If your body stops responding for a long time and ] [ there is no brain activity please die. Setup will continue after you ] [ are reborn. ] \-----------------------------------------------------------------------/ - Quoting Buzh, asr
S/MIME Cryptographic Signature