Buchan Milne ha scritto:UW can't be tossed without a migration path. But maybe if someone can test uw->dovecot migration now that it's in contrib.
uw -> dovecot migration is transparent (but only if users
place their mailbox in "mail" or "Mail"); but this isn't
in latest uw-imapd (after removing of mailsubdir path). Nor
there is a way for getting it totally transparent, since dovecot use the $HOME/{mail,Mail}/.subscriptions while uw-imapd uses $HOME/.mailboxlist file.
well, for the .subscription/mailboxlist issue, name can be trivially changed in src/lib-storage/subscription-file/subscription-file.c, and i believe it is based on the root of the mailbox. so if we want to be compatible we could change the name to .mailboxlist provide a default_mail_env like mbox:%h/:INBOX=/var/spool/mail/%u:INDEX=%h/.imapidx put in a readme the steps that should be taken to change the mail directory and have the startup script for dovecot bitch until it is fixed.
Anyway it seems promising but I don't see it so better performing than imapd. It uses
indexes for mailboxes, and that's speed up, but on the other hand the INBOX is in /var/spool/mail/ and needs to be reindexed every time, since INBOXes are accessed
at the same time but the MTA (which could be postfix, sendmail, etc.).
well, mta fortunately only append data to the mailbox, so the index does not have to be rewritten from scrach. a totally different issue is users with shell access that manage their mailbox with mutt.
i agree with you, but would avoid a mbox vs. maildir holy war on cooker :)IMHO one of the cause of the claimed poor performance of uw-imapd mailbox based vs maildir(s) one is that end USERS don't understand
(this was magic IIRC)
Personally I my vote is for cyrus, but I know others prefer courier. (How about moving cyrus & courier into main?)
featurewise cyrus is much more complete than courier (mailbox acl, shared folders, filtering language, full text searches (if you have the disk space that is)) but it is an overkill for very small systems and it shares the same problem with courier, that there is not a clean migration path from uw. This is why i proposed dovecot, it should receive broad testing to see if it is a viable solution for replacing uw, but i don't see it easy for 9.2.
Why don't do a serious test with uw-imapd, cyrus-imap, courier-imap and dovecot, but with a certain amount of messages, i.e. more than
Well, I have migrated some hundred users (60GB of mail) from uw to cyrus same hardware, it got from unusable to good performing (i don't have numeric data for this, but user response was good) Is there an imap benchmark tool we could use to do these tests?
btw speaking of cyrus, what about putting the dirsync patch in the kernel?
regards, L.
--
Luca Berra -- [EMAIL PROTECTED]
Communication Media & Services S.r.l.
/"\
\ / ASCII RIBBON CAMPAIGN
X AGAINST HTML MAIL
/ \