- dovecot - dovecot.org

[Dovecot] Master Password
by Marc Perkel 13 Apr '06

13 Apr '06

OK - I'm confused. Trying to use the master password feature. What do I need to do to generate the password? What do I do to make the /etc/dovecot.masterusers file? thanks in advance. When I get this to work I will update the Wiki so that someone like me can understand it. auth_master_user_separator=* passdb passwd-file { # Master users that can log in as anyone args = /etc/dovecot.masterusers master = yes #pass =yes }

1 0

[Dovecot] BAD Error in IMAP Command
by Adam L 13 Apr '06

13 Apr '06

Hello. I have dovecot .99.14, sendmail, squirrelmail installed. Squirrelmail logins in fine and works with the IMAP server. I set up Outlook to connect to the imap server and it hangs. I telneted to the IMAP port and receive the following: * OK dovecot ready. a1 login aalang a1 BAD Error in IMAP command received by server. BYE BYE BAD Error in IMAP command received by server. I try STATUS and same thing happens. I searched the Internet for a bit and I didn't find anything that seemed helpful. Any ideas?

2 1

[Dovecot] Dovecot-LDA & permissions
by Duncan Hutty 12 Apr '06

12 Apr '06

I set up postfix/dovecot with the dovecot-LDA according to the dovecot wiki, in particular: http://wiki.dovecot.org/LDA#head-dacb9b9a1f19c3ea86bb6f8caa3d64e3ddad9ef8 For separation of privileges, I want to have one uid per virtual domain rather than have the vmail user own all directories under the mailstore. Therefore the LDA, deliver, in order to deliver must fulfill one of the following: a) it must be started as root b) it must be setuid c) it must have write access to the user's mailstore even as an unprivileged user itself. But a) postfix declines to this: postfix/pipe[26267]: fatal: user= command-line attribute specifies root privileges which is probably a good thing. b) will work, but I'm really wary of making anything setuid c) could be done if the group that deliver runs as (as defined in the user= attribute in master.cf) is the same as the group of each mailstore directory and that this group has write permissions *and* the sticky bit of each mailstore directory is set so that the file created by deliver is owned by the uid of the mailstore owner not deliver's uid. This is not exactly a question as such because I believe, subject to a little testing, that I have a working mail setup. However, I believe in doing things the right way wherever possible and I'm sure someone must have come across this before, so I'm submitting to the list inviting comment. How are most of you doing this? Are you just using a single uid for all virtual users? -- Duncan Hutty System Administrator, ECE Carnegie Mellon University Please use informative subject lines

1 0

[Dovecot] password length
by Mark Richards 12 Apr '06

12 Apr '06

After setting up a very sweet Dovecot installation (with OpenLDAP) we're finding something strange: passwords are accepted but only the first 8 characters and anything less than 8 characters is an invalid password For several reasons, including allowing longer passwords to make them more difficult to crack, I'd rather be able to use longer passwords. LDAP seems to have attributes for minimum and maximum password lengths. Ours is set currently at 256 char maximum. So my guess is that Dovecot may be doing something, or there's a configuration item I missed. Checked all the docs and the wiki. Anyone? Mark Richards

1 0

[Dovecot] strip domains
by Netlink Tech 12 Apr '06

12 Apr '06

Hello, Running FC5 x86_64 with dovecot-1.0-0_8.beta3.rhfc5.at Using system user/passwds for 99.99% of users...have sql user/pass enabled to test virtual users from mysql. How do I strip off the domain when users provide user(a)domain.com so that my userdb and passdb use just 'user'? (disregarding sql auth...I'm not concerned with that right now). My dovecot.conf auth config: passdb pam { args = dovecot } passdb sql { args = /etc/dovecot-sql.conf } userdb passwd { } userdb sql { args = /etc/dovecot-sql.conf } Can I just pass userdb the 'user' using an args= statement? %d? ...and then also to the passdb? Thanks, Curt

6 13

[Dovecot] Master Login
by Marc Perkel 12 Apr '06

12 Apr '06

How do you make the master login work?

2 1

[Dovecot] Re: strip domains
by Roger Weeks 12 Apr '06

12 Apr '06

So if I understand this correctly, %u matches the whole username in the sql database, but %n only matches the name before @? One thing we discovered in migrating from courier-imap to Dovecot is that courier-imap supported user@domain as a valid login name, where by default using %u in the SQL query, Dovecot does not. So it would have been useful for us to avoid a number of tech support calls for users who have user@domain configured in their mail clients. -- Roger J. Weeks Systems & Network Administrator Mendocino Community Network On Apr 12, 2006, at 7:16 AM, dovecot-request(a)dovecot.org wrote: > Date: Wed, 12 Apr 2006 16:05:24 +0300 > From: Timo Sirainen <tss(a)iki.fi> > Subject: Re: [Dovecot] strip domains > To: Netlink Tech <tech(a)netlinkcom.com> > Cc: dovecot(a)dovecot.org > > On Tue, 2006-04-11 at 09:36 -0500, Netlink Tech wrote: > In general I don't see this as very useful feature. Why are your users > logging in with @domain? Or why not change your passdbs/userdbs to > just > contain the domain, and set auth_default_realm so everyone always logs > in with domains? > > > Here you can modify your SQL command to use %n instead of %u. > > >> userdb passwd { >> } >> > >> userdb sql { >> args = /etc/dovecot-sql.conf >> } >> > > And here again you can use %n. >

1 0

[Dovecot] Scheduled Outages
by Roger Weeks 12 Apr '06

12 Apr '06

What's "after hours" for an ISP with customers all over the world? That's why we have regular once-per-month scheduled outages. Same time, every month. Email reminders go out to customers and are announced on our web pages. People can't argue with a month's advance notice. -- Roger J. Weeks Systems & Network Administrator Mendocino Community Network On Apr 12, 2006, at 5:49 AM, dovecot-request(a)dovecot.org wrote: > Date: Wed, 12 Apr 2006 07:27:36 -0500 > From: Les Mikesell <lesmikesell(a)gmail.com> > Subject: Re: [Dovecot] 1.0.beta6 released > To: Reuben Farrelly <reuben-dovecot(a)reub.net> > Cc: Timo Sirainen <tss(a)iki.fi>, Chris Wakelin > <c.d.wakelin(a)reading.ac.uk>, Jens Laas <jens.laas(a)data.slu.se>, > dovecot(a)dovecot.org > Message-ID: <1144844856.24974.1.camel(a)les-home.futuresource.com> > Content-Type: text/plain > > On Wed, 2006-04-12 at 06:51, Reuben Farrelly wrote: > >> That's why there's "scheduled downtime" and "after hours" to do >> these things, in >> case stuff goes wrong, as some users of horde with dovecot have >> found out lately ;) > > What's 'after hours' for a company with users all around the world? > Interesting concept, though... > -- > Les Mikesell > lesmikesell(a)gmail.com

1 0

[Dovecot] http://wiki.dovecot.org/ IS NOT RUNNING
by Alexandre Busquets 12 Apr '06

12 Apr '06

3 2

[Dovecot] Full text search indexing
by Timo Sirainen 12 Apr '06

12 Apr '06

SEARCH is probably the slowest command that Dovecot has, especially when searching texts from message bodies. There exists a lot of text search engines, but unfortunately as far as I know there exists only one that can be used with IMAP. The reason why IMAP is special is because it requires that when searching eg. "ello", it must match also mails which contain eg. word "hello". Most search engines support only matching either complete words or matching from the beginning of the words. So, the one usable engine which I know of is Cyrus's squat indexing. It works basically by indexing the mails in 4 byte blocks and storing them in a trie. So if there exists a string "he llo", it indexes "he l", "e ll" and " llo" strings. When searching it then gets the first 4 bytes of the search keyword, looks them up from the squat index and then once it gets a list of matched mails, it reads the mails and does a proper search for only them. So the squat index doesn't give exact results, but only narrows down the number of mails which have to be searched. The biggest problem with those squat indexes is that it doesn't support incremental updates, so there has to be some nightly runs to rebuild the indexes from scratch. And with lots of users that's not practical, so I think that's the reason why the squat indexes aren't all that commonly used with Cyrus. I was thinking about taking the basic ideas from the Cyrus squat indexer but making it possible to update it incrementally so Dovecot could do it automatically. Also to reduce space I was thinking about sorting the 4 byte strings so instead of eg. "abcd", "dcba" and "abdc" all being indexed, they could all be stored as "abcd" (so the search key would also have to be sorted before use). But that might cause larger result sets making the search slower, so I'll have to test if it's worth it. Unicode characters might also be a problem. If the characters are stored in 16bit characters, it would waste space. If they're stored in UTF-8 the 4 bytes might not be enough to give a useful index (as a single character might take even 5 bytes). One thing I was thinking was to not index in 4 byte blocks, but 4 character blocks in UTF-8, so the trie could be 4..20 levels deep depending on what unicode characters are used (in finnish text it would probably be max. 7 levels deep). So, now would be a good time to tell if you can think of some better ideas how to do this, especially if you know about some new algorithms specifically designed for this kind of searching. :) Pretty much all the algorithms I've found about this were suffix trees/arrays and this basic trie thing.

6 14