[Dovecot] Several issues with 1.0-stable

[Remy Zandwijk]

All,

We have Dovecot 1.0-stable deployed in our 3000 users environment as of
last weekend. Up untill now, I have 3 issues:

-	When stopping Dovecot with: kill -TERM `cat /var/run/dovecot/master.pid`,
	the 'dovecot' process itself dies, the 'imap-login' processes die, but the 
'imap'
	processes (there are a lot of them) don't die, but become a zombie. I 
think the
	way I stop Dovecot is the way to do it. The zombies can be killed by doing a
	'pkill -9 imap', but that's not quite elegant.

-	Dovecot is running in a Solaris 9 environment. It turned out that Dovecot 
crashes
	when more then 256 filedescriptors are being used. This is rather normal, 
since
	Solaris only support 256 descriptors per 32-bit process. After starting 
Dovecot, I
	do a 'plimit -n 8192, 8192 `cat /var/run/dovecot/master.pid`, which allows 
Dovecot
	to have 8192 filedescriptors. Timo, is it possible to introduce a config 
file option to
	configure the number of descriptors?

-	Pine and Outlook users don't see the adresses or names of the people who sent
	them email. In my email yesterday, I said I remembered that tests with 
Dovecot 0.99
	showed no problems. Today, I installed both Dovecot 0.99 and 1.0-stable on 
my test
	box. It turned out that 0.99 indeed works OK and that 1.0-stable is the 
problem. The
	user mailboxes are 'Maildir'; we do not use mbox format. I like to see 
this problem
	being fixed.....


We have 2 namespaces in our setup: the default namespace and a namespace which
location is set to ~/.Mailstorage. When not using a prefix in the email 
client setup, users
can make new email folders in the default namespace and the 'Mailstorage' 
namespace.
We want users only to create mail folders in the Mailstorage namespace, not 
in the default
namespace. Questions:

	- Is it possible to force a prefix to clients?
	- If not, is it usefull to have an option in the namespace configuration 
that allows or
	disallows the creation of mail folder? If so: consider this as a feature 
request.


Thank you for reading, best regards,

	Remy Zandwijk
	Free University of Amsterdam
	The Netherlands