Hi-
A couple of comments regarding the POP3 server (I realize the POP3 server is probably low priority though..)
One problem is that the "LAST" command is not supported. Some mail clients end up downloading all messages every time because of this (or so I'm told). The biggest source of these remarks is people using the yahoo webmail front end. I wonder if this could be added?
Also, under BSD/OS, the "auth plain" mechanism doesn't work. Perusing the mailing list archives, I see that the symptoms appear just like those that were reported earlier for another BSD system. I'm curious as to whether there has been any progress on that front?
Yours, -mm-
On Wed, 2003-02-26 at 23:04, Mark E. Mallett wrote:
A couple of comments regarding the POP3 server (I realize the POP3 server is probably low priority though..)
One problem is that the "LAST" command is not supported. Some mail clients end up downloading all messages every time because of this (or so I'm told). The biggest source of these remarks is people using the yahoo webmail front end. I wonder if this could be added?
http://www.remote.org/jochen/mail/info/pop3-last.html
I could add support for it, but how exactly? Maybe return the previous message of the first unseen message?
Also, under BSD/OS, the "auth plain" mechanism doesn't work. Perusing the mailing list archives, I see that the symptoms appear just like those that were reported earlier for another BSD system. I'm curious as to whether there has been any progress on that front?
All BSD issues should have been fixed by now. I just tested with FreeBSD 5 and Linux, AUTH PLAIN worked just fine in both. Does USER+PASS work? Try enabling auth_verbose = yes?
And what errors exactly do you see now? There's been several different BSD errors :)
On Wed, 2003-02-26 at 23:26, Timo Sirainen wrote:
http://www.remote.org/jochen/mail/info/pop3-last.html
I could add support for it, but how exactly? Maybe return the previous message of the first unseen message?
No, RFC1460 defined it differently. I think I'll do it like UW ipop3d. \Seen flags are used to track the "last" message. If RSET command is used, it would clear the \Seen flag from all messages. So it doesn't interact too nicely with IMAP. Maybe this should be optional..
On Wed, Feb 26, 2003 at 11:26:48PM +0200, Timo Sirainen wrote:
On Wed, 2003-02-26 at 23:04, Mark E. Mallett wrote:
A couple of comments regarding the POP3 server (I realize the POP3 server is probably low priority though..)
One problem is that the "LAST" command is not supported. Some mail clients end up downloading all messages every time because of this (or so I'm told). The biggest source of these remarks is people using the yahoo webmail front end. I wonder if this could be added?
Yes, of course. Duh. I was just having a conversation a few hours ago with somebody about how the concept of "mailbox order" is not always a valid one. Sometimes I fail to connect the dots.
Also, under BSD/OS, the "auth plain" mechanism doesn't work. Perusing the mailing list archives, I see that the symptoms appear just like those that were reported earlier for another BSD system. I'm curious as to whether there has been any progress on that front?
All BSD issues should have been fixed by now. I just tested with FreeBSD 5 and Linux, AUTH PLAIN worked just fine in both. Does USER+PASS work? Try enabling auth_verbose = yes?
And what errors exactly do you see now? There's been several different BSD errors :)
I'll have to investigate further. Unfortunately the main symptom is that people call and say "it doesn't work" -- enabling auth_verbose may help. I had some indications that the dovecot-auth process died prematurely, but that could simply be operator error.
BTW for a completely random remark: you might get configure to check whether "lchown()" exists and use "chown()" if it does not (in safe-mkdir.c)
-mm-
On Wed, 2003-02-26 at 23:45, Mark E. Mallett wrote:
Also, under BSD/OS, the "auth plain" mechanism doesn't work. Perusing I had some indications that the dovecot-auth process died prematurely, but that could simply be operator error.
I noticed that too while testing. It crashes if client sends invalid data to auth plain. It shouldn't affect valid authentications though, except by aborting auth for other people who were just logging in. Well, I'll include the fixpatch.
BTW for a completely random remark: you might get configure to check whether "lchown()" exists and use "chown()" if it does not (in safe-mkdir.c)
Oh, didn't notice lchown() wasn't posix. I'll make it use fchown() and fchmod().
participants (2)
-
Mark E. Mallett
-
Timo Sirainen