Not to pick nits but pop3+leave on server does not mean you have all message from the dawn of time stored on the server. Outlook and presumably other MUAs have "remove from server after X time" and "remove from server when message is deleted" options when leaving pop3 messages on the server is enabled.
I agree with the general principal though that if you wish to keep messages on server as general practice you should use IMAP. IMO pop3+leave is really only ideal while traveling using other clients/computers or maybe if IMAP is not offered at all.
-----Original Message----- From: dovecot-bounces+jkrejci=usinternet.com@dovecot.org [mailto:dovecot-bounces+jkrejci=usinternet.com@dovecot.org] On Behalf Of Charles Marcus Sent: Tuesday, December 29, 2009 6:19 AM To: Dovecot Mailing List Subject: Re: [Dovecot] pop3+leave messages on server
On 2009-12-29, Papp Tamas (tompos@martos.bme.hu) wrote:
The protocoll imap is not the same as using pop3+leave messages on server.
That is correct. The POP protocol is designed to delete the messages from the server once they have been POPPED. The IMAP protocol is designed to leave the messages on the server all the time.
He wants to use this scenario anyway as he did for many years with no problems.
The fact that he has had 'no problems' in many years is purely the luck of the draw. One minor bug in either the IMAP/POP server or the mail client during an upgrade or other maintenance, and boom - he will *really* be surprised when *all* of his messages from the last few *years* are downloaded again, not just the last few hours worth.
One real worl example of how this can happen: his Outlook profile gets corrupted, and he still has all of his mail, but he has to recreate his account settings - boom, all of his mail downloads again from the date he started leaving them on the server.
Same thing goes for when he install Outlook on a new computer - which, in some cases, might be desirable, and in others, not...
The bottom line is, POP is simply not designed to work this way. The fact that it can be *manipulated* to work this way doesn't change the nature of the protocol, or the inherent problems.
--
Best regards,
Charles