Re: [Logicmail-devel] IDLE not supported by IMAP server

[bartek b. <maj...@gm...>]

2010/6/2 Derek Konigsberg <dko...@lo...>

>  Ahh, this one:
> http://www.logicprobe.org/proj/logicmail/ticket/181
>
> Looks like the first thing I'll do is get bincimap up and running on my
> little test server VM (its an OpenBSD install with Cyrus, Dovecot, and
> Courier).
>

I could help here if you run into difficulties.


> Getting it to work actually won't be too difficult on the coding side.  The
> real question is how to change the behavior of the mail connection handler
> to cope with a lack of IDLE support.  I'm pretty sure one of the desktop
> e-mail clients I use (I think Kmail) doesn't use IDLE, so I'll have to
> observe that one again in a packet sniffer to see how it operates.
>
> The overall behavior of the connection handling code is described in a
> state machine diagram near the bottom of this web page:
> http://www.logicprobe.org/proj/logicmail/wiki/design-mail
>
> Hopefully all that will have to change is the specifics of how the IDLE
> state is handled.  Right now, it does something like this:
> - If no pending requests, go into IDLE
> - If 5 minutes have elapsed without any new requests, exit the idle state,
> do a status check, then resume IDLE (I think this is also how Thunderbird
> works)
> - Code is in:
> org.logicprobe.LogicMail.mail.IncomingMailConnectionHandler.handleBeginIdle()
>
>
To handle a non-IDLE-supporting server, I'll just have to implement a
> similar timeout/poll loop without the IDLE state.  This will obviously make
> it less responsive to new messages, but should still work.  (and will have
> to do it for IMAP-only, since POP effectively locks the mailbox against
> changes while you're connected)
>
>
Yes, I made it to this point that this is where IDLE logic takes place. What
you say doesn't sound too difficult, if I understand correctly, just issuing
certain command at certain interval. And probably also changing the
implementation of ImapClient.hasIdle() to actually check imap server
capabilities. I agree on the responsiveness issue, probably the polling
shouldn't also happen to often to take battery life into account.

>From what you write I can see that you would like to implement this by
yourself, anyway I'd be glad to help develop this great product:).

Kind regards,
--
Bartosz