Issue with Outlook and IMAP IDLE

[brong]

From: Sébastien Michel
Bugzilla-Id: 3396
Version: 2.5.x (next)
Owner: Bron Gondwana

[brong]

Attachment-Id: 1354
From: Sébastien Michel
Type: text/plain
File: outlook_workaround.patch

A light workaround for Outlook and IMAP IDLE

[brong]

From: Greg Banks

Hi Sébastien,

I just noticed this Bugzilla. I've recently reworked the IDLE code
not to use signals, which means your patch will no longer apply to
the master branch :(

I think it would be good to either workaround this in Cyrus or
definitely refuse to, so I have a few questions.

• is this still a problem for you? I assume so, given you raised
  the bug only 12 months ago.

• why have an imapd.conf option?

• what does the workaround do to power-sensitive mobile devices
  like smartphones?

• is it possible to detect a buggy OE in imapd and only apply the
  workaround in that case? Does it, for example, use the ID command?

[brong]

From: Sébastien Michel

1. and 4)
   I hope less than yesterday and more than tomorrow !
   We can't identify these outlook and outlook-express clients since there is no User-Agent header in IMAP. And those old clients don't support ID command...

I dislike endless workarounds and I hope to remove it in the future

IMAP IDLE is not supported by android (nor the latest ICS), nor in IOS 4.X and I don't expect either in 5.X.
it consumes too much battery power.

[brong]

From: Greg Banks

> 1) and 4)
> I hope less than yesterday

Ok, so we have to do something.

Sigh...I had hoped that by not working on code that runs in Web browsers I could
avoid having to work around Microsoft brain damage :(

> We can't identify these outlook and outlook-express clients since there is no
> User-Agent header in IMAP. And those old clients don't support ID command...

What about the "X-MSMail-Priority" and "X-MSOESRec" headers which are FETCHed
in your above examples? Are they fetched only by MS clients? I would assume so
judging by the name. Will such a client try to fetch those headers before it IDLEs?

> I dislike endless workarounds and I hope to remove it in the future

Sure, but that's not a reason for an imapd.conf option, that's a reason to add a
comment like "Please remove this hack when the last Outlook Express < 2007
client dies". Does a sysadmin ever actually need a Cyrus installation to not
behave in this new way?

> IMAP IDLE is not supported by android (nor the latest ICS), nor in IOS 4.X and
> I don't expect either in 5.X.
> it consumes too much battery power.

Fair enough.

[brong]

From: Greg Banks

Also, what would be the impact of either

• disabling the 30 minute server timeout, or

• adding an imapd.conf option to allow sysadmins to tune the timeout
  to some arbitrarily large value?

[brong]

From: Greg Banks

I may as well stop pretending I'm not looking at this.

[brong]

From: Sébastien Michel

(In reply to comment #3)
> > 1) and 4)
> > I hope less than yesterday
>
> Ok, so we have to do something.
>
> Sigh...I had hoped that by not working on code that runs in Web browsers I
> could
> avoid having to work around Microsoft brain damage :(
>
> > We can't identify these outlook and outlook-express clients since there is no
> > User-Agent header in IMAP. And those old clients don't support ID command...
>
> What about the "X-MSMail-Priority" and "X-MSOESRec" headers which are FETCHed
> in your above examples? Are they fetched only by MS clients? I would assume
> so
> judging by the name. Will such a client try to fetch those headers before it
> IDLEs?

Good question. I will check also with some newer versions of outlook in case these FETCH also these headers, that will be too bad because newer versions don't send ID command either ...

> > I dislike endless workarounds and I hope to remove it in the future
>
> Sure, but that's not a reason for an imapd.conf option, that's a reason to add
> a
> comment like "Please remove this hack when the last Outlook Express < 2007
> client dies".
OK

[brong]

From: Sébastien Michel

(In reply to comment #4)
> Also, what would be the impact of either
>
> * disabling the 30 minute server timeout, or
>
> * adding an imapd.conf option to allow sysadmins to tune the timeout
> to some arbitrarily large value?

The timeout will be reached on one network equipment like FW or LB.

Don't worry about this issue, it seems that not lot of people are complaining about that. I will update the patch later after you commit your refactoring.

[brong]

From: Adam Tauno Williams

Referencing "IMAP IDLE is not supported by android" (from comment 1). This is not entirely true. IMAP IDLE isn't supported or used by the Android 'default' non-gmail mail client, it isn't really "unsupported". The excellent K-9 IMAP/POP/SMTP client for Android does support IMAP IDLE and works with Cyrus IMAPd.

As for it consuming too much battery, I don't see that on my phone; K-9 consumes less power than many other applications according to the power meter. This is probably just another thinly veiled excuse for Google providing inferior support for non-proprietary services.

[brong]

From: Greg Banks

Sorry, I have to throw this one back in the pool :(