Multiple duplicate bug mails received when both subscribed to an external list which is the contact for a team subscribed to a bug and subscribed to the bug in some other fashion

[Scott Kitterman]

In many cases I'm both on a team that's a bug contact and signed up individually as bug contact on packages. As of this release I now get two mails in this case. This didn't used to happen.

[Graham Binns (gmb)]

Can you name specific bugs for which this is happening?

[Scott Kitterman (kitterman)]

There's one bug open for clamtk in Ubuntu. That's one.

[Graham Binns (gmb)]

On Thu, Jan 31, 2008 at 12:08:43AM -0000, Scott Kitterman wrote:
> There's one bug open for clamtk in Ubuntu. That's one.
>

Okay. I found two bugs open (one Confirmed, one New) for clamtk in
Ubuntu. I'm assuming that you're talking specifically about the New bug,
bug #113384, here, which is an interesting case because you're actually
subscribed to it three times, once as ~kitterman, once through
~ubuntu-clamav and once through ~ubuntu-bugs, of which you're an
indirect member.

I don't know whether that's something that's worth focussing on but I'll
keep investigating.

How many times have you seen this now, out of interest?

--
Graham

[Scott Kitterman (kitterman)]

It was this one:

https://bugs.launchpad.net/ubuntu/+source/clamtk/+bug/113384

that had a comment today. I've seen it a few other times since the latest
LP rollout. I don't get bugmail for ubuntu-bugs, so it's two for two.

This kind of regression is a significant user annoyance. I can't imagine
how this wasn't caught in testing.

[Björn Tillenius (bjornt)]

Please double check the To address in the bug notifications. One of them should be to your personal address, and the other one should be to the team's address (both are @kitterman.com addresses). Is this right? If not, please send me the two notifications with full headers.

[Scott Kitterman (kitterman)]

Yes. That's right.

Received: from adelie.canonical.com (adelie.canonical.com [91.189.90.139])
 by mailwash7.pair.com (Postfix) with ESMTP id 7F0CEAE86
 for <email address hidden>; Fri, 1 Feb 2008 08:26:57 -0500 (EST)
Received: from forster.canonical.com ([91.189.90.190])
 by adelie.canonical.com with esmtp (Exim 4.60 #1 (Debian))
 id 1JKvuz-0001Cz-MG
 for <email address hidden>; Fri, 01 Feb 2008 13:26:57 +0000
Received: from localhost ([127.0.0.1] helo=forster.canonical.com)
 by forster.canonical.com with esmtp (Exim 4.60 #1 (Debian))
 id 1JKvoc-0003tO-09
 for <email address hidden>; Fri, 01 Feb 2008 13:20:22 +0000

Received: from adelie.canonical.com (adelie.canonical.com [91.189.90.139])
 by mailwash7.pair.com (Postfix) with ESMTP id B2697AEB3
 for <email address hidden>; Fri, 1 Feb 2008 08:26:57 -0500 (EST)
Received: from forster.canonical.com ([91.189.90.190])
 by adelie.canonical.com with esmtp (Exim 4.60 #1 (Debian))
 id 1JKvv0-0001Cz-3Z
 for <email address hidden>; Fri, 01 Feb 2008 13:26:58 +0000
Received: from localhost ([127.0.0.1] helo=forster.canonical.com)
 by forster.canonical.com with esmtp (Exim 4.60 #1 (Debian))
 id 1JKvoa-0003tC-TF
 for <email address hidden>; Fri, 01 Feb 2008 13:20:21 +0000

[Christian Reis (kiko)]

I don't see how we could avoid sending you the team mail. Or are you proposing we don't email anyone who is a member of a team already being notified?

I think this bug is a dupe, btw.

[Scott Kitterman (kitterman)]

I don't care which I don't get. I just don't want two.

[Björn Tillenius (bjornt)]

On Sat, Feb 02, 2008 at 02:01:41AM -0000, Scott Kitterman wrote:
> I don't care which I don't get. I just don't want two.

There is no way we can know that you will receive mail at both those
addresses. For example, you are also an indirect member if Ubuntu Bugs,
which get notified about all bugs in Ubuntu. I assume that you don't get
those notifications, since you said you only got two. But that means
that one of the teams you're a member of you're getting e-mail, and one
you don't. You see my point?

If you remove the contact address from the team, you won't get notified
twice.

[Christian Reis (kiko)]

If the team is managed in a Launchpad mailing list, I think we actually get this right, though. So maybe that's an easy way of solving Scott's problem.

One additional thing we could do, Bjorn, is exclude a subscriber's direct notification if he is already a member of a team getting notified. That wouldn't be very easy to do with the current code but would be possible. More generally, what we'd do is take into account team participation when uniqing the emails. Reason I don't know that's a good idea is there's a problem: people might be expecting to receive the direct email (for instance, if they are members of the team but not subscribed to the team's mailing list) will stop.

[Scott Kitterman (kitterman)]

Yes. As an example with postfix and ubuntu-server I only get one mail.

Received: from adelie.canonical.com (adelie.canonical.com [91.189.90.139])
 by mailwash7.pair.com (Postfix) with ESMTP id 0255BAE46
 for <email address hidden>; Sat, 2 Feb 2008 14:20:53 -0500 (EST)
Received: from forster.canonical.com ([91.189.90.190])
 by adelie.canonical.com with esmtp (Exim 4.60 #1 (Debian))
 id 1JLNv4-0001o2-7M
 for <email address hidden>; Sat, 02 Feb 2008 19:20:54 +0000
Received: from localhost ([127.0.0.1] helo=forster.canonical.com)
 by forster.canonical.com with esmtp (Exim 4.60 #1 (Debian))
 id 1JLNv3-0006IA-Ru
 for <email address hidden>; Sat, 02 Feb 2008 19:20:53 +0000

So I was suprised at the inconsistent behavior.

[Björn Tillenius (bjornt)]

On Sat, Feb 02, 2008 at 07:03:58PM -0000, Christian Reis wrote:
> If the team is managed in a Launchpad mailing list, I think we actually
> get this right, though. So maybe that's an easy way of solving Scott's
> problem.

I'm not sure we are, and I'm also not sure we are on an agreement what
'getting it right' means.

> One additional thing we could do, Bjorn, is exclude a subscriber's
> direct notification if he is already a member of a team getting
> notified. That wouldn't be very easy to do with the current code but
> would be possible. More generally, what we'd do is take into account
> team participation when uniqing the emails. Reason I don't know that's a
> good idea is there's a problem: people might be expecting to receive the
> direct email (for instance, if they are members of the team but not
> subscribed to the team's mailing list) will stop.

I'd say we can't do that. What we could do when we have a LP hosted
mailing list, is to only send a direct notification, and not send one
via the mailing list to that. If we wouldn't send a direct notification
at all, we'd break the workflow for many people. I, for one, subscribe
to quite a lot of Launchpad bugs, even though I'm subscribed to the
launchpad-bugs mailing list. If I wouldn't get a notification sent
directly to me, subscribing to a bugs would be useless for me, and it
would be harder for me to manage my bugs. For example, I probably
wouldn't participate in this discussion if I would get the
notifications via launchpad-bugs.

BTW, there is no right or wrong here. Some people prefer getting two
notifications, some don't. The only way we could solve this, keeping
everyone happy, is to add a preference, so the user can choose for
himself.

[Matthew Paul Thomas (mpt)]

Adding an option wouldn't keep everyone happy. It would make happy only those who found the option, understood it, and made the effort to use it.

This bug seems like it should be fixed by structural subscriptions replacing bug contacts. Having a single bug contact forces the creation of dummy teams, which then causes duplicate notifications to members of various teams. With structural subscriptions, a person could be indirectly subscribed, directly subscribed (trumping any indirect subscription), or neither. Apart from MOTU-like processes (bug 179857), it would no longer ever be necessary for a team to be subscribed to a bug report.

[Björn Tillenius (bjornt)]

I'm marking this as Triaged, but we'll only be able to fix this for mailing lists that are managed by Launchpad.

[Robert Collins (lifeless)]

This is already fixed for lists that are managed by Launchpad AIUI. We need an interrogative protocol for external lists to (e.g. - to sketch one, a link in the mail you can click on if you get it twice, set on mails to non-lp-managed team contact addresses) to solve it for other sorts of mail-expanders.

Alternatively, perhaps we could lie slightly and set the same msg ID for all our messages even though they are different; that way if you're subscribed to two lists and your mail client does msg-de-duping (which many providers do) you'd only see one.