Messages containing only the word "unsubscribe" should be filtered

[Ed K]

Users, particularly of Ubuntu, often accidentally subscribe to a bug, not knowing what they're getting into because apport encourages them to do so. Once they realize that they don't want the e-mails, they barrage the subscription list with hundreds of "unsubscribe" messages without realizing that their messages will have no effect other than to annoy other subscribers. For whatever reason, people continue to send the e-mails even after multiple e-mails have gone out about the correct way to unsubscribe from a bug.

These barrages will often force users who actually care about legitimate messages to unsubscribe from the bug, thus damaging the quality of feedback and interaction between legitimate bug testers and developers.

A proposal exists (https://bugs.launchpad.net/malone/+bug/48863) to give a direct unsubscribe link in an e-mail, however this is not enough. Messages which have only the word "unsubscribe", (and probably even messages containing one word) should be filtered.

[Michael Rooney (mrooney)]

This would be grand indeed. One issue is that a lot of these same people also typo unsubscribe so you'd really have to account for those for this to be useful. Also, sometimes people ask other things like "please remove". I think a bayesian filtering approach could go a long way here, picking up on the key words and the length of an email which determines someone wanting to be removed. These could then go a moderators list where the owner of the list either unsubscribes them or marks it as legitimate and the email continues on.

[Erick Brunzell (lbsolost)]

Agreed! Possibly limit messages to a minimum # of characters?

To me the proper way to unsubscribe is quite clear, but that barrage today was totally ridiculous!

[Martin Pool (mbp)]

I think this would be great too; I just saw another instance of it in bug 357001. (Though strangely enough the comment is not visible in the bug; maybe it was deleted as spam after the mail was sent?)

It is valid and not all that rare to send a one-word reply to a bug, especially for existing developers.

However, I can't think of any case where someone would set the mail subject to "unsubscribe" or "remove" and not want to be unsubscribed. It's probably common to have some text in the message.

[Martin Pool (mbp)]

So, re-reading the summary, I propose actually taking this a bit further: don't just block the message but unsubscribe the sender. If they're not subscribed (maybe because they're using a different address) at least reply with instructions.

Ideally this should be done in a way that also covers code review mail etc.

[William Grant (wgrant)]

On Thu, 2009-04-09 at 03:50 +0000, Martin Pool wrote:
> So, re-reading the summary, I propose actually taking this a bit
> further: don't just block the message but unsubscribe the sender. If
> they're not subscribed (maybe because they're using a different address)
> at least reply with instructions.

You can't unsubscribe them. That would be performing a potentially
dangerous action, unauthenticated.

[Martin Pool (mbp)]

2009/4/9 William Grant <email address hidden>:
> You can't unsubscribe them. That would be performing a potentially
> dangerous action, unauthenticated.

Maybe Launchpad should send a mail with a one-click URL to unsubscribe.

--
Martin <http://launchpad.net/~mbp/>

[Martin Pool (mbp)]

I think this is a dupe of bug 286193?

[Ed K (ekohlwey)]

Its not really a duplicate. This bug proposes filtering, the other bug proposes a reply-based unsubscribe mechanism. As William (above) mentions, these are really two different issues - one is to prevent e-mail with useless comments (particularly those requesting unsubscription) from bothering legitimate users, and one is actually to provide a mechanism such that unsubscription can be processed by e-mail. Also, as William notes, this is something that could be potentially dangerous (exposing changes in preferences/subscription settings to unauthenticated input), and should involve more planning that simply filtering readily identifiable spam.

[Martin Pool (mbp)]

Ed,

You're right they're not quite the same. I see them as different
phases of addressing the same issue.

I think just dropping the messages would be a poor step, because it
will likely just provoke people to send increasingly shrill messages
until they do bypass the filter. So I think it should at least start
by sending some kind of reply, even if it just has static text
describing how to unsubscribe. Later it can develop to give a
specific unsubscribe URL.

--
Martin <http://launchpad.net/~mbp/>

[Graham Binns (gmb)]

This bug should be fixed at the same time as bug 286193.