See also https://phabricator.wikimedia.org/T58525 and
https://wikitech.wikimedia.org/wiki/Lists.wikimedia.org#Fighting_spam_in_mailman

On Tue, Aug 25, 2015 at 5:50 PM, Platonides <platonides@gmail.com> wrote:
> This may be an excellent corpus to analyse the kind of spam received by WMF
> lists and prepare some global rules.
> Generally mailman thas a great job filtering just by ensuring mails come
> from suscriptors, and I bet 98% of what's waiting is unwanted. Some ham will
> have been triggered by the "body too big" rule, and there's the occasional
> subscriptor that replies with the wrong account, but it's mostly content we
> are really not interested in.
>
> Some address found to have been spamming many of our lists, and not being
> subscribed to any (to rule out a cross-posting wikimedia) should be added to
> a global discard/reject list.
> In fact it is a pity that we don't keep a global remitent blacklist and
> several list admins need to review their crap, each one on its queue.
>
> We could simply have a milter that adds a "X-WM-sender-type: spammer" header
> (an opposite X-WM-sender-type: contributor could be used, too), and opting
> out would be as simple as removing the filter that discarded it by default.
>
> Yes, I consider such configurated would have to be provided by default. In
> fact, there's a SpamAssassin scoring emails that we could take better
> advantage of, but I'm not that in any has it configured. Again, it'd be
> interested to gather stats about the spaminess score given to such
> presumedly-spam corpus.
>
>
> In general, we are fighting list spam standalone, with very little
> coordination. Plus, I'm sure we can better use our tools.
>
> For instance just, automatically holding emails from:
> ^info@
> ^.*-owner@
> (I should probably have added noreply, too)
>
> did wonders for improving the cleanliness of certain list accepting
> non-members posts.
>
>
>
>
> Sidenote to list owners: Please delete any subscriptions of my email to this
> list using plus aliases. They work very well for organising what comes from
> mailman (mailman-owner addressed mail, retentions, bots crawling the
> address…) but interact badly with this list as I was multiple subscripted,
> since -when combined with gmail "feature" of delivering just a single email-
> emails from this list may randomly arrive on any of several list
> administrative folders.
> For instance, Daniel original email was filled as related to mediawiki-l.
> And I spent some time looking where Patrick email had gone before
> discovering that it wasn't in the archives :P
>
> _______________________________________________
> Listadmins mailing list
> Listadmins@lists.wikimedia.org
> https://lists.wikimedia.org/mailman/listinfo/listadmins

--
Tilman Bayer
Senior Analyst
Wikimedia Foundation
IRC (Freenode): HaeB

_______________________________________________
Listadmins mailing list
Listadmins@lists.wikimedia.org
https://lists.wikimedia.org/mailman/listinfo/listadmins