 Welcome everyone to this talk by Alexander Wirth. He will give us an update about the Debian list servers and how to distribute all the messages to 100,000 people. Give him a round of applause, please. Yeah, hello. It's nice to be here. In fact, this is my first talk in English, so I hope that it works. So this is a famous list master talk. We want to tell you some things about how we are doing lists, what we did during Debcon and what we plan for the future future future. Who am I? My name is Alexander Wirth. I'm a Debian developer since 2003. I'm a list master. I'm an area of admin. I do maintain a few packages, one about 40 packages like IP root, a single, and some other stuff. If I don't do open source, I got paid by creditiff, where I'm a senior consultant, and I do help other people to implement open source solutions. But back to list master. We are a team. We have some members. At least two of them are here in the auditorium. That's Martin Zubel-Hillers, called Berman, Don Armstrong, who is also doing bug Steven Ork. Yeah, that's me. And the not so active ones are Martin Joey Schulze and Pascal Hakim. The two haven't been seen for a long time. So there's a reason they are an italic. What are we doing? List Steven Ork is run by Smart Mail and Post Fix. For everyone who is not familiar with Smart Mail, Smart Mail is a list manager from the last century based on Proc Mail. So it's basically just a gigantic set of Proc Mail rules. I would say a few thousand rules. Sure, it's not just, of course. We only see them as Proc Mail. We also do bug subscriber management. So if you subscribe to a bug, this is also going to list Steven Ork and is managed by Enemies of Carlotta, which is also a project to get replaced in the future because Enemies of Carlotta is out of maintenance for some time now and in fact it's never finished. So we are dealing with a bunch of bugs in it. We also do compliant and abuse management. There are always people that are surprised that posting to a public mailing list gets you on a web-visible archive and so we usually have to deal with such people. There are also people that don't behave like they should behave on a mailing list. There's also this last part about childcare, especially in the last year with the system deflame wars. Life wasn't that nice on lists. I don't know if anybody of you reads Debian User or Debian User German. We had to do things to get big to normal like banning people, warning people and so on. So that's also part of the things we are doing. We also maintain the list archive. So if you want to read older mails from the lists on the web, this is done by our list archive which is run by Monarch, which is also a software of the last century. But there aren't any good replacements yet. Some numbers. We are dealing with mails, a lot of mails. In one month, we distribute about 34,000 mails to our users where we have 100,000 unique subscribers. This makes up delivering 16 million mails a month, which is not that bad. On the other side, we do reject 390,000 mails during SMTP time. We have a bunch of advanced post-fix rules that try to get out the users' bammers. If mails make it through our SMTP server, they want to spam assess and amavis, which removes another 51,000 mails. And we also have progma rules which remove spam and other broken mails from another 65,000 mails. That means we're protecting you from more than half a million mails a month just with our spam rules. That is, for such a lot of lists and a lot of languages, a really good number, I think. On the other side, we as listmasters get around about 15,000 mails a month, which we deal with, like complaints, abuse, questions, bugs, and so on. Yeah, misguided subscriber and unsubscriber requests. It's surprising how hard it is to unsubscribe from a list. We had some listmaster sprints during the StabConf, where we implemented, fixed, and did other things with lists. We also have some numbers and stuff. We managed it to close to 100 bucks on our list-debin-org package, which is a really good number. We updated list-debin-org Alias Bendl to Jesse. We updated Monarch, which is also a big task because our Monarch is heavily patched. We enhanced our list pages with statistics and a direct inline search. This goes in hand with our improvement of the list search. List search was broken for one and a half year now. Thankfully nobody cared about it, but, however, it works again, and you are now able to do much more improved searching and mailing lists now, so you can search explicitly for specific lists, specific dates, and other stuff. We have to give a big thank to Odi for this, which was one of the Sabian upstream developers who really helped us in improving that. In fact, he did all the work. We also improved our Dikim handling. I guess most people aren't familiar with the subject Dikim. It's derived from Yahoo domain keys, and it's a method to sign headers in the email with a public key, which enables receiving mailers to verify that this mail is really coming from Yahoo or Google or something like that. The problem with Dikim is, in the past, we added footers under our mails with unsubscription links with archive links and such and stuff, and this modifies the mail. This has a consequence that if the mail was signed before, the signature is invalid afterwards, and if we distribute the mail to a mailer that honors Dikim, it will reject the mail because it's not a worded mail from Google anymore or from Yahoo, and that leads to problems. Therefore, we removed every footer in general, every body modification we had before in this Steven arc. Some of you may already noticed that the footer is gone. In days with MIME, most people haven't seen the footer most of the time because it's just lost somewhere in the MIME tree. We moved all those information into the header, so if you are looking for an archive link or an unsubscription link, look into the header of a mail. We are also following the relevant AFC at that point, so there are also some mailers that present you buttons for unsubscribing or list joining. There's an extension for Thunderbird, for example. We also created some new lists and closed some complicated lists. What's there on our list? Unfortunately, time is limited and we weren't able to do everything we wanted to do. So, what's left? In the future, we want to also honor Die Kim at our SMTV server, which means if Google says, no, Yahoo! Yahoo is the only one that does it. They said, if you receive a mail from Yahoo.com, which has an invalid Die Kim signature, reject the mail. It is commonly known as P is reject. That's coded into some DNS text entry. We want to do this in the future. We have a spam review system on that later, that we want to integrate with some single sign-on, especially we want to get all of users into the spam review system. We had a Google Summer of Code project in the last year, which produced some maybe Monarch replacement called PLIST, which we have to evaluate to see if it fits our needs in every case or to improve it and maybe get replaced Monarch. Some of our mailing lists do signature checking on receiving mails, especially for those announce mailing lists like Debbie and Devil Announce, Deepin Security. They do GPG checks on incoming mails and our check is not very sophisticated and is always failing on some of my mails that need to get improved because inline signatures for GPG is not really the method you are using today. There's this debconf.org. Two years ago, the debconf team decided we wanted to move this debconf.org into this deepin.org infrastructure. That should be done. There are some political reasons that isn't done yet, but we hope to get this done maybe up to Cape Town. Yes, it's not the listmaster problem, it's a political problem inside the debconf organization. Some of our mailing lists are moderated. The methods, how this moderation works is very special. We should really get something better like a web front or even a good mail front and at this point be it to improve high-volume mailing lists like Deepin User with some moderation. That is currently not possible yet. In the last year, I have written some band manager to cope up with those bands to implement things like timed bands and commands and a web overview of who is banned, why he is banned, and so on. We need to integrate that at 2-0. It comes to my next topic. We need help. We need a lot of help. If you like mail postfix, maybe even ExSim, there are better things to learn. If you are able to deal with strange old software like smartlist, prog mail, shale scripts and if you are interested in being a listmaster, please get in touch with us. We really need help. If you are not able to answer all those questions with yes, there are some isolated tasks you maybe help us with. That is part of the zig-zag tool, as I said, that is currently written in Python, which is not the language which is spoken from the active listmasters. So, we need help. If I improve it or revive it in Python, that is also fine. Back to the P list I talked about. It is template driven and the templates are currently very basic. We would need some web guys who are willing to do this with some HTML and some CSS to get a proper looking web archive ready for the year 2015, which is not the case with the monoc setup. Yeah, this is the other one, the spam review for it and it should work with an Aliet front-end. At the current point, we only allow dvn-developers to do the spam review and we want to get more help there. Just a little bit to the techniques. On every list archive page, there is a button where you can report spam. And if the relevant number of people report a mail as spam, it gets into a review point and we urge active dds or later add-of-users to review such reports to decide if this is really spam. And if it's really spam, we remove it from the archive. That's the point where we can do everybody who is able to read a mail. And if you are able to read Korean mail, Japanese mail or something like that, you get bonus points because that are things that are not that far distributed. That's basically from my side. Now, do you have any questions or feedback to the Listmaster team? Tell us something. I have a question about more statistics. How many people are banned and how many ban requests do you get per month or whatever? That's hard to say. Not everyone is banned on every list and not everyone is banned with every subject. There are some more or less complicated programming rules to decide who is banned. I would say at the current point maybe 40 or 50 people at Zoom. Yeah, there are some spammers and it's hard to differentiate between those mail addresses we block for spamming or we block for other reasons. Let's say just for behavior, we have maybe 10 to 20 people. You mentioned the Kim for signatures for the domains and I'm interested if you want to check other mails arriving into the system and if they want us to reject the system we reject or you want to tell others if they are without signature from the Debian domain, it should be rejected. No, we won't sign outgoing mails because that gets us into other problems. I talked to some dekem experts they told me that would lead to dekem stapling as being mails signed for several signatures which is in Siri possible but no one is doing it yet and they are not sure what will happen. So we decided on not signing our mails outgoing. Any other questions? We still have a minute or two. I can only support Alex on that we really, really, really need help in the Listmaster team because I need to admit that I'm more or less inactive and just jumping in if help is really urgently needed otherwise I'm mostly busy with my DSA stuff. Anyone who has spare time and likes to read Listmaster mail come speak with us and see what you can do. How much spare time are we talking about? It would be possible if we get a full-time drop out of it but if you have spare time we will be happy to use it. So if you were able to invest maybe half an hour a day or an hour a day or every second day that would be great. Reading such a lot of mails is usually a really time but if you have not that much time we will still be able to use it. Okay. Any other final question? No? Thank you, Alexander. Thank you again.