 the current and future challenges Debbie and Edo has. Again, like yesterday, I didn't have much time to prepare this talk, which is mostly, I'll explain later why this happened here. This is also the same, and I just say what I do in Debbie and Edo. I started as a contributing member, developer. Then we discussed setting up an archive, proper archive suite, and I was one of the people proposing duck, the Debbie and Archive Suite. I became FTP master in Debbie and Edo, and then I started managing the edge release, so I'm also a release manager in Debbie and Edo, so that's no formality title. And again, Debbie and community still needs help. If somebody is interested, please talk to me. The original idea of Debbie and Edo was to provide a whole distribution with a complete platform for schools, so that the schools can set up their systems with less work and therefore costs, maintenance costs, give schools a collection of services which is pre-configured and reduces the administration needs, and also to increase the computer availability. Those are practical goals, or ethical goals. I think you know them, you mostly share them. It's teaching kids that sharing software can be legal and good. How to use the software and to improve the software, to show the quality of free software, and also to provide the environment in their native languages. By now, Debbie and Edo is, or by the end of this year, Debbie and Edo will be used in all schools in Extremadura, in many, many schools in Norway and Germany, more in Norway still than Germany. Norway, almost more than 50% of the schools, Germany, I have no numbers, but also way more than 100 schools, I think. And it's used in all the world. Yesterday night I learned it's also used in Argentina, in Madela Plata, which made me happy, I didn't know that before. And it's also used in universities and offices. The schools have some specific needs, which I summarized here, it's low maintenance out of the box setup, because mostly the administration is done by some teachers who do it as part of either their teaching, or unpaid in the afternoon. Schools also often have little resources, old machines. Localization is important for schools, and there needs to be applications for kids to express themselves, to share what they do, and also for making multimedia or art or other stuff. For older pupils, the source code availability is also an interesting or good feature because they can change it and learn how software works. And schools also need office, Java, Flash and multimedia support more and more, because many tests are now written either in Flash or in Java, so the browser needs to support that so that the exams can be done with the software. And this is all not so really school specific at all. Most stuff is also needed in offices or universities or I need to have the same needs. Debbie and Edward Terra was the code name which we released and when it was actually released and Easter, so it was released in July 2007. We had a simplified Debbie and Installer where we trimmed down the number of questions even further. We had, we supported three architectures, I386, AMD64 and PowerPC. The default desktop was KDE, and we also provided the Kiosk mode pre-configured so they were different if the users belonged to the students group, there's certain applications were missing and the same for the teacher group. And we also built live CDs with Debbie in life. We have a network architecture so that when the whole network is installed, the person doing it doesn't have to make up some IP addresses or it's always the same that we have a router with a fixed IP address which is not part of Debbie and Edu. So we say get some router or usually the school has one. The main server always has this 10.022 IP address always and it has DNS, DHCP, Syslog, Mail, Proxy, LDAP pre-configured and holds the home directories. And we have workstation and thin client server on the same network and thin clients on other networks behind that. And if you can choose the type you want to install, you can either say this is a main server installation and it sets up the main server with this IP address or if it's a workstation, it sets up also this IP address for the workstation and a specific set of applications. And if it's a thin client installation, thin clients, they get automatically are installed in a different subnet. Yeah, this is what I said. The server has all these services listed and the nice thing is that all works out of the box so which really makes the installation a lot faster. It's not, for people who have done it several times, it's easy to configure DNS and DHCP but if you do it the first time, it's nice if it just works. And we also have this web browser based admin tool, Elvart, which is used to add users and machines to the system. So there's not so much work on the command line. There's still a bit left because we cannot get the DHCP server and the DNS server configuration from Elvart so you need to use command line tools for that. But for adding machines to the net groups and adding users, you can use this web browser tool that we used to use Webman up to Woody but since Edge Webman is not part of Debian anymore, so we came up with Elvart, which only has a subset of the Webman features and this is the most area of complaints at the moment that teachers want to configure everything with a web browser and we don't have a good solution. Not yet, also not found in Lenny yet. Microphone is coming. Have you tried GOSA? We tried GOSA but the problem with GOSA is that it stores the password so that the users can access it. GOSA requires a server which has no users on the server and that doesn't conform with our architecture. So for Lenny plus one, I hope that we have several tools coming up that we find something, but for Lenny it looks like we will use Elvart again. We also did some other changes to Debian default desktop install, like per default install with use flash so that it's a nicer boot process. We had dash as the default bash in Edge release already. ETC is automatically kept in SVK so all changes are automatically committed to the version control system so that you can go back in time and see the changes. When we partition the hard drives, we only use, we need 15 gigabyte I think is the minimum install with the partitioning layout we have, but even so if there's a 200 gigabyte hard drive, we only use this 15% and then we have different AVN volumes and they get automatically resized according to what is used. And since Edge, we also have this Debian edu doc package, which I will explain in a few slides later, which is a documentation written in the wiki so that it's easy for users to contribute to it. Yeah, it's written on wiki.debian.org. Then I review the commit mails, see that the style of the documentation is kept really as a manual which fits for Edge. Sometimes people commit stuff, how to make it work in Ubuntu or with packages from testing and this stuff I remove usually because it's the Edge manual. So I try to keep this documentation focused on Edge and now the new one on Lenny. Then we use Moin Moins doc book export to create XML from the wiki, use PO for all to create PO files and then translators translated. That worked really well. The French translation came up as a wishlist back to the package. A month ago it had the 100% translated French translation, which was really nice to get. The documentation is 100 pages, I think, as PDF. And then we create HTML and PDF versions from the translations and put it into the Debian edu doc package. So it's installed on this system and we also install it on a web browser, on the internet so that people can read it there. And since October 2007, I'm maintaining it and upload it once a month with new translations mostly. Yeah, Moin Moins doc book export is pretty broken. Somebody wants to work on that, that would rock. The Debian edu documentation with 100 PDF pages creates 2000 XML errors. We have another document written in the wiki in Norwegian which creates over 10,000 errors on 400 pages. There is now the newer Moin Moin version which Franklin Piat has set up. I have no time to test it, but it still creates errors but less. In December 2007, we made a point release. Fixed 40 bucks in it from wishlist to critical, all kinds of bucks. And updated 18 packages after reviewing all the changes and created new CD and DVD images which were then extensively tested. And many people choose to just install this and wait to skip the RCR release because they didn't have the need to upgrade immediately. There were no school holidays and they just went with this. Edge is now in complete bug fix mode. Except for the manual, there's not much work on it done because Edge really works nicely. And yeah, so this is the overview of packages. You read that, yes. Which were not in Debian Edge main. In that word, it's the next thing. Oh yeah, this, yeah, of course. This packages were not included in Edge. The Debian Edo packages were not part of Edge because of this bug, I will explain later which we now downgraded so the packages are part in Lenny. And Flash was not in main. We put it in our main because we don't only have main. It's called local in our repository. And the other stuff is Elvat and the LTSPA file system so that the SYN clients can access USB sticks. And we have more packages which were modified from Edge because we wanted to get bug fixes in which were important to our users but were not important enough for the release team to include them in Edge. And so in total, it's 25 packages and that's this nice percentage. So why do we want to be part of 100% part of Debian? Because it's less work and gives better results. Less work for upgrades and also for the archive maintenance. At the moment, there are three people maintaining the archive and reviewing newer packages. And we could just spend our time directly in Debian and not do this small fork. I also, personally, I see Debian as a test bed to do changes in Debian. It was very easy for us to introduce use flash because it was just a decision between four people. And Debian is more like, I want to see the kernel messages scrolling, other people just don't care. So Debian Edo is good to make changes but Debian Edo is an official sub-project from Debian. We are a custom Debian distribution. We had 25 developer gatherings, mostly in Europe, to get us no better and to discuss things easier, faster. We had 30 Debian Edo developers working on, directly working on Debian Edo packages in the last three years. Though the current core is five to eight people and there have been 115 translators and documentation writers since 2001. And of course, we also have 1000 Debian developers which do most of the work for us. So in edge, there were nine packages which were not in Debian and 16 we modified. For Lenny, we had the goal to have zero packages differ. And I will explain now how the current status, the reality is. But first, the contributions to Debian. There are several people who are now Debian developers who started as Debian Edo developers. Learned packaging in Debian Edo and then went through NM. We participate in the CDD effort. Debian Oscola Linux Foundation, money was used to fund the kickoff of Debian installer and testing security and free Java and Gnash meetings mostly. Mostly the meetings for Debian installer and testing security, also people were paid to work on that. And we developed the preceding for DI and the LTSP packages as they are in Debian now are also the result of Debian Edo work. LTSP is the Linux then, what is it, Terminal Server Project, which is the SYN client software now in use. The development model is the same as Debian. We try to have everything in version control system. We use a bug tracking system. We try to work with upstream, follow licenses, nothing new here. We use a mailing list on list Debian org. We have several other lists scattered for local support on other list servers which has historical reasons mostly that people set up the list and people get subscribed there and we never bothered to bring them over. We extensively use Vicky Debian org slash Debian Edo. We use an IRC channel and we have a alias project. We also have the own archive where we have five distributions. We use Sarge, Edge, Edge Test, Leni and Leni Test. The test distributions are what is set in Debian. At the moment we upload to Leni Test, use this and then create CDs for Leni Test. So we can get packages much faster in our Leni because we can directly upload there and test the changes faster. Edge Test and Leni Test are also auto built for the two other architectures we have or we also auto build I386. And packages then need to be manually removed from the test distributions to the non-test distributions and Edge and Sarge are untouchable because long released. We also give upload rights to non-DDs within our policies which are largely the same as Debian and on SVN commits new CDs are built if there are changes for the CDs and DVDs. So sometimes at the moment we build 15 CDs per day because there are some changes. For Leni, the Linux distribution started to deploy Debian Edu based already this summer. They based it on Leni. They froze Leni themselves and we'll have to deal with the upgrade pains. We also have point releases. I'll come to that later. And we have the numbering issue which I really have no idea how to solve except we skipped one release because we called our Edge release 3.0 and Debian was 4.0 and I think we should just fix it as fast as possible and just make Debian Edu 5 with the next release. These slides I reused them and for the reasons explained later there are some partly outdated. For technically we wanted to put DNS and DHCP configuration into LDAP so that we can edit that with the web browser tool. We want to apply our configuration changes in the policy complained compliant way. This is this 311188 bug which I have three more slides to. One complaint we had is that this fixed network setup is very nice for people who don't have a network but for people who have an existing network often the IP addresses are wrong and it's not so easy to change it and to make that easier. Then the workstations get the users from LDAP which works nicely and get the home directory. The thing with laptop is you take them away from the LDAP server and then the home directory is gone and that was the fix we still haven't done. And the other challenge due to Linux is that we now want to support the GNOME desktop as we support KDE. We got that. Then our new features, we have a GNOME desktop now. We also have a SUGA desktop. SUGA is the desktop environment from one laptop per child project. The GNOME desktop is really polished as the KDE one. The SUGA is not working so well but it's there as the base to work on it next year. SUGA is also not available as default. We used to have four installation options for this main server, workstation, thin client server and standalone and now there is one can select whether GNOME or KDE. That's the question we added. We configured DHCP from LDAP, DNS not yet. That's new in the Debian DHCP server package that can take the configuration from LDAP. We automatically set up a PXE putting environment for the workstations and the thin client so they don't have to be installed manually with the I anymore but just can boot from the network and we fixed bits of the hard-coded IP setup. So there's now less work to change the network but it's still not all in one place. We also made some progress on this bug but we haven't fixed it yet. There are too many packages involved. I'm not sure if we decided to use Ganesh as the default Flash player, I just didn't follow that. We have free Java finally. This killer feature is a feature teachers really wanted that automatically idle users get killed after some time and so we have a killer feature now. And of course we have everything else which is new in Lenny. That's our current diff to Lenny. That's four packages which are newer in Debian Edu than in Lenny and one package which is not in Lenny main. Host policy group is a tool so that we can say, okay we put these host into these group and then disable network access for them in the firewall which is used for exam mode when students should not access the internet. But as there are, I don't think it's really needed for Lenny release as there are other changes. As it's not always a use case to be able to disable the internet. Also it's not the only change you need to, if you want to make an exam, you also want to have new users for the exam otherwise the students can just put some documentation they shouldn't have in the exam in their home directory. So you need to do other changes for a proper exam mode. And this famous buck is the, we don't violate policy by the word, we just violate it in practice because the Debian Edu config package modifies other packages config files but only if you install it and either then manually run a script or if it's part of the Debian installer you choose that CD and by that you choose to run the script. But the effect is that upgrades are sometimes painful because we modify the configuration packages and then the new changes, the configuration files are not propagated to the system. Often our administrators don't want to read the read me or the new Debian file to learn about what they have to do at upgrades. So we have this wiki page where all, where the current status of the buck is described. And unfortunately we started in the beginning of the linear release process, we worked on that buck and then we somehow forgot it again. And so this was, is the status which was at FOSSTEM six months ago. The bugs are mostly make whatever config file automatically configurable, which is nicely solved in Apache with this Apache ConfD directory where one can just drop a configuration file which gets read, but not all packages support this mechanism. And this is not only useful to have for Debian eRule, but also for other customizations and for local changes. I like to just keep the package default and drop my config file there and get used. But this, this buck is nothing, when we, if we can convince the package maintainers, then we can solve this buck with the help of the package maintainers and mostly the package maintainers were helpful and just said, I would accept the fix if you send a patch and we haven't done so. So that's our fault. Then three weeks ago, I was very happy in the morning when I read that Lenny was frozen. And then a few hours later, I was very unhappy when I read SVN commit mails with changes to Debian edu packages, packages which clearly didn't meet the release teams guidelines about inclusion into Lenny. And that really stopped, I didn't know what to do because I know if, I know that if you don't try to achieve your goals, you will not reach them. And there we stopped caring in my opinion. And then the Debian 8 started and I did other stuff. Also because when Lenny was frozen, it was still possible to get important bug fixes in, not only release critical, but only for a shorter time window and I haven't followed Debian release the last two weeks so I don't know where we are. But there will be a time when only RC bugs will be allowed and the bugs we have, for example the OpenELDA package doesn't work with TLS and we use TLS per default and I think this bug should be fixed in Lenny. There's no need that we put our own modified OpenELDA package in Debian edu. I think this bug should be fixed in Lenny. So now we have this hack that the Debian edu package in install in unstable will add our repository to sources list and then so Debian edu can be installed from main and then if you do up get upgrade, a new version from our repository is fetched but in my opinion that's just the same solution that we had with the last release that we just have different packages. And that has the same problems that we need to maintain the fork, need to up do our own security support when we modify the packages. And that's not really what I think is sensible. And what I said this morning already, I think we should change the policy that some packages should be already excluded from the strict freeze guidelines. There's the Debian installer packages, Debian CD, also the kernel, of course we need a good working kernel and in theory the kernel should also be just a normal package and follow the freeze guidelines in my opinion. Especially the kernel. But I think there are other packages which would really benefit if they were excluded from the freeze. The Debian edu packages as they only affect our system are not used by anything else. The fo-fi fully automatic installation which is just another form of Debian installer, some meta packages because sometimes you have recommends in there which you don't know if the package will reach Lenny or not. So these are also just packages collections. And outward packages I'm not sure if they are fine with the current release criteria but I think that Debian should be or not only the critical bugs removed but also Polish more. Yeah, that's, I've been thinking about this release policy that only RC bugs, bug fixes are okay for the release. I think also smaller bugs should be fixable with defined procedures. I'm aware that every bug fix, every change can bring regressions. And I know that the release team doesn't scale to review all the MDIFs but I think it's very strange to release with small boxes which are easy to fix. Why doesn't this work anymore? No, my monitor is fine. Ah, in the morning, I think I'm done anyway. Yeah, that was the last slide anyway. Yeah. The other issue I see with Debian Edo that in our point release, as I said, in the beginning we included more bug fixes, also these minor bug fixes because they were important to our users. And I think it would be good for Debian to do something like that. We did this with Lenny and a half, a bit for certain packages. There are several ways or several thoughts how to make Debian more attractive for end users. One is also another one is this testing security thing that you can install testing and have a supported release. The other way is this cut approach from Joey Hess. I think his third way is to allow more fixes for smaller bugs in edge or in stable. Because you just notice very low hanging fruits which you can fix. And I think it's a pity that it's sometimes not fixed. See, we have five minutes left. I would like to hear your comments, suggestions, comments, microphone. Do you think that it will be ever realistic to just use the Debian archive without any local archive for yourself? I certainly hope so. Why don't you think it cannot be possible? I mean, you have a special client base with special needs. So do you think it will be possible to just use Lenny plus one and the archive and then get your fixes into every point of time and release and stuff like that? Do you think that's realistic? And yeah, or doesn't that create, does that create more, yeah, doesn't it take more resources to get this accepted in Debian than to maintain your different repository for your local changes? Now, changes are mostly this installation changes and a bit of branding and a selection of packages, but we don't modify any core packages. We modified for KDE some translation updates, but I think these translations updates are also useful in Debian to have fixed translations if they are there in a point release. Hello, I have a totally non-serious question. Do you have any interesting stories about students hacking your system to circumvent your exam protection and things like that? Sure. Care to share? No. Okay. Has there been already some serious discussion about this? I mean, I haven't followed Debian release all the time and Debian during the Lenny release process, but do you have a statement from the release team what they think about that? Is that just at the moment like a thing we should discuss or was this discussed? It was a bit discussed at FOSSTEM where I already brought this up. This morning where I brought up this idea of other packages like to have less victory and freeze guidelines. Only one member of the release team was in this room and that member didn't want to comment. Well, I'm the Debian release manager. First I want to clarify that meta packages and artwork improvements are definitely accepted in the freeze. So if you know of them, just send a freeze exception request. For Debian, Edu and five packages, I cannot say that they're just exempted of the freeze but please do try with freeze exception request meals and we will look into them. Thanks. Frank, I just want to come back to your question because you seem to doubt that it's possible to be 100% it is what I also told in my talk and is there any specific reason you think it would be not a good idea to do? Not really but I think if you have a, I was just wondering if you have a special client base and want to have a fix for them and I was just wondering whether the removed flexibility you get when you only rely on Debian stable isn't in some cases not worth, it's worth, it might be worth the effort to have some mechanism to update single packages without having to go to the stable release team and Debian or stuff. I mean for security fixes and like that it's obviously good to have them come from upstream so that everyone benefits from them but yeah. But maybe that's just a pre-conception coming from the somewhat painful experiences in the last few years that Debian had with the stable release process and the point releases where it has several times severe problems with releasing point releases and stuff like that. So maybe Debian can change in a way that it is possible. At the moment I don't see it possible like the current process that we have but maybe Debian should change and not you that could be the answer to my question but I was just posing it. My opinion is yes Debian should change and should be more flexible, yes, it's one point but the other point is if you look at the target audience what do you think how often the installation in the school might change, teachers do not want any change at least this is my experience. And this is one point and the other point is if there is some change in the school it is in a kind of a yearly period to make one school year and so and I have especially in this target group not real problems with this. And there are other target groups which also are afraid to change anything. So stable is completely fine for me. I think so too that I run stable on my laptop or my desktop since 12 years and I'm usually very fine with it. And schools especially they don't want changes. And I think everybody wants bug fixes. So I think Debian needs to change and Debian needs to change. I'm a bit confused about the time because there were two people holding up signs and according to one the time is over and according to the others there's eight minutes left or five. Takes a time you'd like more. In the back Peter. So I'm just curious Holter if you run stable on your notebook how do you watch video streams from this event with it? I use VLC and I use the XORC back port to have XRUND R. The truth comes out. For anyone who cares I run unstable and experimental on mine so. This is the reason why your HP laptop really works. Actually regarding Frank's question I'm quite happy with the answer because I would have been fairly disappointed if basically you had said that Debian AD can't be in Debian because it would have been Debian is not flexible enough to be customized for specific purpose and having a positive answer is good news. Having a what? A positive answer saying that it should be possible to get 100% of Debian AD in Debian means Debian is still flexible enough or can be. Yeah I think Debian is in a lot of ways just Debian Polish and I think it's pity that we are not able to manage Polish Debian as much. Okay if there are no more questions then I thank you for your listening and see you around.