 The long answer is, well, I had to find the driver for the network device, which was not included in Gabron, had to fix some bugs in this driver and had to ask for some maintenance because I was not able to fix this problem on my own. This took me about one day and further the not supported 3D graphics of the ATI device was another task for two days to get the 3D acceleration working until it was not real fun. And so I can understand if some people are not really happy about the interior, but we cannot do much about it. So we cannot force ATI to release their code or whatever, but it's just a fact that there are outside users out there who are not really happy with the interior. These users might also be a lot of media enthusiasts, but there is not enough support currently, but we have no real chance to help them. These are just two users who probably will not lose data, which is not really our fault, but we could at least try to give them some hints and some documentation that they could do if they really wanted to try it out. The lucky user's fuel is, well, he was lucky enough to obtain some supported hardware. The installer went fine and he finally reached the task there, the selection of some 10 items also to select from and most users who are not really comfortable are wondering what to do now, but I want to do some desktop, what desktop is it? In my opinion, the start-sale selection is not really as user-friendly as it would be, because it works for many people, or for millions and loads, and you just have to do the synaptic or whatever to continue from this step on, but I'm a little bit missing the kind of needing hand for all users to make the system just nice, just as I wanted it. The different part of you is the upstream developer. The upstream developers, so those people who are developing the software we are including into Debian are not necessarily the Debian users. They might be Debian users, but in one way I think there are also Debian users even if they are keen on some other distribution because they have other users of their software that are working with Debian, and so they use Debian as kind of quality assurance for their software. A Debian developer, for instance, noticed that there is no manual page or whatever for the software. Most Debian developers are writing many pages, so you do not choose to do it, and forward this approach to the upstream developer and thus the upstream developer has some profit from those Debian users. As well Debian is supporting upstream software to other architectures, and very often there are some problems detected by these supporting teams, and if these problems are reported to upstream, these people also profit from Debian. In this very abstract view there are also users of Debian. Well, the Debian developer's view is as high as the kind of missing link between upstream and the end user because we have in between the software which is anywhere at the top and the machine of the user, then it starts to use the software. And if we are doing a good job, our users will be happy. And also the upstream developer is also happy in the sense I tried to describe before. Well, there is some principle of free software. You stay just independent from commercial interests of a producer. I mean, if I would be an employee of Red Hat 2 or whatever, I wouldn't have the freedom to pick whatever upstream software I would like to integrate. I have to follow some... I have to care for those software which makes a big market share and tries to make the distribution which my employer is promoting become a wide user base. Wide user base you can get easily if you stay in a large market and not so much in nice products. If you think about Debian review which is caring for schools, the schools are not really known for paying much money for software. So the main distributors will not really care about schools or in my case with this Debian main stuff. These people will not care about medical stuff because there is not so much software to bring to the medics. And so in Debian I have the freedom to say, well, my idea is I want to support this special field of software to bring it to the user and I have the freedom to do so because nobody would say, well, do something more useful. In one hand we are some kind of hunters and collectors of free software because if there is a piece of free software and Debian will be developing, oh, this is the right thing that packages and integrates. But in principle we should rather be more less designer of a comprehensive system and not just cherry picking anywhere. And my point of view is try to make some consistent system which really helps the users instead of just packing done software. Then finally we have the driver's view. They say, well, Debian is really nice but it does not really fit my needs. I would like to change this or that to make something else. And they just use Debian as a nice point to start from and then takes the freedom to make some other system which is find the license allowances. But I think it's not always the clever approach to go. What means universal? Try some search and work that's dictionary and what means comprehensive, general and also adaptable. I think these points are quite valid for a user as it is and it is really adaptable because we can adapt Debian for several things and the developers are doing it but they're doing something else. So another try has also three meanings if you are asking for the adjective. And here is one point I stress a little bit with what you call it. It is adapted to various purposes such as form and operation. So if you have, for instance, a screwdriver which is adapted to, well, you can mount any screw or you have a tool which can be used for many different things. So in this sense I have limited problems with Debian because it is not really adapted to anything. It is adaptable but not yet adapted to certain means. Special cases, for instance, the Debian Edu, the Castle Debian distribution it is really adapted to use inside tools. It can, for instance, handle students' accounts. It enables students to run educational software games, whatever and it does it in a really good way and it is also a good tool to teach students basic skills and informatics science and so on. And so Debian Edu is adapted from a teacher point of view. The adaptation is about a few points on the scene. But if you are, for instance, a scientist you find in Debian very interesting applications. They are included. There are many outside but well, it is quite interesting. And if you have nobody, please be quiet over there. Thank you. So Debian has a high potential to be a full featured scientific workbench but it is kind of served in pieces. If you are going to a low budget market and buy a workbench you have to assemble it yourself. And this is more or less what we are currently providing. It would be really good if it would be high quality tool provider who brings the workbench to the customer and puts it in their workroom to be ready for us. So we have no real concept to guide scientific users. So it is just only adaptable for the scientist point of view and well it is slightly better for biologists because in Debian we care for these biologists but there could be done some more for other scientists and also other fields of... We have this multi-media project which is not really active inside Debian. We have this Debian Lex which could be do something for lawyers and so forth but these projects are not really in life and so I think we could cover those special things much, much better if somebody would agree with me and start some work with us. So what about development inside Debian? Do we scale well in every direction? Debian is kind of multi-dimensional room and we have an increasing number of people involved which means not only the debut developers we have many people who are providing work to Debian as translators and bug reporters and sent patches and so the number of people who are working on Debian is increasing and the number of packages is increasing from release to release the number of architectures is not so drastically increasing we got some more we even lost some but the tendency is also an increase but the number of bugs also increases the number of bugs who are closed is not so high as those who are opened we have an increasing number of users we also have an increasing number of derivatives and last but not least we have an increasing number of labels which is not really nice but these are some points we are increasing numbers mean there is some change if you move forward into quantity at some point in time you just gain a new quality which is not very new and the problem is to find the right moment in time when this critical point is reached where the quantity makes a new approach to new quality and it is hard to find out this point I think we just stepped over this point Debian has reached a new quality than we had in maybe 2000 and we go into some process of evolution to move to something different I think it is a positive move like well if not we would not be here we would not be at the Debian conference and would produce new things but we have to be aware that we cannot approach as we did before we have to find some new methods to work on Debian and so I think that passing Debian distributions are ideal to deal with nearly all these dimensions of growth I mentioned before some quote from 2003 I hope this year number is fine compared to my talk number is when Manoj Sivastava said for the most part Debian is a bazaar of casinos with a few producers in place to override the low level casinos in exceptional situations each developer has within reason full control over his packages when you are following Debian technical policy just creating the low level procedure the technical policy and the general resolution protocol means of overriding developer decisions about their own packages this means more or less we have kind of a flat structure with some controlling mechanisms and the structure means we have one-to-one relation between maintainer and package this has just changed because we have some group maintenance for several complicated packages so we have an end maintainer to one package relation one package means if you have the X packages the result are several binary packages but many people are caring for the X packages which leads to some reliable teamwork and this work on the phase 2 means we stay on a technical level but I think we will evolve further on where in custom distribution we have an end-to-end maintainer to package relation that means several maintainers care for a set of packages which builds some kind of close subset inside Debian and so we have some complete suites covering use case and we have users and all users have their own use case so we have to find out which users are keen on using Debian in their special field by the way, did anybody solve the puzzle? it's up, it said anybody solved it so why do I think that Debian scale is higher? we have the increasing number of people so if we try to apply some substructure to Debian all of our number is the same but if we have smaller projects people are working on who are focusing on a certain field we have some better focus on these packages in these fields if we are subgrouping the number of packages we have closer packet sets and these subsets are easier to maintain in the overview and so the number of packages problem which is growing continuously might be solved well, the number of packs will increase more or less we cannot do much about it but if you have a number of maintainers who try to care for a number of packages at least I hope that they will try to have more motivation to fix those bugs who are relevant for their custom Debian institutions also the number of users which is increasing in general is reduced to some specialists for the Debian education of the teachers and so the maintainers try to cope with the specialist issues so they try to talk to teachers and ask what they can do for them also the number of derivatives is growing as I said but if we do some reasonable customization we could perhaps offer some means that the drivers might say well, there is no real need to derive because everything is inside Debian and I prefer to join Debian and work inside Debian than just arriving and start from scratch all the time but one thing I did not mention in my list of items is the time span in between the releases because this was for the last release not increasing we became better, this is really good this idea that we could release the time span could decrease the time span between releases this custom distribution was born because it might be possible to release subsets of Debian as custom distribution for instance if you think of schools they have a certain frequency to upgrade maybe the school year so it makes sense to have such fixed releases for certain subsets there are no techniques to do this currently there is just the idea but maybe it might work in the future so what about these derivatives Debian has according to probably some old numbers 129 derivatives, Fedora has 630 and all others has less than 30 what more Fedora has 63, sorry and well, the question is, is it a good time or a bad time? who thinks Debian is a better distribution time? distributions are a good time other people think it's a bad time it is good if you see large numbers but I think it is a good time that people regard Debian as a high quality distribution which makes sense to you and to derive from but on the other hand these 129 people think Debian is not good enough for my use case and so there is a pro and con in these numbers and in this talk I would like to stress the not so good side because if something is good we don't have to change it it's fine, if people like us it is fine I know what the puzzle is the puzzle is needed in a high stack yeah the puzzle is needed in a high stack what do you mean? you are brave enough, one free beer for me and so I would like to express some ideas how we can make those people who derive from Debian even more happy than they are currently what kind of deriver do we have? there is some kind of poor diligent deriver he has some misconception about Debian he thinks Debian is just one way he takes Debian as a running product which can be changed in any way so this is a few of the the use of some probability of software who has to accept what his provider of software hands him over but Debian is something else you can work on Debian and you can tell us what's not so nice in Debian and what you would like to change and then this poor diligent deriver has perhaps some problems with human beings he is sending a patch, please change this or that and the maintainer says go away with your stupid patch, I have no interest in it this is not not nice but we are doing human beings and please try again and perhaps the maintainer has changed his mind the problem for the diligent deriver is that he continues to work to keep up with the development of Debian and so I have observed several derivatives of Debian or the most famous derivative is Knopix which has 50 derivatives of his own those projects tend to die at any point in time because the deriver has not enough time to keep up with the speed of Debian development so my hint is rather try to cooperate than to make something else then there is the impatient desperate deriver he knows that Debian is something you can work on it's something you can influence but he faces certain showstoppers in Debian not every maintainer is responsible as he would like to be and if the deriver has a timeline he has no choice to change Debian in the way he wants to be he has to derive and he knows the principle of bureaucracy that those who do something are those who decide what is really done and so he just takes Debian and does what he wants to do but here I have also hint to leave open the way back to Debian because if the showstopper is solved and if it's possible to reintegrate your stuff just try it then we have some possibility employed derivers I named here some from Munich and Vienna and some other Spanish derivatives those people are gathered by the government to do something else and maybe it's known fact that every authority thinks it is very special and cannot do the same as some other authority so they need their own stuff so this deriver is more forced by social reasons to derive from Debian an alternative would be to continue the Debian eGaff passing Debian distribution guys it is much a sketch what could be done in a diploma work and there is not really much substance behind it a similar situation is our funded project the DeMuni project the Debian multimedia distribution was founded by the EU and the EU wanted to have a product and Debian is not their product so it's at least something else then we have the lucky life cities life cities are really cool you can demonstrate something you can test your hardware your favorite OS on random computers live DVD from Klaus Knobot has saved my life and I was sitting in the middle of Iceland and I had to find an internet cafe to fix something at all and we have very good usage of life cities in key systems and if you perhaps want to give a Linux course in a computer lab you are quite fine with life city the problem you have to keep your system up to date and this is really a problem I've heard there is even a book of 200 or 300 pages how to remaster Knobot it's really nice to remaster Knobot but if you have done it once you don't want to do it again because it's over and over the same work and doing the same work over and over this is not really fun for people like us so just use Life Paper I've heard about Life Paper here in this conference and it's really great and just use what you want to do give this a list of packages you want to have on your life city and some configuration you want to have and enter some line and say make it, make my life city and you are ready so perhaps you can join this effort regarding this two orchestras principle and make the life ever even better to fit on these and so on lucky life city creators could use plain Libyan stuff then we have the clever commercial Libyan those people want to turn Libyan into a sellable product Libyan is not sell anybody can use it for free but there are many companies who base their work on Libyan stuff they know the names they know the secrets and these people they add some value to Libyan and for instance this might be non free drivers to help those poor users I mentioned in the very beginning or print some manuals they can also sell some service for enterprises what we cannot really do inside Libyan because we are just volunteers and so it is basically about rebranding Libyan and add some value I think Libyan supports those drivers or should it even more support to make their work as easy as possible but they should be aware that Libyan even competes with those drivers so if they are not good enough use Libyan instead of Libyan in general I think the drivers should one lesson learn the difference between Libyan and the derivative should be very small smaller than some epsilon and over all time so if you derive more and more from Libyan I think the success is not as much as it could be so keep the diff always small and remind we are on your side talk to us if we can help you it's fine we would like to so finally I would like to say something about special applications inside Libyan we have to be aware that Libyan becomes larger and larger so the question arises how many packages are good for Libyan and I don't know I don't really know but I think if we bring some special applications under quality control we are able to attract the wider user base and this can't be really bad and so I think if there is a maintainer who cares for his package or a group of maintainer who cares for their packages it can't be really bad if we have more and more packages if there are packages who are not maintained we can get rid of them because they probably have also not a large user base so how many packages are good for our users is a real question and I think if we include special applications only if there are small user base it's really nice for our users and we can spread into areas which are not covered by other distributors finally the question is what about large user data yesterday I heard the talk about Zangra and with the GNOME projects they have really huge data and so it makes no sense to put them on every deviant in the world but it would be interesting to have a solution for those people to perhaps have a data area inside the deviant there is a long standing open bug number 38902 which is just tagged won't fix without any command which is not really nice because I think we should do something about those people who need large chunks of data so what is the basic goal of the custom period distributions we have more than 15,000 packages and our users are not interested in all of them they are interested in the subsets and we have some groups of specialized users who are need and very easy installation and configuration of these subsets of packages so while deviant stays a general distribution this is the main point, this will not change it should support specialists as well to make deviant really the universal system only if we are able to support every user who wants deviant support it will work for everyone so we try not to do a derivative from deviant and the basic idea is do not make a separate distribution but make deviant fit for special purposes in the future I think custom deviant distribution could solve structural problems inside deviant and I have to say I do not really like this name custom deviant distribution because most people who hear this name think this is something else in deviant so if somebody has a better name for this I am open for the couple it should make me fit for user interests and in my opinion it makes deviant stronger if done it in the right way what is the right way? I hope I can elaborate next year more about some common techniques which should be shared which my first attempt did not really work but I hope that I learn some lessons and make it better for next year and so the future of deviant is also known as the last final step towards total world domination I think I totally agree with Tosten who had to talk about this before me it was nice, I was not able to join and so to our right it is a need and a high stack and I think the custom distribution is kind of a need and a high stack for those users who are seeking for needs inside deviant I want to show them immediately and not just by wrapping all the huge high stack packages just show them what they really need and makes them happy to become well you know a future base so now open for discussion and I hope you like this idea and I hope you like this idea Okay, so firstly I should describe who is that I am employed in direct and by-marked shuttle work so I don't think that that's what I'm about to say so I agree that CDDs are a good thing and they should be encouraged and so forth but it's very important to realise that I think there's nothing wrong if somebody chooses to derive from deviant this is what free software is free software is software that everybody is allowed to modify and that everybody can modify and when people choose to exercise that freedom they have a lot of reasons which may or may not be something that is maybe wrong with deviant that's the case then obviously we should improve that but sometimes they have their own reasons maybe they disagree with us about something and it's definitely unreasonable if we say that people should not if we say that somehow we disapprove all these people doing something wrong in fact the biggest problem that we have at the moment with derivatives is that it's fantastically difficult to do a derivative and we really need to put in an awful lot of effort and we can't keep it up today properly basically because our source code management tool is so poor and I would really like to see some improvement in this area this was in the wacky ideas both earlier and if anybody has any comment ideas please talk to me about it really there's nothing wrong we say I want more people to take deviant I want to easily be able to derive it I want some random person in their house to be able to make a derive distribution for their friends and still have reasonable security support and for it to all work properly and we're a long way from that unfortunately I have a simple question maybe I don't understand the difference or the boundary between deriving a distribution and customizing a distribution to me it's really interesting to customize a distribution I should have mentioned we have a certain term custom deviant distribution that is a name for customization inside deviant this name is badly chosen because it's misleading can I explain because you've actually had to mention that CDDs do not have different versions of the packages that are already in deviant so CDD consists of some additional packages in deviant and some extra attention is paid to certain packages that are interested in but all the time it's the same versions as our deviant problem it's customization that's what CDD it means and derived distributions are ones where there is a different source code different binaries for at least some of the packages this is what's called deviant internal projects and we wanted to make some separation between technical projects and these user customization but we succeeded completely in confusing people who are just wanted to derive deviant and think custom deviant distribution is something they want so we have this term found and if we find a better name which is not so confusing I would be really lucky but I think we have to explain it anyway so if you have deviant dvd or stack of deviant CDs we have everything which custom deviant distributions provide it's all completely inside the environment this is what this term means I'm from the city of Mene and actually what we made program I think is that we cannot cope really with the development of CDDs we have to take the stages there is no development for stages and that's actually good for us this can say ok we have problems with this kind of hardware because it's new and this can change for example XOR the 810 driver from new hardware that's just one change to one package but we would not be able to say ok look all the time we have two or three people who remain development and it's possible for us to run with the deviant and the city has to work so we cannot we have a release cycle inside too we have change processes and instant processes which make a new release every once three months something like that and we just take stable and have some new features for us but I don't see as long as there is not something like development of stable there won't be many people who really stay in media I'm in real favor of keeping stable for such purpose and I think you should at least try to make sure that the next stable distribution contains everything you really need if you reach this for this particular moment and then there's also this backport effort which at least I've heard about it this is quite good and supports the main features of new versions which are not stable but you can perfectly use it it's also signed by deviant developers so this backport effort is something you could at least got to get some help from we use it some of us most of the time we just stay in the stable and we don't want to change all the system every three months I'm not sure how currently unstable but if they maybe the whole distribution is new every three months or a half a year every second has changed and that's impossible to test I think I think you did some local adaptations which is besides this keeping up with this adaptation and customization if you bring this inside deviant for for certain purpose other people can just learn from it because I think Munich is not Vienna but if they share some code between they could have some common profit sure but that would be for stable not for infrastructure to really level up for stable in Vienna now once we will release Lenny at least I hope so if there is some common code in Lenny you can have common profit we might be attached at the end of the year maybe they agree maybe there is a way to really do some development for the stable too in some extra territory or something like that where people can share it themselves they are perfectly well in reason to work the rafts this is a perfect development but try to keep the tip small I would try to my name is Paterenek I'm from deviant u developers we have I think we are actually starting these derived things in deviant 6 years ago and we have always focused on making as few modifications as possible as much cooperation as we need and making sure that the deviant packages are stable in two door fixes so we don't have to modify them the next place and this has worked pretty well we started out with four key special packages and at the moment I think we are down to eight hopefully for landing we should go down to zero but it's a long and slow process to convince all the for the maintenance of the packages we need to do the modifications we need it's as simple as how do you automatically configure other servers during install time a lot of maintainers in deviant do not really understand the program they just say well you just fire up your image and edit it but that doesn't really solve our problem we also want to make sure that all changes to the configuration files are kept during upgrades and that doesn't really work very well at the moment we often give the system the choice between a configuration file that used to work with the old version and a new configuration file that doesn't really do what they want for the new upgrades so I really wish that the deviant maintenance would keep this in mind and fix their packages to have the upgrades more riskily and have the overall modifications but my point is that yes we cannot modify the stable distribution that much but actually there is the edge on the half at the moment hopefully there will be a process to get the new return on the new X server even if you can modify the stable distribution you can get your fix into the unstable and then you wait for two years and then you actually have modified the stable what are the other experience John is called in Indonesia I have a question what is being done to encourage people that are making these derivative distributions to give back their improvements to the deviant project I can say exactly holding a talk on devconf holding a talk on devconf but I try to explain how to give back something we cannot we have no handle to tell those people just backport your stuff I can only say we are open try to give something back once seems like they should realize that there is a moral obligation to do so yeah but what should I do if somebody just goes and derives deviant but how should I have a communication channel to the person he is free to do so he is fine to do so but if he does not ask me to do something for him is there any kind of license that would require them to give back no there isn't that would be a non free license yeah exactly so deviant does exactly this to upstreams you might as well ask the same question you know what do upstreams do to try to get us to tell them about the changes that we make and the answer is most upstreams do nothing at all and it is up to us to go and tell those upstreams you know here as these changes please take them and well some deviant developers do this very well and other deviant developers do it best well and I think it is very much the same derivatives one interesting observation is that it is the derivatives advantage to actually push the changes back in I think that is the mechanism that makes this work if they are deriving too much the maintenance load will be too high and they will just not be able to keep up that is why a lot of deviant developers are pushing changes back into upstream and that is why school unions even in EU is pushing our stuff back into deviant we don't really want to have that high maintenance load but other than that I don't think we want any harder enforcement mechanism because that would make the software not really but the point is here deviant competes with those derivatives so if they want to to be as good as their competitor they should provide something back maybe I don't know if this is your answer I think actually that is kind of the opposite of the answer some of the derivatives feel that if they send their patches back that they will be helping their competitors more than helping themselves I think that they are mistaken and well everyone seems to suggest that the people who don't do that don't always last very long I think you are the option of standing on someone's shoulders for their toes I know what I would choose and probably my answer was bias by the deviant how many minutes left maybe for the next two days therefore there are only questions