 Sorry. Our next talk is Knut Irvind's talk on custom Debian distributions. Thank you. I just have one warning first. Is it possible to get this sound a bit lower? Because I'm always speaking too high. The sound people. Thank you. If you find strange spelling errors, it's just my handicap. I'm totally dyslexic. We will have to leave it at that. I just tell this to the teacher because they know how to make red marks on my things that I write. Okay. I will talk about client technology a lot. And time spent maintaining custom Debian distribution installation in huge municipalities. I also want to talk about economy and user experience and what next. What do we believe in? School of Linux Debian ADU will be an important path further. And some of this, I have also this warning as a standard thing, is too complicated for teachers. So if it's two teachers here or three, you will probably have some questions afterward. Okay. What's School of Linux Debian ADU? You have probably heard about it, but I will mention it. It's a network architecture out of the box, operational concept, digital user profile, open office and 75 other user programs. It's tailor made. And it's also one important thing that it's made for our mother tongue. In Norway, people, pupils in regular don't speak English. And for the most of the world, that's the case. It's about 344 different countries. And it's almost over, it's almost 400 languages that are official out there. So most of the computer programs in education has to be in your native tongue. And it's also made for the school budget and we have one to two hours to install and configure. Try that with Red Hat, try that with Windows Server. Good luck. We have also had some user evaluations from different parties. And we don't, in this presentation, do any, what you call it, kind of saying that Linux or Free Software or whatever is good enough. Everybody has said that it's good enough in our perspective, in our experience. So I will go to the economical realities. This is the situation in Norwegian schools. The headmaster has to choose between hardware and people. The cost of operating infrastructure with ICT as an add-on to the other teaching activities costs one to two positions. The municipality in Nittadal in 2002, 2001, they said we're going to standardize on Windows 2 98 because that is the only thing we have the money to run and the hardware. After they introduced Linux in 2002, they haven't the money to choose Windows because of the prices of the hardware. There you have it. And I believe this is the same situation in Sweden, in Denmark, in Finland. And if you can think that this is countries that is almost the richest country in the world. You can just imagine how it is in Spain, in Africa, in India, in China, in Italy. Thank you. So to believe, as somebody mentioned in my room, that well it's just to buy a new computer. It's not, the money isn't there. So to believe to buy a new computer is out of the questions for most of the schools. Now I'm going to talk about client technology and I will go fast because I have something up in the end. The traditional computer architecture in the 90s was or are thick clients and they run everything on the client. But the most important things is that you have considerable administration of software on each client. The program is run locally and are administrated on every client. And this gives good support of application. You have one structure of saving files. You can do it on a server and services can be centralized. You use moderate bandwidth, but you must have newer hardware. And the single structure of running application is around 150 euros annually to operate every PC. And that's market prices in Norway. Not included cost for hardware, network licenses or ICT person at the school that doing the pedagogical work or helping out, shifting out old hardware that don't works anymore. So the schools don't have the money. So they should suggest to, oh yeah, we have no money. We will use Windows 98 and we will put on the top a graphical terminal, Citrix or FreeNX. That will breathe new life into your computer. Yes, but it's expensive and I'm not going to explain why. This is the basic concepts. What about FreeNX and Citrix? You must have a local thick client to run a graphical terminal on top. The user application is run centrally, both local and central operation. You probably should use FreeNX. It uses half the bandwidth of Citrix, but if you have limited bandwidth anyway and places your server centrally, you have saturation and other complication that really kills off the client. You don't have the support of sound really good. It's not just working as the pupils expected to work with media rich application with animation, with small movie clips and such. That is very popular in the first and secondary, the first years you are studying at the school. Because you have to also support media rich application, you also need a lot of workstation, then you get double structures of running a user application. The price, the cheapest price we have in a huge installation running Citrix is 240 euro per machine. Then it's market prices, it's huge installation, not included cost for hardware network, licenses or ICT contacts at every school. So what did we do then five years ago? We said that real thing clients will make new life without a hard drive, no moving parts on the clients. It will really solve the problem using the hardware. All the applications and all the installed software is run from one server. The thing clients just handle keyboard, graphics, no local administration. And we also newly have deployed a lot of installation with this workstation. Then we can really give support to everything, video, USB, DVD, more complicated Java application, flash, games, etc. And you get as little administration as with the thing clients. So here you are, the programs are run locally but it's administrative centrally. And this gives a simple structure for running user application and the price running this is 150 euro annually operating the system. And the reason is mostly because of the software where you run it. We have done some calculation with the municipality of Oslo, the main city in Norway, with three schools as an example, 400 users, 150 clients at every school. And we have one advantage that maximum 60% of the machines is in concurrent use. And here you are, the amount of servers. This will be a good one for future Siemens or HP or Dell. They will sell a lot of servers. So if you go for Citrix, you are there. You have a lot of servers. But the school next effort is about saving money for the school. So we really recommend thick clients or half thick clients, even with Windows. And the most important cost factors are concurrent users, the amount of servers, the maintenance of client software, and here are the market prices. Have you seen a picture like this before? Okay, it's the Italian humor. The thing is that when you put this on the graph and you look at 30 clients, 60 clients and 120 clients in a marketplace and you ask for the prices as we have done. You can see that if you have an operation at school with 30 clients, you ask the market price to run this, or you have 60 or 120. And compare the prices running the server. It's about on Linux 24,000 original crowns. I don't remember what's in Euro, but it's around that. You add also on what it costs to operate every client because it's concurrent user. That is the price factor on operating the system. Then you get some interesting results. Graphical terminals are three times more expensive than running workstations with Windows and Linux, if you have 120 clients. Graphical terminals don't scale as well as they do with workstation. The same thing you have about client is that it don't scale so well as workstation, but it's cheaper in the first place because of the software running on the server. But if you look to less disks workstation, then you really have scalability. And the reason for that is that you're running software from one place. You can almost run 150 clients with this less disks workstation. Yes, the difference between think clients is that when running the application on a think client, every application runs on a server. Running the applications on a diskless or less disks workstation, every user application runs on the client hardware. So it's more like a workstation, but the hard drive is placed on the server. This is all technology too. It's like Windows 3.0 with Novell. But it was not possible to do that with Windows 95 because of registry. So it's not a new thing. It's about 18 years old way of thinking. I believe that they did it on Unix too 20 years ago. And what does this tell us? It tells us that this less workstation or half thick clients are the most cost efficient solution if you have a server infrastructure. With running cost that is 40% less than all other alternatives. Using Citrix, Windows Terminal Server or FreeNX costs more than three times more compared to any other client alternative. Laptops are probably the same price ratio as think clients. Windows workstation has almost the same maintenance cost as Linux workstation if the hardware is identical. No schools have identical hardware after one year. And then you have to have different images to soft up the clients. So we know this is kind of disappointing, that one. Because Linux could be as expensive to run and to maintain as Windows thick clients. So if you choose the Extramedura architecture you get higher maintenance cost. Because they have a lot of workstations installed. The price difference between Linux on this alternative is that you don't need to buy software. And if you're running Linux workstation the software in the schools for the price of cheap hardware after five, six years the licenses to Microsoft with school prices is the same as hardware. So you can say that if you go for workstation you will after five years have paid all your hardware if you are running Linux and compared to Windows you will get twice as much hardware for your money. There you have it. And it's also one more thing in this story. From a political standpoint if they're going to promote Citrix the suppliers have heavy interest in the solution they recommend to the school because they earn more money. The other thing is that functionality of the user applications heavily depends on where to place the running of the application. It should be run near the user. But we have in Norway this discussion of a lot of ICT services in the municipalities that really want to run Citrix because they believe it saves money. It's not saving money in the school but they believe it does. And when they do that they remove the application from the user kind of way and they have to transport graphics and everything on Citrix or FreeNX and then there you have it. You don't really are able to support media rich application because of the bandwidth limitation even if Norway has this grant plan of giving everybody every school one gigabit bandwidth. And the most important factor is to educate the teachers because they don't know how to use the equipment. Now I will go and show you some results from the municipality and the workload in medium sized Norwegian municipalities and large ones. And if you look to the left you will see that Hurum has 200 clients consuming 450. They have increased that to almost 600 today. Nittal have 506 clients and Citrix also have planned to in 2008 to have 25,000 or almost 26,000 clients. The interesting thing that they use one half a position just 50% of a week to operate these installations. And they also have a central pedagogic resource and at the school they have or have not an ICT contact that helps out if some technical thing breaks. In Hurum the municipality, the centralized operator does every maintenance of the hardware himself. If they call him he will change the machine. On Kongslinger they use almost three and a half hours a week with five think clients at every school. In Nittal they use one to two hours a week with 50 think clients and also they have 12 work hours a week. 150 think clients. This is the workload. And I have to ask you one control question here. How much is it, what is the, if you have 100 clients on your workplace, how many persons are employed running that operation? What's the usual amount of work hours? You probably know this. You know somebody that runs 100 clients, 200 clients. Nobody knows. Give it a shot. Yeah. One man for, say it once more because he has a microphone and everybody will listen to it. We say one man for 50 to 100 clients. Here you have it. You have one man and also a pedagogical responsibility for 450 clients. Or half a position. Locally if you add that to 10 schools you also have a week. So you have almost one and a half work week. I call that that in the region. I'm sorry. But anyway you have one and a half position running 450 clients. Okay. There you have it. At the schools they use one third, one fifth of what's the regular manpower running an operation. We have put that up in a charter when running LTSP clients on older PCs. ICT contact central operation, what it costs annually. And it's very efficient. I have learned them as myself. This is my guys. Hurum has a lot of money. So they have built one gigahertz infrastructure. They have more money for every client. They have bought newer equipment. Kong swing is between. Half of their schools have one gigabyte fiber to their schools. And four of them have just 1.5 megabyte to their school. But anyway that's not so important. The important thing is the annual cost of every client. Here you have it. And if you add on the hardware you will see that just 60 to 70 percent or as much as 70 percent of the cost is manpower. And we also have calculated the cost in 2008 doubling the amount of PCs because every municipality really does that. They double the amount of PCs because it's mandatory to have computers in their education. And the interesting thing is that there are in fact a reduced cost here for centralized operation. So the person running the installation centrally they want to double the amount of the clients. But they don't increase the manpower because they don't believe it's necessary to do that. So School of Linux is pretty scalable. And we have also one more observation. It's more cheap or it's cheaper to do centralized operation than a local one. So if you have an ICT operator on every schools running the operation and not have anything centrally and don't centralize it's more expensive. And it's more expensive running Windows than think lines. And the parties that say this is Teleplan is a kind of private but independent firm that just investigates what things cost in a telecom market and the municipality of Oslo. They used 2.5 million crowns to investigate this. And what those experiences in municipality tells us? Central operation is the most efficient way to maintain the system. The school in its scales well. Think lines are cheaper to maintain than think lines in most of the cases. The cost of Microsoft licenses over 5 to 6 years is the same as the hardware. It also tells us that supplies have to cut the amount of clients or reduce the cost of operation to match School of Linux. Why pay more to get less? And this is a market thing. The last one is a marketing question. Because School of Linux is propagating and operating in a marketplace. It's not a kind of national effort. It's a kind of you have to compete in the marketplace. And this one is probably the most important thing for the decision makers. If you have to choose for one infrastructure with Linux compared to Windows or proprietary solution, you have this advice to the decision makers to make a budget showing them what Linux costs and show them what Windows costs. Okay. I will just go through some objections we will meet in the usability space or user space. Yeah, you can ask a question. I was wondering if maybe on those cost analysis you saw that there was a strong impact in terms of cost because of the technical training for the operators switching from Windows terminals to Windows-based terminals. I mean, if you had more cost because of that. Yeah, the thing he asked about is the cost of training for Linux operators. Is that correct? Okay. I will take that because I am coming to that now. And I will take the objection that we meet all day. School of Linux don't support some. OpenOffice makes a mess. Using School of Linux don't prepare their pupils on their future jobs. You have to be a Linux guru to operate the system. Here are your questions. And there are no pedagogic programs on Linux. This is our work every day with DB and EDU. Okay. Objections about the client. Some does not work on School of Linux. Some works on thin clients, but you have to turn it on. And it's not always there. So you have to ask your vendor or your supplier of reused thin clients to turn it on and have sometimes to find the right sound card. It's not easy, but it's doable. With this test and thick clients, it just works. Remember, clients are intended to use on 10-year-old hardware. That's machine that you will just throw away anyway. So to expect that to give sound, I believe you should place it in the areas in the school, in the library where it's expected to be sound-free. You shouldn't have sound at all. You just choose half thick clients in the music lab. The place is where you need sound. So there you are. Have it considered this workstation that will give you the sound. The other objection is about OpenOffice.org. OpenOffice makes a mess when people take document home from school. Okay. The people who have MS Office, the parent MS Office, have their home. What to do with OpenOffice? The most easy thing is to turn on the support of Microsoft Office 97 as a default. Or teach the people to save MS Office 97 for their home use. It's easier to do that in secondary school. Or give them a CD with OpenOffice to use at home. Tell the parent what to expect. Strangely enough, in Norway we really promote, we really have done a lot of political work to make the government support open standards. And we have this one question. If I'm going to have a dialogue with public offices or to interchange some data, whatever, should I be a customer of Microsoft? We have asked this question a lot of times. And they say, the government say, we should not be a customer of Microsoft. You should choose every operating system you want to. But anyway, coming to the daily work, the teacher are not there. So how should we manage that? And the easiest way is to turn on OpenOffice support, it should say 97 here, as a default documentation if you're going to have an easy life. And this is a huge objection. We must learn Microsoft because of work possibilities. And the answer is my kid should not learn niche products. Well, the main objection is my kid should learn niche products. But then we have this argument and it's a bit difficult. The first one, we did not know 15 years ago that word perfect should be replaced by Microsoft. We don't know what to expect in 10 years. So the people should be able to handle change. And that's why it's a good thing to learn people's different solutions. The other one thing does not exclude the other. It's the other, what you call it, helping to get rid of the objections. And most of the teaching programs is platform independent. You get it on the web. And here you are with your question, the Linux guru question. We don't trust Linux people that are interested in the cause. If you are a person that are into Debian, you are a danger for your community or for your municipality. Have you heard that one before? Yeah. Yeah, it's dangerous. Look at them. Strange shirts and you know. But many municipalities that have started with school have no prior knowledge. They run big installation today without being a Linux guru and they also support windows on some clients. They rely a little on external support that is easy to find in 2006. That was not the situation five years ago, much have changed. And experiences with the municipality you have seen here is that you need, they say that to meet two things. You need less support. The person that had started rolling out school Linux have no prior experiences. And they have done it anyway. And they have done a bigger installation that they were able to do with competitive proprietary software. And they have just learned the thing through the net and some small courses. So every money spent on doing this is the opposite that we tell them. We tell them that you need more money to learn Linux. The truth are that the municipality that has started with school Linux and roll it out and supporting it uses less time learning it, less time supporting it, less time to buy external consultants. So who is telling that Linux is more expensive? The person that sells support that are more expensive than Linux. But this is very difficult to explain because people don't believe that what you call the second runner up is cheaper than the biggest and most supported platform. Here is the last one. There are no pedagogic programs on Linux. We use drill pro is a popular program in Norway. And other programs running on Windows that don't run on Linux. Well, we run those programs with Vine and it works. Most of the pedagogical programs is used in the web browser. So most of the vendors now test it out on Firefox on Linux. And the most of the school Linux has completed the national exams with success using the national exams in the web browser. And I will sum this up. I have two more minutes. Five, okay. We have some difficulties and this I will propose and this will be a kind of we want to work together with in Debbie and the EDU just to fix some things that are problematic for us. The first is that in school Linux 2.0, the swap and the client was turned off as default. All of the clients through the two megabit RAM crashes in some municipality. It talks about 50% of the machines. It's about 250 to 500 clients that don't work anymore. And that's not good for your reputation, is it? Upgrades from one... Version one of school Linux 2.0 is too difficult. Multilevel configuration will handle a lot of that. Do you know what multilevel configuration are? Is that known to you? Multilevel configuration? Yeah. Okay. It's a bit... Sometimes it's up, yeah, five minutes. And I ask because we really need help on that because when upgrading the configuration breaks we have also done some hacking in school Linux that have made it more difficult to upgrade. But we really have some... We need some solution that makes upgrades easier. And the hardware support is not good enough on laptops, even on servers. So we have tailored Kubernetes with laptops with the configuration files we need that connects the laptops to the school's network. And we also in the future probably will be nice to have support for multi-architecture that can support different clients because we could have AMD 64 processor on the server and we have half-thick clients with Intel i386 support or Intel i686. And my last advice on this talk is don't talk negative about different tests to pollution. It's application that the user won't talking negative about other reflects back on yourself. That said, people that has user windows tend to like KDE. That's just an observation. It's not a science study, but it's the feedback from the teachers. You can't... If you have come from a window as well, it's a bit easier. Our competitor is proprietary software. The ICT service in the municipality or the headmasters could be bought. The low windows, okay? It's safe. It goes after the lowest hanging fruit. We help people that want help. Why use our scarce resources on people that won't have any help. And here you have some other guys in Norway. You always up there someplace. Andreev. You probably have some questions. Thank you. If no one has, I have one. There's not much time left, but if you can tell what... I mean, we've seen the Scholar Linux is successful. We have seen that Scholar Linux is successful. It would be interesting since we are all around here with lots of Debian developers to know what Debian can do with Scholar Linux, where it works well, where problems are still around. And if you can give suggestions to make the life of Scholar Linux easier. The three suggestions we have is to turn on the Swap on the LTSB. That's the shortest win. The second one is the multi-level architecture. And we have a third one. Well, not multi-level architecture, but multi-level configuration. And multi-level architecture in the future. But the other things is more handleable. Handleable. My original English. Thank you. So, on hardware support, with units, we have always some objections when the bosses in the municipality don't... Debian don't work on their laptops. They tend to go to Red Hat. They tend to go to Kubuntu. That's a step backwards, because they should use their Debian that they use on their servers. It was one more question. Over there, I suppose. Okay. Yeah. If you can shortly say what kind of applications or the list of applications that you include. Well, I can say something about... They include a lot of the applications, but the most used one is the web browser. Okay? They don't really know yet what to use in their laptop on the... They don't know about computers. So the biggest problem we have now in Norway is to get the use of the systems. They have installed it, but it's a kind of thing that the use of the computer have a bit stagnated in the seventh and ninth class level or school level. So it's the web browser, then it's the office application. The pupils hate the office application. They shouldn't use it. But anyway, they have to... They are enforced to using it. But it's a lot of other applications that are not very in use, but it should be used. It's squeakland with e-toys and making simulation, that kind of thing. K-stars, that kind of science application. That's proper in the science department, but it's not widely used. So we are measuring that to catch up what's really in use and try to promote lectures and sharing lectures using the right software. I have to wrap up, probably. Okay? Thank you for listening. All help we can get from the Debian community will be nice to get to school links and we will promote it as hell in all of Europe, in Africa, in India, whatever. So help us, please. Thank you.