 Welcome back to the science track. Our second talk for today will be Debian the ultimate platform for neuroimaging research from the neuro group Michel Hanke and Yavuz Lavalchenko from what is it? Dartmouth, right? From America. Okay, thanks. Thank you So this talk is about Letting you know that Debian is the ultimate system for a neuroimaging research in case you haven't noticed that yet and The talk is going to be split in half. So I will present what we have done over the past Five years and then Yarek is going to take over and talk about what we are trying to get at Not necessarily over the next five years, but sometime in the beginning Just few words You know who we are and right now we are postdocs doing neuroimaging research of one or the other kind and but we have been More or less involved or at least using Debian for quite a while before we got into contact with Neuroscience or neuroimaging research, but then we both happened to you know somehow Have to get PhD degrees in neuroimaging on different continents and different universities, but the goal was the same so In case you never wondered what neuroimaging research is about it's about figuring out what happens When and where and why in the brain, right? So you probably if you if you read those popular science blocks, for example, you heard about People figuring out what's the difference between Democrats and Republicans That's us not the difference, but the guys who do this that that's us so and Most of that type of research is done using a technique called Magnetic resonance imaging if you're into skiing you've probably seen one of those, right and So this it's essentially a big magnet a couple of times stronger than the Earth's magnetic field like a couple of thousand times and What they typically do is they make pictures of ankles and knees without having to cut people open and if you turn the body around and Put them head first in the machine, then you can also take pictures of Brains right without having to cut people open or otherwise torture themselves. So it's completely non-invasive technique and Once you have those pictures. This is the anatomical Image of a brain you can then use software to Look at certain properties. So for example, there's software that turns this volumetric image Into a surface so it estimates the cortical surface looks where it is, right? And then you have a two-dimensional thing that you can do analysis on and and try to locate certain properties of you know what the brain does and the second major Way to use this technique is to not only do a snapshot of the anatomy of whatever is in the scanner But also look at what it does over time So here you can see the time series of one of these volume elements that people call voxel in that field and You can see what it does over time So there are periods where it's above the average and periods where it's below the average And then you can associate that with what the subject the person was doing While the scanner recorded the data and you try and figure out what has that to do with the you know with the brain did and Eventually you might be able to use that data and train more sophisticated algorithms on it that you know from from computer science research and machine learning for example and Make them predict what the subject had done at the time or Associate that data or the properties of that of data from magnetic resonance or resonance imaging with other Data acquisition modalities like e.g. Which measures the electrical field of the brain activity so and With that you could do PhD in your science your imaging and So this is pretty much what we're what we were trying to get and when we You know we're Beginning that process. I had seen a new imaging research Institute in Germany in 2002 That solely ran on Debian and they did all their research and data acquisition on Debian So I thought this is this must be it right you just Use Debian and then it's going to be you know six seven months and then you're done could be used to the next step but When you actually looked at the the state of the archive in 2005 No surprise there were you know There was a lot of support for general-purpose software like you know computing environments like are which has been mentioned before Python has been mentioned before general-purpose visualization software but then Those scanners they don't give you the data as a you know comma separate value thing Right, but she then could load into something but they use dedicated data formats and only two of them and not the most You know important for that type of research that we wanted to do were present in Debian and worse There was few to know no, you know specialized application to do Data analysis right without having to develop every algorithm Or at least glue existing algorithms into an analysis pipeline So It didn't work right in six months So luckily we had people that tolerated us staying for a while in their labs because to do the Debian way you obviously Need to go look what you need package all of that right once this is done you can eventually start doing research and So we needed software for the all Pipeline of doing things right so you need software to collect the data in that field That's typically you conduct an experiment, which is a series of visual displays or people you know Getting sounds played over headphones while I'm in the scanner that all needs to be done and Then you need software for data analysis You need to read the data and the format that you get it from and and finally you need to do something It's to be able to You know to put in papers right you don't get the PhD without having published so at some point you need to document what you're doing so you need to visualize and that wasn't there right and so But luckily What we call upstream was active right and people developed software for all these things and it was just waiting to be packaged slightly simplified and What I went to walk you through is is a simple example That's not a simple example, but it's an example of an important software package for for our Research and you will find you know many things very familiar because that's most of the stuff is happening in other packages too With this one, it's it may be a little more extreme because it accumulates all the problems that that you might get if you want to package something so The example is is a relatively comprehensive software toolkit for neuroimaging or magnetic resonance Imaging data analysis called FCL a University of Oxford produces it and its advantages are You know first of all it's well documented so you can you can read the documentation Understand how it works and they provide what they call a feats suite which is essentially in In Distribution terms it's a regression test That's a very precious thing to have because whatever you do to it You can run it and it will tell you if you don't do it or not Right and so you can you are actually in a position to figure out whether you compile it and use it the right way On the downside it's in deviant terms non-free software and it's it's most of the Not most but a substantial portion of the of the software that we have to deal with is Non-free licensed because they have terms in the license that prevented for use with or in clinical Applications so the doctors may not use it to treat patients in most Countries this is forbidden anyway because if the software doesn't have Passed a certain certification process you are not allowed to use it anyway But they put it in their licenses which in deviant terms makes it non-free so we have to deal with non-free software and Other than that it's very similar to other scientific software projects They have multiple developers they come to their PhD develop the piece and then they go and as a result You have very heterogeneous code, right? so you cannot do Grab and set something throughout the code and then you have fixed the problem because every single sub project does it in a slightly different way and mixture of Languages they have a public mailing list But they have no bug tracker and no public version control system Which is it's puts quite a burden on on the package So if you look at the actual content of what you can download is You have free and open source software. That's their visualization tool 25,000 lines of code and it's it's GPL license because it depends on cute Which is nice and The rest is non-free as I said non-clinical and non-commercial in that case too, and it's a hundred sixty sixty thousand lines of code So it's quite quite a bit and at some point, you know Looking back the past five years It only had binary only stuff that of course That's a no-go and you have to strip it fortunately It wasn't that important and got replaced and as usual they have convenience copies of a lot of stuff And because they have been convenience copied quite a while ago They are outdated and then at some point it was easier to modify the copy than the original and so there was there's diff And you have to take care of that Something Debian right now doesn't yet cope very well with is the data they ship about a gigabyte of data packages which are Not all of them are essential But you know if you if you use it for certain purposes then you need them and there is there's right now right now No way to do it and we're waiting for that data that they've been to work to become available and in 2005 We also had Things that would have to be packaged Before FSL could go into Debian non-free and FSL view into into May So just a quick summary of what what that packaging meant in terms of our Biography So the whole thing started early 2005 and then 2005 in October There were Preliminary packages ready that scientists could actually use in their research, right? So this is about the point where you think it's it's it's valuable and upstream was Quite happy to to see that packaged and they they also agreed to support it, which is it's a very good thing and then one after another the The essential dependencies were packaged Unfortunately by other people then than us which and now the VTK I don't know if you do Visualization stuff in science you probably have heard of VTK that's now an essential library the half of the visualization stuff depends on it and then We we added the essential data format libraries and finally about two years later the first Component of FSL got got uploaded into Debian primarily because at that time Yarek became Debian developer and The the problem of asking for a sponsor for that package wasn't existing anymore and Again a couple of months later FSL got uploaded to Debian and we are now at this point Still waiting for the data packages. So the whole suite even after five years is not Completely usable just using Debian sources because about a gigabyte of information is missing and As well that the regression test suite is packaged, but hasn't been uploaded to Debian because it's it's again quite quite heavy First conclusions that I want to draw from this is that we for us needed to have a separate repository from from Separate from the main Debian archive because we had the package ready for a couple of years And we would not have been able to you know make people use it without having a separate repository Which is quite easy step nowadays. We also use it to provide Backports of everything that we do Backports to stable because that's where what we run in our cluster and all the mission-critical Machines, but also Backports and they're all backwards essentially to Ubuntu because most Debian users run Ubuntu in our field and But we could not stop at that point because otherwise Probably wouldn't have been here because we would have given up by now because you simply cannot cannot do this for a larger number of packages if you have to Do the building and testing and everything all by yourself and by pushing it into Debian you get Access to all the people who do the QA work and they do it for you and and and any package that goes in there It's neuroscientific research software Is no other no no different from any other package get tested the same way and that's it's very Valuable thing and then of course we get Debian infrastructure for free. That's something we could never Reimplement or or even buy ourselves, right? That's that that's why It needs to go in there and of course it's visibility not everybody knows about our secret repository and once it gets into Debian and and Eventually in Ubuntu people will know about it. So after five years It looks pretty much like this is the status of 2005 and now we pretty much have our neuro Debian, right? It's the the system that you can simply install Debian and you can From that minute on start doing neuroimaging research and because we have a lot more support for Purpose packages that are very useful in neuroimaging research. We have batch-curing systems and only one we have a couple of Critical libraries we have all of our we have tons of Python and in addition we have the most important neuroimaging data format supported by by libraries and we have a ton of special purpose neuroimaging research Packages and we have more coming So that's about the state of what we have. Oh, so other than that That's all in Debian and not in our it's the backwards on our repository But this is this is what Debian has right now And this is what it will get but there are licensing issues to figure out and they will eventually take ten years, so who knows and So the second conclusion I want to make is we need we need a tailored presentation It has been mentioned before that you know the Debian archive has a lot, but people just don't know right and there is If somebody is supposed to do research in that field they would ask okay What what do they have for me? Right? They don't They don't just decide I want to use Debian and then you look whatever is Debian has and that's your limited set of what you want to do It's the other way around they know what they want to do or need to do or talk We're told to do and then they look what environment is useful for them and On yet on the other hand they want to know whether there has been just uploaded one time and now it's outdated and Bitrotten at some point, but they all they also want to see that it's actively maintained so there is somebody there that cares for them and Especially the guys who developed that kind of software they want to know whether we are serious right that we acknowledge Necessities in the scientific world they that we properly reference right and allow them to get To justify their their funding right because all of them get paid by some Governmental or whatever agency and they need to document. Okay. Where did that go to right? Is that popular is it used at all do you have two years or two two thousand? So and we've seen that before in in the talk and the Debian pure them pure blends developers have done a great job of Presenting what is in Debian in a you know in a way that is useful for scientific purposes So you have you know if you if you go here you can you can see screenshots You can see the number of box you can see the translations You will find links right that you can click and register Voluntarily right to tell the developers. Yes, I'm using this count me on your list for justification of your grant And you find the the canonical references that you put in your papers if you if sight them We'll talk about this later on and you they will also it will also list packages that that Will become available or are unofficial right now and the reason why they are not yet in them Right if you go to that page you'll you'll see a lot more This is just a sketch, but if you go here, and we can do a little quiz. Where do you have to click to get to that page? Yeah, where just does anybody know? Developer corner he knows because he clicks on it all the time and so so what what you need to click on is here or there right so every user first has to realize that there are developers as well and Then they can click and then need to scroll down a couple of pages And it needs three additional well-placed clicks to get to that page, but you will eventually get there so So it's unnecessarily hard to discover what is in that you know how versatile is but as we all know the Actual information is easily available. We have UDD. We have DDE. We have the pure blends task files maintained by many people So the information is there. It's just not People seems have trouble, you know realizing where it is so But we you can you can easily take that information and generate a tailored view of the devian archive for your purpose and You can do that without adding adding adding any information that is not available within that So what we do is we generate something that is available on the neuro devian net and we give it a fancy name It's called neuro devian people feel home because it's called neuro devian, right? It's it's very easy thing. There is no neuro devian actually because it's just devian plus this repository with backports but people feel really nice and So we take information from UDD and DDE and we have a Twitter account and that's pretty much it and we we link with Portal which you are going to talk about so I'm not going to talk about this and for an individual package It looks like this package description then goes to the upstream home page at first right and then to that other portal to the respective entries and devian met and science and the canonical reference then the register link and then we have we aggregate popularity data from Ubuntu and from devian and from that Neuro imaging tools and resource clearinghouse and then you get a list of all the binary packages that are available People just can't just click they get the sources D entry and then they they are set so and Was that any how good what we've done so people use it This is a plot of over the last Three years pop-con devian you'll see it. So this is the actual pop-con count here. They're 300 So it's not X and not vi right, but there's there's some trend You see that big bump that is what happens if sizable cluster gets installed and has popcorn available and enabled And you see that there's there's general trend, but we as I said, we not only have devian we also have other derivatives, so we need to put that little ink to scale and We then we can we don't have as much Ubuntu data as for for devian But this is the that's the same plot of popcorn values For Ubuntu and what I want to draw your attention to a little piece. That's this little rocket launch there That's what happens if a package that's already Available in devian unstable hits a release in Ubuntu Right, so that this is visibility right people just realize oh look and then they install it and that's that's what happened and devian's last freeze actually is here and So if we if we if you look at the data We we estimate that about 1% of all the people who do neuroimaging stuff of some kind as Implemented as they have some package which only makes sense if you do neuroimaging research Right because we we don't our packages don't get pulled by dependencies of general purpose stuff It's about 1% of the people who submit popcorn to devian, right? And if we ask that new so I find I have to say something about it the neuro neuroimaging tools and resources cleaning house is a portal sponsored by the Europe's government that aims at providing or listing any neuroimaging software project that there is right and we we asked them What is what is what is your the rate of of Linux users and they say 15% of the people who come to our page and download stuff Do that from Linux machines and about 50% of those are devian derivatives, so That's not enough for all the work that went into it, right and Yark is going to talk about How we can change that right one thing I'm here, okay So you saw really first Intermediate conclusions withdraw and What needs to be done so we draw more attention to our project well we need to let people know obviously and We cannot know just within devian Community we need to know we need to let them know at the specific scientific mailing list for those people hang out Right or it's actually useful maybe to present at native conferences were also those native people hang out, right? So we did that we went with our little neurodebin poster to cognitive neuroscience conference not far from here in Montreal and It was quite a nice experience. First of all, it was really pleasant to realize how many people are struggling there and When you describe them, oh, you can install any software for your needs with one command up get install They got really amazed and one lady from Florida She told us her story how she had to hire sysadmin and he enforced macOS 10 solution to her And now everything goes just slow in that map. So We are not only providing seems to be tools We are just letting them even to be assimilated in the field So people could use different tools without any hassle on the same system That's why to deliver the message even to bigger audience. We want to Present ourselves at society from your for neuroscience conference, which happens this fall and Since probably you are not also native audience for that conference. I'll just I just want to let you know that it's around 30,000 people attend that conference. It's huge and there is lots of lots of commercial companies there is lots of agencies from the governments and Hopefully there will be Debian which will be somewhat the president in those terms and We need to Not only let people know that we exist, but we also want to give them easy way to try it right and That's why from NeuroDemon Website we provide a virtual appliance image for which could be used on virtual box and It made people really happy. This is reviews from Nitric in the same website portal it provides ability to review the project and quite a few people were thrilled moreover we've used that image on our little workshop at Dartmouth for Pevine VPA users and We ran it on 16 boxes without pretty much any hassle. There was a little bit with the originating in OS X Itself among the boxes, but otherwise it was really fine There were no literally no slow down in we did real computing and it was really nice And with seamless mode in virtual box it becomes just an experience that you have those tools available on your native system They don't even might not mention that this is running Debian So we let people know and actually let them try But there is a little bit more what we need to worry about we need to work together And this what we realized after all those years and you've heard already success stories when Things start to move whenever you talk to upstream right whenever you find a way to talk to upstream and to persuade them that they can get something from us and We will benefit for sure if they start listening to us. So Let's think about Debian as the city And maybe it's a little bit side-of-a-way artistic one, but still so it's a city. It's a ecosystem which is construct which is constructed by people and Sometimes they achieve really nice results really nice buildings and which are constructed engineered and they're just natively Part of the city Sometimes there is also software It might be state-of-art It might be actually well engineered, but still it's it's really hard to put it into this into this common picture and that's why we struggle again again people to Still be in Debian with all those tools and it's not just the city and the buildings It's us and the software and When we come back to this kind of somewhat ugly looking house and Maybe summarize once again what aspects do we see often in scientific free-and-open source software? Maybe it's not dominant in scientific Free-and-open source software which developed by companies, but it's really common in the Education many projects they are developed by one person or maybe under one person's supervision and then people rotate within a few years, so there is no consistency in the code development or even how the project is organized and Very important also that people often don't realize How they have to expose their projects what license in licensing terms to put what copyright means at all and They don't have power to support the project. They're really eager to present the method So they could be assimilated within the field, but they're not presented well, so methods are not assimilated and Project just dies But we want to present Debian to them. How can we present Debian if let's say some people although the whole article or blog Pass was quite positive for free and open source software, but this sentence stroke me. It's so great It's so useful because if I change few words It becomes our goal to describe how to describe Debian in one sentence. It's great Well, if we attenuate it a little bit more I want to say that its ecosystem because it consists of software and people it's not just software it's not just platform it's us and We have clearly defined standards best of breed capabilities. We have integration We have maintenance infrastructure and actual communities to support it cost-effectively well cost-effectively Usually zero plus some our efforts, but then we also See that those standards may be a little bit heavy on people This is what John Pierce psychopath guy said in some private commerce Communication that they've been packaging is rather more involved. You have to read whole books and they've been private policies Why do you think he thought about this way? Then he pointed me out to Debian new maintenance guide chapter one getting started the right way Well, they've been policy. Sure. You probably need to read Debian policy once in your life regardless either you're involved with Debian developers reference it's a good thing but The appropriateness for starting the right way is already kind of aside. What would be your call for the next one? So it gets a little bit hairy, right? But nevertheless, we do have standards We have constitution social contract and Debian free software guidelines, which are really really important documents and As such especially Debian free software guidelines. They provide as they make Debian more than just a platform distribution they Provide upstreams with legal assurance clearing house. This is how we put it because FTP master is there Just this is their food right to find something which is not properly mentioned. They've been copyright and To get package, right? You need to go through all this interaction with upstream to remove some Not matching code and licensing and it's a lot of work and we are doing it for free pretty much for upstream Moreover, Debian policy assures that we in that the projects they comply with common standards, which are present in the field and Moreover, it provides uniform and robust deployment across different platforms and how? packages behave in their real life and there are even more Standards we have which are maybe not so official. Maybe there are recommendations, but still we have standards and That actually leads people upstream sources sometimes Stream developers to say that. Oh, that's great. We'll try to follow those Debian guidelines when we work on new version of the product So I think it's kind of one another successful story and We do have capabilities. You saw this picture. We have a lot to Provide for users and for developers. We have all kinds of Programming environments. We have different base libraries. They could use and then they don't have to come up pretty much With this hairy structure of their packages where they provide all externals they need for their projects Well, that's another anonymous reviewer They are just thrilled to use it whenever they realized that it's there And integration infrastructure you've already heard about many things and this is what we're delivering for upstream now they can benefit from binary builds for 12 architectures Quite often if project even if it claims that it's architecture independent the code is nice They don't have means to try it across different architectures And if the packaging is right and they provide unit tests We easily assure that if you need test field during during the build on some Interesting architecture we inform upstream upstream never can have such facility in there at home And we provide archiving delivery, which is really important also because we just deliver this software For free across the world multiple mirrors It's really easy for upstream. They don't have to do anything about that And we provide support But it's not only support that we are supporting users We are not there to actually support users in terms that educating them how to use the software We are there to provide proper deployment deployment experts we ship software to people and We take deployment burden away from the upstream. I think it's a huge plus for them They don't have to worry about All these questions on the main lease or back trackers that something went wrong in his system on the user system And we are also Usually people who use this software so we eat the same stuff, right? So we encounter the same bugs quite often, but moreover besides this support we provide to users We are providing support to upstream to some extent this expertise transfer of what we've learned by looking at More than one software product We're looking usually at many software products We realize some common patterns and how code should be organized how it should be installed and we Sometimes no tricks which one specific person from some computing lab might not know So we're providing this exchange of ideas and many Projects in education which come from education. I think they're desperately need that and She even those who are experts in their field They think that even if we provide all those kind of hints and we bust them too much it's still useful and to Kind of absorb this all in in one another Document we started Science deployment guide which gives little hints and we are welcoming contributions and discussion on this regard So we have standards capabilities integration infrastructure and support and people so together it creates this distribution ecosystem in which we have people all those little dots we have upstream development which might be More closely connected to Debian whenever maintainer of the project is also a developer of the project Which is ideal situation and there are some disconnected projects, but as you see by those all lines Those people who deal with Debian through the maintainer they don't have to care about deployment problems which are deployment is blue, right and They just have to deal with those problems with other operating systems just Nice picture to kind of put everything together that we have also different teams within this ecosystem and we interact Among each other and with projects So conclusion Debian is the ultimate platform for new image in research because it got electrified Well, that's wrong message probably maybe in 200 years, I think that was that will be about right, but the main conclusion probably will be the same sentence that we do provide this robust system and actual ex system of us developers and all the software which we provide and Debian is so rich in versatile and if you care you can make it tailored towards specific domain like we did with neuro image and neuro Debian portal and It provides really robust deployment People this is what also aspect is important for scientific research that You need a stable system. You don't want to run unstable. That's why part of our kind of duties when we package something is to make packaging Easily or transparently back and backportable. So we don't need to patch it to make backport So we just can build it for many systems. They've been derived systems without any hassle and that creates really really nice system Or the core is really stable. It's Debian stable. Everybody knows that it's stable and just scientific software it's bleeding edge, but it's there in a stable environment and We try to work with upstream So what could be better it's ideal platform, but we need to do a little bit more We need to have better coverage of neuroscience software at the moment We are concentrated on this neuro imaging, but there is resilience of other tools Which are used in other fields of neuroscience research Although we have targeted web presence on our website. I still think that we still think that it's Would be beneficial for Debian itself to come up with some design of an appearance on the web or it Becomes more prominent that we didn't have to go through three links to get something of interest for you and release notes for every let's say even task in Blants we can easily craft release notes for every upcoming devian release which are tailored for specific domain of interest So people know what kind of new software comes into Debian for their interests That's another aspect we need to worry about sometimes People need to have multiple versions of the same software installed Because maybe they analyze data with previous version and they want to replicate the results Or maybe they want to replicate somebody's else results, which were done with the same version so that is also aspect which we want to keep in mind and The software FSL which you've seen already it's on the way to come up with version multiple versions packaging although it's such a humongous one, but you will be able to install multiple versions once and alternate between them and Everybody already referenced that we need Ways on how to reference the board so for that come back in 10 minutes when we have next session and means for reproducible research either we provide some version control facilities or we just Advocate on how to use virtual machines and snapshotting. That's the easy solution They could do and it works the snapshot the system when they publish the paper later on they need to reproduce it's there they just take that snapshot and large data packages once again and extended quality assurance We don't have any Infrastructure to do regression or heavy testing we try to put some testing while building But it's not solution for let's say FSL with its huge battery and it would be great to write it on multiple architectures and of course we need further to Disseminate Debian around the ball and if you want to contact us and we want to express our Gratitude to other teams within Debian and our advisors Steven Hansen James Hexpen Steven Paul and Thanks a lot for a great talk. I think it's really great that you won't go out and Two conferences and then push that stuff. That's really something one so far once so far But apparently we'll go for the second time and I think that's that's really that's really a great thing which Really will help in the long term any questions Once Can you stand up, please? So you're mostly neuroscientists or did you also get to become Debian? developer or the Debian maintainer Or you are just Debian user and do a lot in promoting Debian in your Environment, I'll just answer for us both We have different stories entirely Michael is psychologist. I'm computer science by major and we do neuroimaging so we just also not psychology not computer science and Michael is a new maintainer in new maintainer queue, but He's bored with that stuff. Sorry. So and he has me All this stuff goes through me into Debian anyway You're a Debian developer. Oh, yes. Yeah, just just to make sure So I'm in new maintainer since middle of 2007 But there wasn't much time left other than that to get through it eventually maybe Further questions. Yes Yes, I was just curious about your use of virtual box and VM snapshotting. It seems to have lots of applications and whether or not you would use that to maybe not circumvent but Increase the freshness of what's in Debian That is would you consider snapshotting off of testing or something like that with versions of your latest versions of your packages? That's so you mean snapshots of testing distribution what we have which is Going to be a booth, right? We were talking about snapshotting. It's snapshotting of your complete system what you are working with now So whenever you have installed all the packages and data Sorry, I just meant to make the distribution available to other researchers Essentially a blend that is it's not based on Debian stable, but instead a snapshot based on What we what we do is That virtual machine We do nothing to it We add other than adding the line to the source of this and run up get update for the first time And other than that is playing Debian stable and since it runs in this virtual environment There's really nothing that it couldn't do right and and people just download it and then they click on the Seamless mode that it vanishes and they have macOS and they're maced and they run their software that they need and they have no Clue how it works internally the machines are so powerful that you don't have to tell them right and and we found that the Performance is really for most of the things exactly the same so as he mentioned we we ran a course on on Multivari pattern analysis of neuro data Based on a macOS computer lab running Debbie and you really don't have to do anything. It's it's quite versatile You have nothing to do Okay, so the freshness of your software is controlled by you adding a line to app to the sources list Oh, you know what what we do in addition obviously is we we backpull all the stuff that we upload to unstable Also for Len and we build it there and it's in that repository and if eventually You know getting into backports of org will become easier for people like me or Debbie maintenance Then we can also do it there, but right now it would all go through him and then he doesn't see his kids anymore So you should become a Debbie developer. Yeah, so yeah, it's my fault For the questions I have a comment. Oh, yeah So have a comment so everything as a upstream guy If if any of you here are also Source developers you want your to package your stuff. He is he is very good Oh Bunch of debuggers one for bash What's it one for can you make? I don't know what are they? What languages do you know? I Just wanted to comment on two of the things that you said to describe this future development or or Sure future future development or things would be nice to have One is installing multiple versions. There's already the update alternatives I'm sure you're aware of this. Is that something that you just haven't tried or just not or it doesn't meet your needs in some way and The other is testing there is as you identified no real kind of testing or Regression testing and packages infrastructure and Debbie and right now Wonder if we could talk about that a little bit or if you have any ideas so Update update alternatives is nice and if you if you use it for a few tools, then it's it's also convenient But for things like F cell which really have lots of interdependencies with lots of things You can't just combine one version with anything else So we need to have a way to have them available at the same time with their in their respective environments And then usually what people do is they they also rely on Reperscripts that set up specific environment for that specific version with another configuration. It's Complex and then there are things that are You know Packaged by upstream in a way that you don't even know what version you're running and if they release a new version They don't change the version and you wouldn't know unless you figure out You know what what they actually changed and and that's really hard to package that in a way that you can support that if upstream doesn't want to and Sometimes they just don't want to although people want to use it that way So we we can go as far as as we can do it with upstream because at some point the patch size is so large that you can You know Responsibly do this anymore because most of the things as else the rare exception They don't have regression tests and we repeatedly run into the issue that we do good in the devian sense I think we remove convenience copies and link it in a slightly different way, right? and then then the t distribution doesn't look like t distribution anymore and then Then you know some hat needs to be chopped off if that gets published and it's not upstream set because their software works, right? So and so you need to you know have a cutoff where you say okay, then doesn't go into that Because they don't want it and and if we succeed and that becomes as popular You know in in that specific special interest field Then we hope to generate another enough of interest of upstream people to get into it And and even with that single presentation at that native conference It seems to already have get generated a lot of interest so people now We always came begging and say oh wouldn't it be nice? And now they come and say oh I have this software It's been downloaded 80,000 times from the NIT of a nice here see the box and they say Can you can we get that in there somehow and sure it can and then they're motivated enough? You can you give them to do this what what they need to do to comply to the 50 standards And then they just do it and you get it into that easy as that If if they you know have interest I got a follow-up question that I mean Is it very difficult for them because I have the experience or I have the feeling that a lot of that research software stuff Sometimes the copyright isn't very clear You mean there is like various people from various groups collaborating on it And then I've seen a couple of really good in my field software popping up lately Which have a GPL tech on it. It seems to be People do it by default now, but I mean I know that there was various professors from various groups And I don't believe that everybody really signed on the GPL Maybe then somebody had the code and just stick it on is that something that you can enforce or is it easier in your case? Because there's one institution which retains the copyright and they can just change it or how's that working? Or is it are you not looking that hard? Well, we have to be hard, right? the the thing is there is one package which we're trying to package since four years now and they're working on the license and They're clearly violating everything that we know about free software They it's it's it was closed source now It's open source, but the license permits you even from downloading prior to agreeing to something and they use GPL software inside So if we would be of the the neuroscience busy box, we would be very busy in court Right, but it's it that doesn't make any sense at least in this field It's a couple of people who develop software and it's it's not them who make those decisions It's the licensing agencies of their universities who pretty much have no clue and and and so they have to Spend their valuable time, which they cannot do research in to try to figure out, right? And but at some point a lot of them have sold their souls to this University administrative layer and it's probably social work to feel it out court And the second question for me Is there a new year's neuroimaging task already or how did you collaborate with Andreas for that or are you just using we have a Few let's put it this way first of all we have that's the problem somewhat of Problem of tasks in blends that they're not orthogonal some software could be present many of them, right? Yes, so we can and put in a new one for neuroimaging specifically It also doesn't make sense although there is one there is imaging and imaging deaf in they've been met There is also cognitive neuroscience. No cognitive science or cognitive neuroscience in Debian Yes And that one is Debian science Okay, I think what would be really nice is just one second because one problem with the tasks so far is that Once it would be really nice if you could say okay Neurodebian net is these four tasks and then you they could just collapse them and look at these Tasks of stuff because for example now, I think the biology task is basically just Debian met bio and it's just one Met a package so it doesn't really make sense to click on that So there is some problems, but I mean one thing that I mean Andreas told me that for example for Debbie came we now finally have the tasks in So and he said well we could also put it on a release notes when when squeeze release once it's in So that might be also something interesting for you just having your stamp on and maybe getting quite some exposure On the release notes if I don't know whether Andreas will actually manage to do that, but it's something for you to consider I would say the the thing is that so we have all those tasks and that the web page You've seen that's generated by few lines of Python code It just gets them from SVN It get it talks to UDT through DDE and ask for the bits that that we need so and we have that tail of you Doing the same thing in the in the with the tasks interface would mean we would pretty much copy half of it and maintain it in addition Right, and then it would be hidden somewhere behind and now and of course we could do that with neurodebian net some And then add to that but that would mean we would need to Make it talk to our repository as well Right and that's right because it lists also that and and also the stuff that that can never go into Debian How are you actually are using maybe I missed that are using subversion.debian.org or git.debian.go from maintaining everything all the packaging The website code is all an alias Okay, right, so then I think it's no problem for the tasks thing to just take that but okay. There was some more yet What's that another question? Sorry I've got a question you talked about the virtual machines that people are using So are the users are happy running this stuff in a virtual machine? Or do you think that a sort of a life city would attract more people so they can just Try it. Well, they can easily try it with your day with virtual machine, right? If Some people had difficulty running it for some reason then they just installed Debian in virtual machine and Then put one line in APT and they were happily users since after We were going to do little research to trying to figure out either someone migrated from virtual machine Into native running Debian, but we haven't done that yet But many people quite satisfied some people have a not pleasant experience with Running on Vista or some new 64 bit seven into seven, right? But in general, they are very happy and we have the situation that the people who use the stuff do the stuff Our PhD students and they have no choice what they what they should run their stuff on even if they know that David is the best They need to struggle with you know red hat or got for big windows, right? And then those those go crazy if they download the virtual machine and then it just does it whatever they want and and If they go into seamless mode with a virtual box They don't even you know think about that you know it for a week and then if they just use it There is no initial resistance sometimes What I want to say, I think it's a very very good example that you make a pass between the Debian development and the users and I think a lot of Parts or people inside Debian could learn from you that we can do things much better like the webpage I think it's really great the example can you find where to a certain part of software and your web pages much more comfortable for all the users and I think it's also very important to have people inside Debian that have contact to the Developers, but also to the users so thank you for this project And I hope a lot of people will have a look at the video of the talk and more people in different Topics would do the same work as you do. Thanks I wanted to comment on the issue live CD against virtual box. I have some experience on giving Python schools and We were using a live CD system for the students and but now I am We decided to switch to virtual box because with the live CD of course you typically have the for example the firmware problem for the wireless Cards and this kind of stuff so you for the average laptop You're pretty sure it will work and they will be will like Debian They maybe will decide to switch at some point But the virtual box is just much less invasive and you are pretty sure that it will work everywhere and this is The final selling point Okay, then I think we should leave it at this we can still discuss later for things I already noted some stuff we might discuss in a round table from your talk and We were back in like five minutes with the new developments in Debian packaging Lightning sort of talks