 Welcome everybody to the science track and the first talk of the track will be by Andreas Tiller on Debian science. Yeah thank you. Well I was recently asked how many depth cons I attended I was counting and this is the 11th depth cons I'm attending and it's really nice and in 2003 I was the first time talking about something like blends it was not named like this and I was always addressed as a man who was talking about blends and science and that this step comes the first time I'm addressed at somebody else I'm addressed as a man who's taking the photos and this is kind of a different kind of getting famous. So what I want to tell you today it's about the current status of Debian science and I would like to explain my ideas how we could get even better support of sciences by using the blends techniques even better and I would also like to discuss some problems we have in Debian science which are coming up from time to time. Well the history is like I told you I was on several depth cons I also visited depth con five in Finland and it was but maybe the start was by given by Helen Faulkner in a talk which ended up we create a mailing list where we discuss topics of scientists inside Debian. Then some competing packaging teams evolved to some people started with packaging scientific computation and so on we have different teams and was not really clear who is who and who is packaging something else and so the teams finally were merged in quite an effort and some flamish discussions and this was more or less ended in three years ago and we are using this blends framework which was invented for Debian Edu Debian made and so since 2008 who is familiar with it with the term blends please rise your hands okay who has not raised his hand okay that's good so we have a complete set now and well a blend is something which is definitely completely inside Debian it's no derivative the funny thing is always if I talk about blends and I have these slides which is in bold face no derivative in the end I get the question why are you doing a Debian derivative to do science no I don't I never had an idea because I'm personally definitely convinced that if we do things inside Debian if we adopt Debian itself to a specific purpose then we have in the end way more less work on the long run in the short term it's very easy to take Debian add some scientific stuff and be done that's a quick and dirty approach a blend is simply going over all the hurdles you might face in Debian make it properly and have it done on the long run so it's about forming a packaging team about around the topic well derivatives are frequently done by a single person and this single person has usually quite low running over by both factor in the best case this person person is has children having children is cutting quite a lot from your spare time and this person will not be able to continue with the derivative and the derivative died I've seen this with so many derivatives in the field I'm observing quite closely which is biology and medicine so you have countless derivatives of Debian it's a most derived distribution but a blend is doing a derivative the right way and don't derive but do it inside Debian we have developed some tools it's kind of advertising the topic that we are doing it so you can tell the world Debian is doing something about science and we have also tools which are helping to do some quality assurance and science packages it's called the web sentinel I will show parts of the web sentinel depending how much time we have depending how much question you have and this should help on advertising and quality ensuring bug fixing and so on in my eyes Debian science is in so far a specific plan because science is not a defined topic right so many people can define themselves as scientists and the fields of science are very wide so in principle my idea about Debian science is it is only an umbrella to be a spin offs to have smaller topics like we had for instance one famous offspring is Debian Astro there were some astronomers inside Debian who are working in the Debian science team at at one point in time after I was nagging a lot please do it yourself they created an own blend and this is not very successful they have gathered a team of 10 15 people who are working closely on the topic of astronomy this would not have happened if they would not have formed a specific team because astronomers astronomers all over the world use Debian but are not aware that they can bring in their own work into Debian by working with the Debian Astro team well for a brand successful blend you need a critical mass on one hand of packages and on the other hand of maintainers doing the work and it's a very good thing if they're not only one maintainer works on one package but at least two or three maintainers and this is also kind of a team building inside a plant and this blends framework should help to to create this critical mass if you see or we have this list of packages this is interesting this one is missing I can do something to fill this missing gap and I joined the team that's basically the idea what specific plans do we have as I said I'm working for Debian made I actually initiated this project and I started with all these techniques which I also adapted from other plans like Debian Edu and Debian Gis we have also Debian Gis this is about geographical information systems we have Debian for chemistry they existed in parallel to Debian science and we are not really offsprings but I have said Debian Astro is a real offspring from Debian science which really followed the idea I have so what I would like to see is a list of 10 more of these plans say Debian mathematics physics electronics robotics whatever we have but currently there are not sufficient there's not a sufficient amount of maintenance who could form a team you should start maybe with three or four people who are running this strongly and then the other people will come and and join you at least this happened so why it's not more as I said the idea of Debian pure blends leads yeah there's a question can you please use a microphone no no it's for the video recording there's a microphone I just care about I mean I only care about that but I particularly care about neurodebian neurodebian yeah is it one of the blends like that is the same as all the others well I I think neurodebian is a potential blend they they are working together in the Debian science team they have meta packages in Debian science they are working also together with the Debian mid team but I think neurodebian is a perfect topic which is a kind of close topic around but it's only maintained by two very very active people it's michael hanker and jarl slav halchenko they are doing a lot of work but in my opinion they should try to gather more people and use the framework themselves to advertise their stuff there they developed interesting techniques which are different than what we are using we should merge them into the blends framework for instance they have a kind of back porting system they back porting for every Debian release and Ubuntu release this is quite cool stuff but it's not really integrated and so a blend is integrated and I talked to them several times please do so and yes if you are more conventional than me then I would be really happy so this was the answer to the question other questions so I continue the advantages are very widely unknown this is at least as I told you this is one effect why the blends topic never goes in the main hall because if other developers would know about how to organize their team or their topic in a blend I think it would be more known even if I'm talking about it and I have a video proof this is Ashish LaRoya many people will know him he quoted or there's a quote in the video at about minute 33 we should try hard to run around asking users and developers is there a topic you are care about and create a blend today this was in 2013 and they asked him 2015 did you ever created a blend oh no I didn't and yes but it it's work actually right you need to do some work there was some time when I even did the work for other teams for instance I remember I think it was in Edinburgh or so for the Debian games team where I when there was a Debian games talk running and I was sitting in the talk and I was creating all the meta package we are using and I was finished in 45 minutes and presented them the work and they grabbed it up two years later and now there's also this framework for Debian games while I was mining two years is this it's no time well we are developing slowly this is one release cycle so it's it's not bad it it's good if they do it it's good one release cycle is an appropriate time and you need people who are just doing the work but if you do it you see more people are coming and I have well I'm doing team matrix graphics and you see that if people are continue to use it more people are coming I can prove this statement on some graphs I can show them later if you want and the idea to come back to Debian science to to be an umbrella should be more more better advertised from time to time I'm posting to the Debian science mailing list please while you are some people who are working for say robotics you could do your own plan next next question so I guess one point where I haven't been sort of motivated to to really participate in this is and I think I'm maybe you know could be disconnected from from the target audience but as I sometimes think of myself as a mathematician and the idea of meta package for all of mathematics doesn't even really make sense to me so can you I mean do you have feedback from various user groups that these meta packages are really useful or I mean people tend to be focused on one or two applications that they need for their work I don't know so I it seems like maybe you want to sell me on this one set well I agree with you that sometimes a hard decision how to categorize packages this can only be done by in the example of mathematicians by mathematicians and it's always kind of yeah gut feeling how to do it in general the task of the meta package are overloaded by programs you will not use but who cares about this place these days and you can be sure that that you really install everything what what's covering the topic if it's more or less designed in a deviant meet we make it very very simple we put close to everything which is biologically related into one meta package and it's installed you have lots of package which are not really needed but for the for the first installation we have in my institute we have 50 users who are doing we are using one machine they have are using quite different programs and once it's once you install mid bio you are sure everything's installed for them this is not always the application you might have in mind because you are running your own machine and you know what you need but it's for people who don't know and also if I'm going to do to this page this is the biology task it's also for advertising what we have so if you want to explain any mathematician what kind of mathematics programs are inside Debian and you point them to say it takes a deviant science topic mathematics and you can show what's what's inside Debian I don't have no idea how many packages is and if you get at more structure to this very general mathematics that it's quite long here I don't know you have a lot of packages right and for instance I think this one this GM few is just an education tool you don't need in numerical analysis is also yeah so you would find some better structure and you can use these pages as an advertising of what we have so these pages are answering the question what you are doing in Debian what can I expect from Debian it's not only about installing this it's also about showing what we have or you see we have screenshots and also a very good argument I think it's going back to the Debian main thing we have citations here so if you want to approach upstream I want to package your your nice software or upstream is not cooperating it finds not setting any text which are helping you and all this not willing to it's a license they said well what have you what's our offer to you we provide your publication amongst other one in a kind of a structure's context he sees application close to other relevant programs he sees how many users are using this program you can see a screenshot you can have translation we have also some upstreams have some registration for their wrong to know how many people are using the we can also link this here if we have some metadata we can put it here we can the thing is that with a blend you can go from a single package few to a group of package few with all the metadata connected in one place and I think this is this is kind of an advantage which is not only the installation side but only the inside to provide proper information and attract people and connect people like developers and users it is a proper answer to your question so I think yeah there's a few important points that you mentioned and I think to stress I mean to motivate developers to stress what's the value beyond the package descriptions which already exist I mean we're mostly pretty expert at finding what we want in Debian and and to point out you know some of this sort of curation yeah yeah it's a curation yeah you have the string search in up get search or get searched or do you have an index also but the string search is not always helpful this is a curated list I mean I mean one thing we all want to avoid is just work which doesn't have obvious benefits right so we should really stress yeah to you know compile this metadata is an extra task yeah to maintain these tasks is an extra task yeah somehow we need to technically the list is generated only from package names you just need package names so the task list has package names and the framework extracts all these information from UDD so this is done automatically okay next question my question is is there a concept of a popcorn for the blend itself for the meta package itself you know what correct yeah but well probably yes it's it's a normal package shows we have popcorn but I never minded about today oh sorry minded about this data yeah it's good question yeah yeah we could go at this maybe I should make a mental note yes I would also love to to have some popcorn data maybe from Ubuntu also because we have a lot of derivatives who are using these packages and maybe it will be also some extra information I do not really want to advertise Ubuntu here on this page but it's it's some relevant information about the package do you know something it could be interesting thinking about the popcorn and similar to see I don't think there's a way to do it but to correlate who has a how many cases of having a high percentage of one blend cases installed versus how many times the blend itself as a full yeah maybe I'm I've not not thought about this idea but strange to put it this way because what one of the issues I was thinking maybe is that maybe a blend installs say you were this is about bio but biology spent so many pack a soup fields yeah many people would want one two three of the packages not not all of it I fully agree well my point is I was nagging my fellow Debian made members to do the categorization I'm a physicist right I'm just not competent to do more fine grained categorization I would love if somebody would do it and there is some work on gay going and thinking and but it also will take us one or more release cycle so ask me again in four years and then maybe I even thought about splitting Debian made and Debian bio because this could force this categorization my original idea was inspired by the critical mass of developers and packages to put both together because bio is very closely related to medicine I'm working in a medical institute but we are basically using the biological packages and no practice management so this had some advantage in the beginning and we never went away from this but you are fully right I also what not really question but a command I mean this is web-based and now we have the situation that newer desktops like gnome don't really explore what you don't present the menu that we used to have so back then in the gnome 2 setting there was this applications menu and then you would have science and people would be easy to discover actually the applications which are on which are installed in the tests and now with gnome it's a bit more difficult you just have to type stuff in and it shows up so is there any way to get this overview on the desktop in a sense starting either a creature or maybe packaging it locally as a package inside the plan well no well I think the answer is in the beginning when I started writing the framework I was writing some blends own menu system with a devian menu this is menus but it's not used anymore right so this should be ported to a free desktop and then the problem would be solved then I guess you have the problem that these menus are not really exposed anymore right yeah okay you have these applications and if you if you type the windows button on a gnome screen then you just have an input thing and this is all not very discoverable so yeah but it's a gnome problem I think well yeah but that's I don't do that I don't want to make a flame but you should think about the standard desktop yeah no no I've just tested but I've seen you can integrate the gnome software center into that search so you get packages that are not installed so you search for a package I don't know if it's really helpful to do a full text search there but maybe that's something maybe if you just use the text so that is maybe some packaging tasks that gnome package maintainers can look into yeah well I think to some extent gnome is using the free desktop org data and we should should also rely on this standard and the implementation is not our problem because we are doing package categorization and not desktop at least we have no manpower to do this but my point is that additionally it would be useful to have something for users which is discoverable so for example Enrico showed me a tool which is kind of like a small gooey with buttons on it for all the packages which I think in that case was a specific tag because he was doing it prototyping depth tags so there could be this one central place where you could launch all the packages from one task from a small gooey and you just help people oh run this tool and then you can see yeah yeah biological things and or I think really we should try to package offline the website and just have it as a starting screen in Firefox or something or at least something which the people have on the on the hard disk and because those websites are beautiful but you people have to actually find them yeah yeah yeah that's correct yeah I tell them and I'm using them because yeah but but you have yeah yeah yeah yeah and I point some sometimes people to it too because you see the translate description this everything with this yellow means work so you can click on this and then you go to a page where you can translate this description and if you have a translated one then this then you can say here below this is translated I have German look here and then you can only fix the translation but the yellow button is vanished because it's done and here you can also use depth tags so I can use this one or edit depth tags if you have so this is some interactive to some extent it's not JavaScript and so it's simple but it should show where users can do some work and by the way what I forgot regarding popcorn the additional information with popcorn provides if you have very similar programs you can maybe sort according to popcorn and see are the most people are using this of all the similar things and so this is also information I think I continue from here okay so as I said the set blends are providing some kind of a straightforward access to the package pool and you have no real need to search for interesting packages because it's just collected for you and we try it to get some engagement to package more free scientific software for instance you could you could do some menus also it's also on my slides or you could do some install off or so you use it normal day we are an installer and then you get a selection list so you want to be on a duty and science they were made or so we were very very close to get this implemented this idea from 2003 for stretch and then somebody said well you implemented this wrongly and it was dropped the feature so we were very close to say install David and science yes but we need another release cycle at least very sad anyway so the web sent in a I have shown you and I can show you the box overview for instance for science mathematics as also whoops I should go today and science so this box overview shows you the following there is some kind of logic behind this this color scheme so in red highlighted tasks you find a lot of bugs in the green yeah it's orange orange reddish and green is nearly nothing and then yellow so so we check just economics no idea what it is and you see here also some color scheme there the more reddish ones are the release critical bugs and the yellowish ones are the important bugs and the green ones are not so there's a legend on top so you can see with this color scheme you can lead people working on some field to this page and if they are interested to help you then they have quite a different few on the bug tracking system than they have otherwise so they find the bugs they might be interesting for them this is the idea of this box overview and we have this so-called thermometer it's it's kind of a list which tells you in what release it's what version and in in backboards and so you can look at it on later on the web and we have this maintainer statistics I have one one example on the slide so you can see it later here so my point is that Damian should more actively reach out to upstream not only taking taking some software and do it but actively reach out to upstream for instance user free license user versioning and something else and though it makes a definitely difference if you are a random person like Andreas stiller who's Andreas stiller I never heard about this guy or I'm writing to you as a member of the Damian science team okay there is a team Damian I've heard about Damian and they care for science is it's totally different position if you approach upstream and they the chances that they respond to you are just higher so as I said we advertise upstream publications we can link to their regular registrations provide popcorn statistics and so we give credit to upstream it's important because well scientists are yeah sometimes a little bit focused on their very specific stuff and don't care around the world around that's not bad it's helpful for their science but to to get the connection between the upstream and the user we are the missing link do you track website hits for the web sentinel I mean so you can tell upstream okay like 20,000 people looked at the no I don't track this no I don't taxes so I realized that the screen is not adjusted properly so we I'm just have a list of some specific meta packages a list is a little bit longer than here my slides do not fit it doesn't matter we have these meta packages in science it's you see for instance brain computer interface is quite different from chemistry or electrophysiology so we have a lot I also don't know what financial engineering is we can look at the task so I just made a list okay and we have another it's it's about 30 30 tasks it could be more because there are more sciences but these are the tasks somebody is caring about this is just a point this is not a very strict selection what we have because we have nobody who really can curate the full list as I said I'm not competent what I'm doing is sometimes I'm checking what package is maintained by a member of the deviant science team according to the the maintain of the package and if the package is listed in our task I have a UDD's query and before we are releasing I'm sending an email readers of deviant science have seen this kind of myth please categorize your package because you know in what category it would fit and one package can fit in more than one category it's not exclusive definitely not and I want people to do this and we have some common science utilities like this data acquisition accessing devices distributed computation and presentation with always with science in mind right it's not the general presentation tools and so this is quite helpful and we have also according development tasks in principle every task could be could have a development task which is basically the development libraries Python modules and so on and some sciences defined this and others didn't so it's it's completely self-organized this task creation it's I have no better solution for this because otherwise one person needs to know the whole package pool and my focus is not really deviant science but more deviant made so I can't solve this so the problem is that many developers are not really aware that deviant science exists so single maintainers are ITP in scientific software and I'm checking all the ITPs and I redirect them to deviant science list if you're reading the list you will see from time to time please join the deviant science team with your ITP and it is frequent that some single maintainers also frequent but it just happens that single maintainers don't care for their packages and they are featuring our C-bugs and they either affect scientific packages because these are pre-dependencies or there are scientific packages but not maintained by the deviant science team and so what I'm doing is pinging the maintainer and if he doesn't respond I'm team hijacked the packages team hijacked means well I would like to put your package into deviant science team and if I don't hear from you after two weeks I will put I will do the work and do it it's well it's conflicting with deviant policies because because we have this maintainer ownership of the package but nobody blame nobody of the maintainers blame me about this I had some some package which was before the Jesse release I had 24 hours to save it into the deviant mid team and I well I set the limits if you don't if I don't hear you after 24 hours I will do this and I did it saved the package for Jesse never heard from the maintainer but some some third person blame me well you should not do this it's hijacking I agree I should not do it but it was helpful for the users in the end so what should I say and developers don't really care for the tasks and this is a UDD query I'm I'm doing usually to see how there's an scientific package this is not categorized and yeah I do some suggestion something I can do myself but yeah some maintainers respond and some don't that's volunteer based right so this is no it's stupid and next talk I need to adjust this better I was I'm badly prepared the thing is we have a graph from deviant science team please assume numbers one two three four here and it is showing this 420 packages are have commits only from a single maintainer and we have 280 packages which are touched by two maintainers we have three and this graph is somehow hyperbolic style so this means is it's a bad team I show you a good team this is Pikachu pearl they have I can tell you 150 packages only by one maintainer 400 by two maintainers and close to 900 which are maintained by at least three persons and this is a good team because a lot of people are caring David microphone I'm sorry for for the bad display I need to adjust it somehow so maybe I missed this but how do you gather this data are you using uploaders field or change law this is a version control system I'm I'm passing git for of some teams not of all teams so and this is from it okay thanks and for deviant science from git and S4n because both is used another question you wasn't okay yeah so this is all around this plan stuff and then team matrix I'm I'm gathering different sources of information mailing this UDD uploads version control system and try to do some relations one last hint because in the deviant mid team we are not very not really here but we have also quite many package which are have two maintainers but I know from my own work that there was formally one maintainer and then I took over so this second maintainers only well just me and so it's in the end only a single maintainer so this graph is not only not not always right but like to show the pattern you can see all the graphs and I'm linking every year to this in the deviant deval list and deviant project list but I can you can ask me and I'll show you all the graphs of all the teams I have so the problem with scientific software is that you have frequently short-term contribution by scientists because scientists frequently driven by projects with a limited time frame to three years or so and then even upstream vanish and you need to upgrade package or the program and adapt to new GCC versions which is currently up to date so we have also quite specific software so there was a lot of efforts to package some software from developed at CERN which was done by a single maintainer and this single maintainer now is also not available anymore it was some Chinese name and I forgot the name but we had done a lot of work and we need to make sure in the deviant science team that this work is not lost and save these packages even if it's complex but it was really good yeah and sometimes you have old libraries for turn and so on yeah so the point is also the solution would be that we should try harder to get more than one uploader this is definitely goal and I think one way to do it is to involve more upstreams from the software this would be really cool so then we have the effect of so-called team or fan package whoops we need to sell wage these packages I mean and the thing with this team or fan package so what I mean is we have as maintainer the mailing list and the uploaders are a number of real persons and if none of these person is not active anymore the packages de facto offend so but it's very easier to to do a team upload than a non-maintener upload non-maintener uploads always you should not do it ourselves the barrier and the team you just fire up dth minus minus team and you upload I have also to make the the concept of plans more known and invented them concept of sponsoring of plans so I have a wiki page and set up said if you need a sponsor so sponsoring concept is known you are not a deviant developers and are not able to upload I'm deviant developer I have the power to upload I can help you but I will only help you if you if you show that you understood the concept and if your package is inside the task packages and you have read the documentation of the blend and it shows up on the web sentinel this is my requirement to sponsor your package and so I was able to help others because usually newcomers seeking desperately for a sponsor and I provided and so we can kill two birds with one stone the package gets sponsored and the sponzey knows the blends concept this was quite well received in the deviant gist team it helped a lot of quite active deviant science members who hopefully will be soon deviant developers and save me the work and yeah there's a wiki page okay and the last question is is deviant science doing a bet Q a I think the problem is we have 1000 per packages and you always find 50 packages with bugs because it's a lot so and we are frequently blamed that's not deal with this back bugs but yeah it's just because you are a lot and if you are really bored by the bug this is the solution these age minus team do a team upload and this as I said the barrier is way lower than an NMU so anybody who's blaming us about this can do something so last slide what we if you want to do some maintenance of scientific software just contact us and it would be better to do it inside the team and if you find you found some these maintaining similar software like you you can leave the umbrella of deviant made and do is your separate plan this is for me it's kind of important point of this talk yeah so you find the talk if you Google for Andreas Taylor talks it is written here sorry for that it's at people that they be an org till the tiller at talks and then you can find this slides and I will enhance them a little bit so I think I've done some questions thank you here are there for the questions questions in the in the middle it's I like this because it's more lively so there was one question from IRC can you install two blends easy geographic at geography and geology because there is a set of yeah I can answer shot yes you can you can install deviant games deviant junior deviant science on one machine it's just meta packages and per definition packages who are optional should not conflict to each other so yes I have not prepared a lot of slides for the science buff it's just about discussing bringing in your ideas and right so there's a lunch break now and we'll reconvene with a talk on bioinformatics at 2 30 I guess so yeah thanks again Andreas and see you later yeah see you