 Well, it's not really round but I hope some more people will attend and I do not want to make a big presentation here just to get some kickstart with some short slides what Deviant Science is and you might have seen the statistics about the mailing list because Deviant Science organization will happen basically over this Deviant Science mailing list and you see the effort started in 2005 and until 2007 there was not so much discussion or engagement on this mailing list but I tried to apply this Deviant Blend stuff to this Deviant Science effort in 2008 and you see an increase of the discussion and I think an increase of the activity in this mailing list and we also applied some group maintenance for packages because we think that there are several scientific packages which are complex and should be rather maintained by more people and so we have also implemented an Allioten project where you see you can here you see the commits to this mailing list there's some very active committer Stefan I just get rid of the last names because some people told me that this is about yeah data safety but whatever you know the people from the mailing list if you are involved in it and here we have some discussion of the maintainers so the names are those people who are active on this list and I hope we get even more in this concern. Deviant Science has developed some so-called meta packages which somehow fit to sciences we are interested in. You can see a list in my opinions Deviant Science is something like an umbrella or a chance to kick off for specific science blends so if you see the biology part that is basically covered by Deviant Made because there is a very strong connection to biology we have Deviant Chemistry it is basically covered by the work of the DB Chem project and there is even something working for geography the Deviant Grass project or Deviant Giss which is working on geographical software in general and my great hope is that we also get people who are actively working for Deviant Physics or Deviant Mathematics project or so that we can profit from the work of the Deviant Science project work together with them and just get more focus and detailed sciences but currently we have not so many people who are actually running such a project and we have not yet the manpower as I'm seeing here Deviant Neuroscience cognitive is also some effort of part of Deviant Made people and we have some common science utilities like utilities for data acquisition, image analysis, statistics, typesetting and viewing these are utilities we need for general in science and I hope that the specific sciences just came back and use this specific stuff and we have also some development packages in where people just found it useful to sort out some application which are only needed for developing the scientific application but not really for the users to just run a certain application and do not want the development. Here are some interesting URLs for those who are not knowing what Deviant Science might be and I would like to ask you at first about your opinions of the Deviant Science project and perhaps everybody can give a short statement I would like to head around the mic and if something remains unclear here in the slides it was intended to people who are just know what Deviant Science is in principle if it's something unclear I'm happy to make things clear and explain some more. Could you take the mic? It would probably also be useful if people who are involved in Deviant Science or even maintain science related packages outside of Deviant Science could indicate what packages in general don't give me a list but like what area of packages or science in particular they're interested in. Sometimes it's not obvious. That's right. So Justin could you keep the mic and just introduce yourself? Okay my name is Don Armstrong I maintain a few partially related science packages in Deviant but I maintain a whole slew of them outside of Deviant currently primarily related to bioinformatics and other biology related subjects specifically a whole slew of R packages so that's what I do that's related to Deviant Science. Any reason why you are keeping them outside? The main reason why I haven't brought them inside to Deviant yet is primarily a factor of time. They aren't yet of quality that I would be willing to see distributed in the archive so I mean I do have them there but I basically packaged them just enough so that they work for me. But it would be interesting if there's a package source somewhere available if you send a quick email that we can at least have. Now with Dirk and the Cran2Deb project they pretty much obviated what I had been doing so they pretty much made it much better packaging set than what I had available. Hi I'm Chris Walker I work at Queen Mary in the University of London and previously my background is in X-ray diffraction so I had an interest in data acquisition and data analysis. I'm now working for GridPP so we have a grid cluster with about 1400 calls and we analyze data from CERN. It's running scientific Linux which is essentially a Red Hat derivative. That's going to be difficult to change I suspect but it would be really nice if some of the grid middleware was ported to Deviant so that on my netbook for example I could submit jobs and I could you know other people running Deviant on their desktop could submit jobs even if our servers actually still end up running Red Hat. Anybody else who wants to? Hi Enrico Vini I'm not a scientist myself but I work for meteorologists and I write a lot of software that does things that I don't completely understand but basically archival weather data organizing weather data kind of software to support research on weather and I used also libraries to parse various kind of formats and I even wrote a few of those so yeah I'm interested in seeing some of those packages in Deviant that there's a whole lot about geo libraries or libraries like NetCDF, HDF5 they're all tricky things because they all have their own quirks in where they are packaged they work they're linked by different applications in different versions and so on so yeah interesting and that's more or less why the data formats and libraries. Hi I'm Mario Lang basically made as a day job I work for the Technical University of Graz and we are operating currently about five smaller larger Linux clusters with about 700 cores we are using Deviant for the system yes and I'm basically interested in yeah as other papers so I have said there are a lot of problems with computer and computational libraries with linking and and different communication systems and especially the queuing system problem is something else I'd like to get forward on sometimes because we don't really have a functional queuing system in Deviant and that's one of the things that's absolutely required for running a cluster so you end up yeah doing doing a lot of of hand work and and manually installing software and it would be great if we could get this some more standardized inside of Deviant. So hello I'm Lionel Maman professionally I'm mainly active with proof assistance programs that check validity of proofs I haven't quite so my tries to get a few in Deviant have mostly repeated out because of uncooperative upstreams and negative ones but in Deviant science what I mostly do is mentor and sponsor of Philippo Rosconi who is preparing a few packages around mass spectrometry and also a bibliographical reference extractor. My name is Francisco Rivas I work for a research group which our interest in is to understand the evolution of the flaws actually we are not involved in Deviant science but we do some studies about mailing lists backtracker systems and source code management to get in the commuter the most important commuter the most active in commuter in the community that kind of stuff we use Deviant to understand how it is growing and evolving. Hi I'm Diego while I'm not involved in any scientific thing yet I have some interest on what happens in Deviant in this particular site in the anecdotic part I was actually able to do this trip thanks to my university I don't know school director or whatever actually using Deviant for scientific work so just an example. I'm Drake Diedrich I used to maintain DQS so I apologize for us not having a queuing system anymore I couldn't get it to build anymore it survived two full Debian releases without me rebuilding the binaries because the ones I built the first time worked and that was non-free software it evolved and turned into SGE came back with a whole bunch of Java dependencies and we didn't have a working Java for it yet so we don't have a queuing system. Yeah I'm Michael Bank and I primarily work on chemistry software. Yeah well I think in the last year some some interesting things on the meetings list came up about how we should handle for instance references so it do we have made up any opinion about how we could do this other there were some suggestions that currently it is more or less come and practice that people add references to their software insights a long description also because you need if you are a scientist and write some software you need some some some quotes for your software and so on but this this this quotes are not really a structured information and we are thinking about adding an extra file in the user share doc or something like have you followed the discussion or have you any opinion how to solve this problem or so yeah well actually I wanted to get back to that yeah I wanted to get back to that I just got reminded yesterday but I haven't reviewed the discussion yet and I think we should really go forth with it at some point but there were quite a few different ideas or opinions how to do that and it might be a bit difficult yeah I mean I think it would be nice if we could assemble references for programs in some central BIP check file also um so that people can check out the papers that are appropriate for this program and then get the background but it that's like a wish list but I think it's really important that we get the proper citations um in in some structured thing because um there's a lot of scientific projects which um seem to be like sort of site where that that means like they're free software but people say well you're required to cite this paper or so if you use it and if we could tell them like okay in Demian it's required if you do like let's say not required but there's like an informal policy of having this citation file or something and then users are supposed to look there and then cite it and so we could get around this copyright involving citation problem if we just decouple it and say well we have it there and so it's there and it doesn't have to be in the license so I really think it's an important issue and it would be nice if we could converge on on some sort of form I think it was mostly about like should it be BIP tech should it be playing tags and would be nice to be good converge is there any other opinions about that I was going to say I mean there are essentially as I see it two issues two issues there's one what um if you are going to cite a piece of software um you want to to easily and quickly find the best thing to cite um and then it might also be useful to have some references for how to use it and other what other people have done with it so and that I think is a technical problem the other problem is the social problem of when should you cite it and that that in in many ways I think is a separate discussion and that's to some extent we can say well it's up to the ethics and the standard practice of people in that scientific field um and I think I mean personally I think the first one is more critical to address first the technical problem that's what we do in Debian but I think it's also useful to have a have a discussion about when when one should cite a particular program I mean the other thing that's really useful as well is to have those citations especially when a program encodes multiple methods and it makes it very much easier for example R has uh almost in all of its documentation files tells you precisely from which paper the method that the function you're calling was written up in so if it's not clear from reading the documentation and it's not clear from reading the source code what it's doing you can go back to the original paper uh and see what's going on so I mean so both in context of citing the original paper but also in terms of additional citations where appropriate to papers that explain what's actually going on so usually I find at least in my area we usually cite the small paper that announces the uh the project so like R you would I forget which citation you would use but it's some citation in 1998 or something that um introduces the R project um and so it tells you exactly which one we suggest that you cite R in your method section you know by putting the citation um and I think a lot of the software does that with these we suggest or works to that effect instead of we require you to cite and maybe to to if we run into upstreams where that's a problem to pull a couple of these examples out to say okay look these are famous examples which are these papers are cited R for example has to be cited at least probably a hundred thousand times by now um and it works if you do it this way and there's enough community pressure behind it um so are people also citing these um secondary papers for R for every package or is it really just citing the R paper it depends so in my case if the standard method is if the package is one that is part of the base of R and you're not making a very explicit use of it then yeah I'll just cite R itself however if I'm doing work uh for example that specifically involves a sub package so I've done work on genetic association and there's a passion package called S&P SOC that actually does the test that I'm using um then I cite that specifically in addition to R well my concern is well not concerned but apparently now we have this really nice R package doing thing and um it would be interesting whether you could also like automatically extract the proper citation as well and then put it into a user share doc that package has a citation um or well well if you have to do it like manually figure out what the citation is and then that like sort of makes this automatic R package building a bit uh less useful at least in concern with the citation so yeah I think R has the fields in the documentation to tell you what the citations are I just they may be in bibtec but I'm not sure exactly what format they're and I'd have to go and dork would know but he's not here so I can't ask him uh once we are there and the documentation issue uh I've observed that quite frequently we have um in the upstream source PDF documentations of the package but no license no source code whatever it is always rejected by ftp master because no source code available um it is quite a stupid issue in my opinion because in most cases out of do not really think about providing their text sort of the pdf source is generated out of open office document or whatever and uh in in principle you you could make some pdf to text uh and then provide the text and ask you for then this is a source or this is not a source and what is your opinion about this so I mean I think here one of the most important things as early as possible try to get the upstream authors to provide whatever they use to build the pdf and license it under the same license as the paper if it's a word document that's kind of annoying but we can edit word documents uh so I mean maybe you've modified a little bit but even in the case of pdf documentation you often want to change it to fix it in cases where it doesn't match uh I mean I know in my case I write documentation sometimes for things that I use uh and what actually the code does diverges from the pdf file that I have in the case of journal articles it's a little weird but even then you still have the submitted file that you submitted to the journal instead of the published version of the journal article it's also quite frequently that your your package ends up in in in non-free or whatever because you want to provide the documentation but I really hate this issue because we it just makes trouble and without uh no no real use for users yeah I mean in some of those cases too though you can rip out the doc and provide you want to provide the docs for the users oh yeah but rip out the documentation and provide the documentation in a separate non-free package so I mean that's yeah that's what we're currently doing but yeah it's sub-optimal and the only way to fix it is to get authors to be aware of the issue because it's one of the things that most people now are aware of the need for free software uh in academia but sometimes they don't think about free documentation because it's only now that there's been such a huge push for uh open publications I mean it's really only the past three years or so that we see in the rise of open publication as big as it is now with the BMC and POS journals yeah in most cases I think the reason is that authors uh authors of software and documentation do not even think about applying this and and perhaps we we we find some framework letter we we could just send out to them so to put something out to that no not everybody has to invent the same letter always right yeah and I guess this too as soon as they've licensed the pdf correctly if they write back and tell you you know I have no idea where the doc file is then just put that in the copyright file and then that's an example of a binary that there is no source existing in the planet for it anymore so then it's like sort of free but suboptimal so that should be fine for main you think uh I believe so I mean it's not the question about whether we think but it's the question whether I have to pay more but if the author writes back and says that yeah we we looked for the source we cannot find it uh if we find it we'll send it to you or something like that then that's a case that any obviously licensed it under free software license right yeah I mean I guess a lot of problems is all the legacy documentation like there's a lot of pdfs from the 90s or 80s back then and well doesn't get probably updated and stuff and yeah it's a pain yeah I think there's some other issue I observed in scientific work because I've heard from from zana institute which is a large institute of biological research and rapid that they don't use our stuff we provide in deviant meet because they want to stick to certain versions of programs they have deviant packages because they have even three deviant developers working at the zana institute but they want to stick to a very certain version which is exactly able to reproduce the the results and it might happen that this version it's not even in any deviant stable release because well it was used some other point in time and do you see any chance to provide some deviant packages with with a certain version for a long time yeah snapshot is and no deviant snapshot but I have no idea how we could have the users to to get it with more simple than just pinning it to a certain word one of the things that would maybe be really really useful is to have some of the users of these packages produce test sets that are expected to produce a particular result and so if they stop producing that result then I mean they may not even need to be distributed in deviant itself but some sort of method of testing so that they can say oh yes these versions are all good they all do the exact same they reproduce the results we had previously and when they start diverging then they can ask about it but outside of that the only thing to do is to stick to version I know I do that sometimes when I publish a paper I fork branch out the repository that is producing the data in that paper and everything else continues on pace in the tree but the stuff for that paper gets locked in place because the people writing the paper have to write based on a consistent set of data so in terms of making experiments reproducible besides specific version of packages sometimes you have specific version of dependencies to be you know tracked as well and even specific version of build dependencies in case of use libraries where functionality is implemented like templates in C++ there is DH build info that will store in a package the information about the build dependencies used to build it and there is report bug that can generate a report of all the dependency versions of a version of dependencies of a package installed in the system so is it could it be a good idea to create a little script that gathers all the DH build info information of the packages currently installed and the and the versions and creates a little text file that you can attach to the research say well I've did this with this whole toolchain so if I need to reproduce it this is exactly what I use and the architecture and so on if you're really concerned about reproducing thing that something comes to my mind it's interesting that you're asking about snapshotting private repositories and testing because I've been working for two years now on a project dev marshal and I've got a um intern putting together the test suite for it which snapshots and stages and tests software so that you could find out which versions past tests you could go back in time and pick up you know the release on the release of debian unstable on this day that you kept so if you installed from your own private snapshots this would then be able to provide them back to you and we're also building a test suite for it and my opinion these are all interesting solutions for people like us who are very fit maintaining a deviant system but it's not so comfortable for for general user to to apply this so if you any other ideas I mean for most general users surely if they use the debian provided packages they can say we use package x on in lenny I mean it's it's if you're doing something slightly different it's surely the particular issue and I think I mean what what's been suggested sounds like a good idea and might be worth putting on the wiki as a you know this is what you can do this is an example of good practice and if you submit a paper it can go in the supplementary information that goes along with the paper for exactly what was used yeah but but sometimes the real life is that this is a version the user want is is not in lenny he has this stick to pre-lening release with some privately built it's deviant package or whatever to this person who does all that for that user or even if it's user themselves at that point knows enough about deviant to at least snapshot those packages that they've built in the build system so maybe just making it like you said and the wiki better available so that users who are doing that can be pointed out okay these are the steps that we recommend that you you do to be able to reproduce the system that you've generated but also I think that I mean certainly we want to have everything in deviant but if it makes sense we could also at least provide binary packages um from a different version if the source is in deviant for like maybe on aliyot or wherever I think that's that's an appropriate usage of aliyot um providing back ports or providing a specific version of user aspirate and they are not able to unbuild them themselves if they need it for their research so just saying well it's not in lenny uh your loss or yeah I think it should be possible for us to do that I think too would it also make it easier is when we uh complete the resupport of snapshot deviant.org so once that we get that machine to do that um so you can at any moment in time go back to any version of a package that we've distributed ever um that'll be a good thing that'll help make this easier okay if there are no no further opinion on this topic I uh have some some question about the stuff here I developed in on this box page the the the box all of you um just feel a few relevant meta package and I tried to find some some measure how many bugs are in in this in the dependencies of these meta packages and it turned out that the measure I found is not really helpful because you have many many uh bugs this is this is not a number of bugs it is some measures because the the release critical bugs are multiplied by 10 and so to to make sure that the release critical bugs are even higher makes makes even more noise on on on this list and the wishlist bugs are quite low priority it um but it turned out that nearly every meta package gets this red line uh and so it's not really that helpful we have uh here in in met metrology are not so many bugs uh but I have no real idea how how we could really make a reasonable measure to to attract people if there is something really wrong and uh perhaps you have some some good ideas how we can do that here is a list how how it is uh calculated so we have in the metrology section we have no critical grave or serious bugs we have um three important bugs in an only suggested package so it's only not so relevant and we have a few normal or minor bugs in in all the other if you look at um whatever the statistics is not so crowded so we have some better overview we have can can you see is anybody able to see this yeah I think the best solution is we crowd here around this table and look at the screen this is probably the best solution because this this is just does not work oh you look at your laptop yeah yeah 2024 but this screen is just shifted so on the in the bottom of of the page I have also some footnote where it is explained how exactly this bug measure is is calculated but I'm I'm not really sure that it's really reasonable because as I said we have most of the meta packages get the red light and this actually makes not much sense and does not really motivate people to to look at this page um currently I I think at least my hope was that chemists go to the just chemistry page and I'm personally using it in the deviant made project quite frequently to to look what packages are buggy and where we should work and I think it is um if you look only on the page you are interested most it it works but if you are generally looking at it this makes not much sense so perhaps you just have a look in in in a quiet minute and and build your opinion on it and we will not find the solution in 10 minutes here I think but at least I was able to show you the problem we have what you what you there's a your is http blends dot elliot dot debian dot org science and from from this you have links to tasks and bugs pages where you can click yeah I pasted the depth of channel in a lower room maybe it's I think it's also linked to from the debian science wiki page yeah it's definitely linked from there probably googling for deviant science might help as well I'm just posting the the enter page for deviant science as well yeah you can also click on these task pages one of the things it doesn't currently do and would be useful if it does although I realize time is finite is for the physics I realized that there were about eight programs that implemented matlab like for functionality and so I've split those out as a separate meta package because physics was including them engineering was including the maths was including them and it seemed silly but you don't then list include the bugs in those depended upon packages um in that list of physics bugs and it might might well be useful if you did so you mean if you if you look in the sense as common science utilities that you say science matlab like in in this in this context okay but you're talking about a sub physics meta package right not a not a general a political applicable no I think I think what you mean is that the physics tasks could just depend from science matlab like and mathematic tasks could depend from science matlab like because it's a general tool not only for physics precisely precisely and data acquisition similarly yeah this makes sense could could you send me a please a list of these matlab like programs there there there should be a meta package I thought I'd created one yeah yeah did you just or you can do it or but because I don't know all these matlab like packages I don't I only know Octave but I don't know the others is the numerical computation packet list thing in in your on the on the bugs page yeah could you post the exact link where you are looking for well okay so if it's blends alley off debian org slash science slash bugs slash numerical computation dot html or it's one of the links on the left yeah that's it so this you mean do you know data language precisely that's a problem or Octave is there I know Octave what what are the other ones you mean for matlab like all of those are matlab like I mean that for example iPython yeah okay combined with matplotlib provides similar of python like functionality that's that similar to matlab the new data language looks like it's a um um provides similar functionality to um idl okay and one of the things I have worked on in the past as you know is um is the wiki to try and improve that and indeed putting them back into your meta packages to try and improve the um the visibility of all of these nice cool packages that are in debian yes and that that actually is one of the things I think we really that I put a quite a lot of effort into um I think ultimately um and I'm glad Enrico is here um the his tags are the way to ultimately the way to go but I haven't worked out exactly the way to use his tags to implement um for a scientist looking at things saying I've heard of this pack this non-free package that does something like this might be interesting what are the free packages that are available what are the free packages in debian um and you know a list of packages that are similar and related um there's a great list on the chemistry website uh chemistry wiki page of different uh drawing packages and there are about 20 or 30 in debian which perhaps is too many but there are a lot of them are aimed at different function different bits of functionality so some are 2d drawing packages some are 3d some have got crystallographic information and so on and it and some um have integration with avanisher packages and it's really if you know rough if you've got a vague idea that you're looking for something but are not quite sure exactly what um what it is you or you you know what you want but you don't want to read the descriptions of 20 or 30 packages to find what does exactly what you want it's really um it'd be nice to have some way of doing that slightly better yeah right we should I should really get to it I know I that that I sucked doing well I always wanted to do these package tags and think about it but I haven't really gotten around it and I hope I'll get around here finally um just one thought occurred to me um could we also have or maybe you just said it and I missed it could we have is it possible to have like package tags which says like like s or like matlab like um is that ethically okay to suggest non-free packages people might want to search for as a user um Enrico what do you think um Daptex perfectly supports absolutely controversial tags um for example um Miriam Ruiz is providing um parental guidance tags for games rating extremely controversial things like violence or sexual preferences or religious things uh the way we do it is as an external tag source you can edit etc daptex sources.list and add extra tag sources to it or you can install files in etc daptex sources.list.d which make daptex include that um and then daptex will merge everything together so you can have you can put like in Debian science site something that just says all this like matlab is like Microsoft Office and so on and then um make in the meta package you just include that bit of configuration and and put it in there although it could be arguable that um I mean if that doesn't fit in Debian you can just put the package in non-free that installs the required I mean if you think that that would be the too much involved with proprietary software and not quite the aim of Debian it's just a matter of making like a 10 kilobyte package in non-free that installs the data file with the tags and adds the source to etc daptex sources.list.d and that's it and then you can really put whatever you want and then you get it gets merged with the regular tags for that package or how does it it does it gets merged with the rest of the tags and everything that uses up the xapian index will find it uh everything which reads the merged data set in varlip daptex will find it the only tools that won't find it is those who get the data from the packages file but they shouldn't that should just get data from other places yeah I'm just wondering is there any comments about should we include it directly in main or is that inappropriate what do other things people think we do science or is there any opinions about it sorry another thing you may want to do it task specific um because there may be conflicting proprietary software names or proprietary software used in different ways in different fields I don't know yeah for a purely legal standpoint I can't imagine a problem from certainly it's not a legal problem it's ethically a problem I would say like should we should we promote s or matlab or something I think when we're promoting it as a replacement I in my opinion anyway it's it's not ethically as challenging uh which when we're um I mean it's when we're usually it's already in the package description I'd say right and yeah and in those cases I can't see a problem with it okay but certainly it's nice to have the possibility of doing it like it just core to me that this could be a distribution wide thing including stuff like open office replaces Microsoft Word and or rice we sort of play yeah um and it can just be one packaging non-free with this information that can just be you know and it they've been game team may want to add to it um just a matter of saying who's coordinating this uh we can have a look uh I I'm I'm not too interested in driving this but I can get patches and put them together if people feel like one thing I found when I was trying to categorize packages is that I I mean some I I'm a physicist by training in crystal with background in crystallography so that area I knew something about but I I looked at the tags that were available and I was convinced that they weren't quite detailed enough for what I thought you know for the detailed categorization but I couldn't work out what the tags one should be using are and um across the whole of science and a consistent way of doing that and perhaps we can have some discussions either now or over lunch when Enrico who understands tags is here um and we who understand the science and work out whether the tags that are currently available are sufficient um for what to get a good categorization um I mean what one of the examples is there's a tag there's a physics tag do we tag software that does physics or do we tag software that's useful for physics and you know things like the matlab like software matlab is not physics matlab is is maths now there might be specific physics libraries which one would want to tag physics but um and again there's some parallel does um libraries which I think are probably tagged both physics and chemistry and I I couldn't work out what the answer was and there wasn't sufficient documentation in the tags to to give me an answer to that and I um you know came a bit unstuck so it would be nice if we had some some discussions and on that and worked out where we could go one thing to keep in mind is we don't need a good categorization we need a useful categorization so it's not an effort in being comprehensive um there was a we came out with a proposal of um some first approximation that will tag like matlab stuff as stuff for doing computation in science or something like that but I can't find the email and I blame myself for not having okay science modeling science data acquisition science plotting science biography science publishing that kind of follows the workflow of publishing a paper and that was kind of an approximation kind of task specific its categories that created not by the intention of um doing appropriate taxonomy of science but to support someone who has a practical problem of writing a paper and therefore you say well I want to write a paper I want to do modeling this set of tags will at least tell me what I can look into for that stage and maybe it can cross-reference with other tags for physics when appropriate but at least you see software for modeling and then you can apply you are looking at maybe a hundred packages instead of 20 000 and that's a big big step the rest can be done with packaging can be done with package descriptions with tags or package descriptions according to but that that already helps a lot just saying that we maybe should stop discussing or just quickly doing package tags but it's only like eight minutes left so after you we move on to the next topic I was going to suggest this might be an opportunity to use citations as your tags you know every program that implements the algorithm of a certain paper can have a reference to that paper as a citation and then when you're looking for something that implements that paper it pulls up a list of packages yeah but that might get a bit crowded unless you it has a DOI um that's rather easy maybe to have it as a package tag but there is some software which you can just reference you have to reference by name author and url yeah and and then it gets maybe the package tags get a bit crowded I don't know but we could think about it maybe we should pick Enrico's brain later on in depconf and move on to whatever Andreas has I have I have some some last thing I just spoke posting a another link and I'll see about the task page because I think this is a really nice and I think you can go to to any of these tasks reference and I go to what we had here it was numerical computation and you see the task page shows screenshots of the the program and I would like you to to look at those pages where there is a upload screenshot button and start the program upload a screenshot to make this page more more living and attract more people so everything is there you see where the screenshot is available and where not and this link leads you exactly to the upload page for screenshots maybe we can have like a list of recommended stuff like what Enrico said using dh build info doing screenshots like some stuff which is not necessarily in devian policy but might make sense and all over science so that these task packages get more useful well dh build info is not for that but anyway by the way you can also use depth tags here I think in this case the depth tags are all set or at least some depth tags are set if you go on this on the example but in case some depth tags are missing there is also a yellow button where you can go depth tagging or so if you see such cases just go for it you are the experts on on these packages and you can the best take us for this okay I think Enrico said he has some more to discuss yeah I have a few more technical issues since I'm not a scientist myself I just packaged the work for scientists I notice a few technical issues that could be special to science packages that I don't know if you're interested in in discussing or if there's any special behavior like package namespaces there is devian has a single package namespace and I've noticed like packages for sequencing chunks of DNA molecules whatever with extremely common names like combine I don't know something like that and I started getting worried like okay then I can combine DNA sequences I can combine bridge blocks and the more software we start to get in the more it's likely to have package conflicts and if there's a way to I don't know that that could be an issue that devian science could be talking about package name package names or file namespace file namespace executable file namespace right is also a problem because every scientist obviously thinks that he's the one or she's the one doing the combine program and nobody else would ever come so that's not doing that and I have two also two other topics I'd like to raise quickly one is we're mostly dealing with fringe packages fringe like packages that not many people are using which may have special properties like is it doesn't make sense to look at architectures and not have a supercomputing software waste time on arm build demons or be prioritized low ever lower priority or deciding that it's a best effort to build it on arm so it the fact that it's been built on arm won't prevent the package to go to testing or something like that or security implications or lack thereof we don't want to bother the security team I mean Debin can even say we don't provide security support for stuff that sequences DNA so the security team doesn't have to bother and users don't put a DNA sequences to get input from an open port on the internet issues more practical not strictly scientific but that could be common to scientific packages I want I want to raise the last point because in beginning batch queuing systems came up and maybe we could have a small box in official box scheduled at some point later in that come for every talk about what makes sense in maintaining this I mean I used to push great engine into the end but I don't really maintain it right now I don't really use it right now but if people I mean there's certainly some people who have like experience with batch queuing systems and we could maybe have a buff what makes sense and then if people are interested we can try to organize that so you can contact me or try to figure it out yeah so we we have to finish I think and so thank you for attendance and see you perhaps in some part for some other unofficial discussion and let's write up something for Debian science or so from this if we're proposing a buff which I think is a good idea can we perhaps have a discussion now while we're all here about when it would be interesting for sure but somebody has to do the work as always you know the principle and one last point the science what it's obviously a fringe effort and so there's a danger of packages getting lost or the maintainer going away as we noticed just before the last release and I I made had some efforts to try and say look this package has got some bugs that the release critical and one package in particular there was a new upstream release that didn't make it into Debian and it'd be nice to do that a little bit more continuously so that we don't end up with a big rush just before the release when the maintainer's gone away and the scrambling around to try to get an old version patched up so if we can if people who have interests on and packages can keep an eye on them and you know post bugs and set post to the mailing list saying this package you know I'm struggling with this package there's a much newer version come out is the maintainer around or if not is there somebody else who's around that can do something about it or do it yourself you know people do it themselves I think the next the next event should have just started now