 Here she is, Helen Faulkner, telling us something about Debian in Skynes Research. I think it's quite easy to tell that this is the last session, and that half the conference is already left, but that's life, and I'm really enjoying my new email address. For anyone who hasn't figured that out, and I might almost be tempted to go and hug Elmo again, except that I'm scared that he'd take it away from me, so maybe I won't. You're right, I could do. Now, how do you work this? Magic. Okay, so about me. I'm a physicist. I do academic research. I'm at the beginning of my career, hopefully to be a shining one, and mostly I do optics, mostly theory, mostly focusing on electron microscopy, which is entertaining and useful sometimes. Most of my work is theoretical, but I've done a bit of experiment in the last few years when I was living in the UK. And my research is, to put it as simply as you can, it's about seeing small things, looking inside really, really small things like crystals or proteins and seeing what the chemical structure of those is. And I just love this cartoon. How many people here are involved in science? Anyone? A few of you. Oh, that's great, because I was really, really hoping that lots of the people who showed up would be people who have opinions about this stuff. I've used Debian for a few years now since I gave way to my housemates' urgings and said, all right, will you shut up if I let you install it? And he did, and the rest is history. I got much more heavily involved with Debian about a year ago, thanks to the Debian Women Project and the early controversy which surrounded it. And mostly I use Debian for my research. That's what I mostly use the computer for, that and IRC, of course. And I'm really interested in making Debian more useful, easier to use for science research because I think it's such a great operating system with such potential for science. And one of the reasons is that there's a really clear match in philosophy, in my opinion, between science and free software. Scientists, research scientists, we're used to collaborating with people to achieve a common, sometimes a very far-off, goal. We're absolutely used to building on the work that people have done before us. It's often extremely incremental. We're used to sharing our work so that other people can build from it. And we're used to making judgments about each other and ourselves based on the quality of the work that we do and to knowing that our success is very strongly connected to our reputation as a good scientist or as a bad scientist. And everything depends on that funding, getting a job, being able to do the research that you want to do, depends on having a good reputation. And I think that's not unlike the free and open source software community. So scientists really already understand very well and agree with the philosophy of the open source movement. If we didn't agree with it, we wouldn't be doing what we're doing. And I think that that means that Linux is a very natural choice for scientists to use in the computer work that we do. And of course, Debian is the best distribution out there. So the question is, why don't all scientists use Debian? I'm not gonna tell you actually, I want you to tell me. But here's what I think. What do scientists want out of an operating system? Well, I wrote down the stuff. That's not all showing, is it? Oh, that's bad. Can you scroll? Well, I can't see it here, I don't remember what it was. Yes, but there's at least one other point. That's a problem. Okay, we have a problem with the, possibly with the difference in resolution between this screen and my screen. Yeah. All right, so I'll be guessing the bottom point on every slide from here on, but we can deal with that. Okay, so I sat down and thought, oh, what do I want as a scientist? And the first thing that I need for my work, which is mainly theoretical, is a coding environment and compilers for the languages that I'm using. And I know that lots of people might laugh at me for admitting it, but I grew up writing physics code in Fortran. And there's a whole lot of people still doing that because there's such a wealth of numerical libraries for Fortran, and it's fast and it's easy. And then I've since coded in C++ mainly because I thought it would be entertaining to switch without having any idea of how difficult that would be, but that's a few years past and I've sort of gotten my head around it. The other language that the people that I've worked with have used extensively is MATLAB. And again, because it's so easy, it's high level. You don't have to think about the grotty details and I mostly don't want to know about the grotty details. The next most important thing that I need is graphing programs which will handle really large data sets, which I collect with my research and produce plots that I can whack into papers easily without mucking around for hours and hours. I want access to high level languages or scripting languages or things within the graphing programs that allow me to do sophisticated or interesting analyses of my data so I can pull out the information that's in it, get the results that are in there. And of course, I need to use latech or sometimes word processors. In physics anyway, there are some journals which will only accept Microsoft Word documents as paper submissions. So mostly I use latech when I can, but I can't always. So we need this stuff for writing papers and for writing talks, I usually use open office but you need some sort of presentation. I often hear it asserted that open office is not on. It says it's on, oh there we go, it's just the levels. I often hear it asserted that open office produces word document format, that is good enough but sometimes academic publications have extremely strict standards. Is it your experience that openofficeproduces.doc files that they can't tell you're not using Microsoft Word and that they find acceptable for print publication? In my experience, it's 95% likely to do so and I admit that my computer, my laptop is dual boot and I check it in Microsoft Word first before sending it if it's important. If it's something that's not very important, I don't bother but if it is important or if it's complicated because I still find sometimes that I open my Word document that was saved in open office in Microsoft Word and the pictures have moved or the formatting has gone a bit funny somewhere or an equation hasn't come out or something stupid like that. So it's close but it's not quite there. Ditto shifting between open office impress and PowerPoint is the same, it's close but in my experience it's not quite there. I haven't tried open office too. That might be better, I don't know. Hopefully they've rectified some of those and people in the scientific community in particular can, we can get them involved in reporting bugs and trying to resolve that. Of course Microsoft will move the bar again their next release but you know. Yeah, of course. Okay and the last thing that I need is access to the specialist software that applies to my field and this is not necessarily going to be available for Linux. That's in fact the main reason why my computer is dual boot in that I have to be able to write digital micrograph scripts. This is a scripting language that you use to talk to an electron microscope. It's very broadly used in my field. It's dead easy to write scripts in but it's not available for Linux. So that ties me to using Windows at least some of the time and I think that a lot of scientists are actually in that position. And one of the implications of this is that I often want to be able to use software which is available for Linux and Windows so that when I'm stuck in Windows for a day because I'm doing mostly digital micrograph scripting I can do the other stuff that I'm also doing in Windows as well. So I mean open office is good for that. I mean obviously latex is fine as long as you've got reasonable editing environment on both sides and so on. In terms of the broader features, broader things that I think scientists want or that I want, I want stuff that just works. I would much rather be doing physics. It's more interesting than trying to figure out what on earth this thing is doing that's just died on me and why it's died and then writing a bug report and then having the maintainer of the package get back to me to say I've got no idea what you're talking about. Can you please clarify? And then I try again and then, you know, I'd rather be doing science. I want software which is easy to find. I spent, I don't know how many hours, ages really, tracking down different plotting programs, trying to find something that would plot a three-dimensional data set. And I tried, I don't know, lots of different things. Everything that I could find that was available in Debian except for the ones that have a really scary command line interfaces because I don't like those. I ended up packaging Labplot because there was nothing really good but you know, this to me is the issue of having to hunt through so many different packages to find the one that actually works for you. To me is a big hassle. I would rather have gone to a website and had someone say, yeah, you should use Labplot. You know, then I could do more physics. I want software that I can hand to all of my colleagues, burn it onto a CD and say, look, you can use this. Say, look, no, you don't have to pay for a license. No, it's free. Give it to your students, all that. I think that's a massive advantage of free software for science. And all right, as I was saying, I want software that's available for Windows. I don't use a Mac, but some people would, but for Windows and Linux so that I can use it myself and so that I can obviously save something as a PowerPoint presentation and email it to my boss who doesn't have open office or whatever. That last slide says anything else. I want to know what other people want. One thing that I've seen needed in my meteorological environment of the moment is small tutorials. People find the software and maybe there's a huge manual. Maybe there's a complicated common line interface. They have no idea if it does what it needs. Maybe if they find it useful, they read the whole manual, figure it out, do lots of stuff, but having like a two page, like do this, this, this, this, this, and see the results just to have an idea of the whole philosophy is really important. It's usually done by peers. Like do this, do that. Oh nice, oh yeah, okay, oh it does like this. Yes, great. But it's rarely found on a piece of document. Maybe useful stuff for that would also be things like screenshots on web pages to say this is what you can produce with this piece of software. I know I find those useful when I can find them. I'm into electronics. And there's a lot of software for electronics in Debian and in Linux in general. And I feel that the thing that's missing is that it should be all integrated. If you want to do something, you have to take parts from all different places. You have to use like 10 different packages and each package works differently so you need to convert one format to the other and it takes a lot of work because it's not integrated. I feel that in my field, that's the main lack. Do you think it would be possible to coordinate the packages either at the Debian level or communicate with the upstream developers of those packages to achieve greater integration? I think that the problem is actually upstream. It's not at Debian level because they are so different that it's difficult to integrate it at Debian level. Yeah, we should talk to upstream and work on that direction. Is it mostly a problem with file formats? Sort of, no, there are different things because sometimes it's file format, sometimes it's lack of data for a certain program. You need like, for example, the definition of the components and one program has some components and the other has some other components and that's tough. Yeah, fair enough. I guess wish list bugs might help but I also think, as far as I can tell with the very little specialized things, it's often someone writing the thing that he or she needs for their own research and if they don't want to use the stuff that you want to use it with then they don't have motivation to spend lots of their research time writing the component that does that. Yeah, maybe we can talk to people more about it. Also, there are some programs missing in Debian which has programs for clusters which is, for example, QS systems similar to PBS. As far as I know, there are no such programs in Debian. There is ITP but no packages. Also, we needed a Globus toolkit. We tried to compile it in Debian but it uses Java and we just failed to compile it and we needed to install Red Hat and there was also a problem that some complex programs only have RPM packages and it is not trivial thing to compile them for Debian. Okay, the one I'm aware of is a thing called ImageJ which does image manipulations and calculations on the data and I've thought about packaging it a couple of times but it's Java and I'm scared of it. But that I would like to see in Debian. I just make a comment about the batch schedulers and then also grid middleware like Globus toolkit and stuff like that is that typically on different clusters people have their own setups so coming up with a package to fit all of these different sites would be quite difficult and also if you've probably installed PBS it's sometimes just easier to install it and then forget about it. Then regarding the grid middleware or grid software people work on these things but then when they apply for funding to work on these things they don't consider things like packaging. So you have people who are writing software and then they package software but they don't they package in a format which is or they make a package that works but not a package that is let's say nice. Running a few clusters on Debian there is a packaging of OpenPBS Torque but it isn't in Debian proper but there is a third party packaging for Debian. Also you can find versions of Globus for Debian just not from the main site. A couple of the centers that do run Debian actually republish their packaging of common cluster software under Debian as Debian packages. I think one thing that would probably be really very very nice indeed when it comes to scientific software not just in Debian but outside Debian is if scientific software held the same standards of quality as any other piece of software. I've spent far too much of my life firstly trying to find out how this stuff is supposed to work at all secondly trying to work out how they ever thought it could have worked because it plainly can't possibly. And thirdly these pressing things seems to be that on the whole scientists can't write software. And possibly shouldn't be allowed to write software a lot of the time. There's a lot of scientists with no formal training in software writing. And unfortunately we end up with a huge pile of software and people find it easier to rewrite rather than to try to get somebody else's software working. Possibly just forming a good pool of easily available scientific software would do something to alleviate that if people could make it. If people had some expectation that they could install a piece of scientific software and then modify it for their own needs or just use it for their own needs then they may be less inclined to spend months writing grant proposals in all suspends years rewriting it. Yeah that's a very sensible idea. Okay so I said down to write down what I think of the pros the benefits for using Debian for science research. And I mean the first obvious one is that there are a lot of packages. If you want something chances are there's already one of those in Debian. And this is good. And Debian gets flack because it's supposed to be too difficult but actually I think that most scientists or scientifically minded people can deal with the level of complexity of the Debian installer. And that's not an insurmountable problem for most people. Certainly in physics most physicists are very geeks at heart. I think that would be a bit different in some of the software sciences psychology for example. There might not be so many people with the computer skills required to install and run Debian. And one of the other benefits is that certainly some of the software available in Debian or Philonix generally is better than what you can get for Windows or for Mac. I'm all about KDevelop. I use that as my development environment and I think it's much better than the environments I used in Windows in the years before I switched to Debian. There are other examples. Debian is cheap, download costs really or the cost of a CD. And this really matters to students and also for people like me and early career researcher. I don't have much funding. When I have students I need to provide software for them. There won't be much money to do that. So Debian is really good from that point of view. I'm not paying hundreds of dollars in licensing costs for some obscure piece of graphing software. And also Debian is sticking around. It's not going to be gone in five years. So you know that you can start someone off on a PhD. They're going to be three or four years maybe longer. The software will still be around. Hopefully it will still be maintained. Hopefully. You know that Microsoft is not going to stop supporting it. So I think that there is a time investment involved in learning to use Debian if you haven't used it before especially if you haven't used Linux before but I think that that's going to pay off in the long run. Does anyone else have other? That was my list. Does anyone else have other ideas of what the advantages are using Debian for research? You're a boring lot. I mean there's nothing else good about it. Yeah. Debian tends to have installation tools a bit more advanced than others like FAI. And when you have to reinstall all the labs or something like that you can build a bit larger infrastructure than that. And while I work in a Fedora environment and well now they've been spending like three months to test the new Fedora Core 4 if it's going to be any decent or not. And then they will have to like install it machine by machine spend like a month on it and they're going to be quite hit by it. Okay. Yeah, that's right. That is an advantage. So the disadvantages I thought of. The first one is that there are a lot of packages and you can spend a long time fishing around for the one that is actually going to be useful for you. And it can be sometimes very difficult to find to get any sort of measure of the quality of any particular package in Debian. I think we need to give them stars or something. This is a five-star piece of software and that's only a one-star piece of software, so, you know. And also it's true that some of the software in Debian as far as I know is actually not as good as the non-free alternatives. And the one that really springs to mind that I'm aware of is MATLAB. As far as I know, Octave, are you going to disagree? I'm not expert, but my impression is that Octave doesn't really compare. Go on, go on. Could you please explain why you say that Octave doesn't really compare? I have to admit that I haven't used Octave, but I've asked people about it. And at the last time I looked at the MATLAB development environment, which was MATLAB 6 or so, I was staggered by how much it had improved since I used it as a student. It's really, really easy. It's something that you can let a student with no programming experience loose on and they'll be able to work it out. Well, Octave is pretty stable, pretty freezed, but you have the Octave Forge part, the one that's developed in SourceForge. And they really work a lot to keep up with MATLAB new functions and everything. And at least as far as I've needed to use it, and I've used it quite a lot for different subjects at university, it has never failed to do anything, everything I needed to do for my subject. Maybe in some specific field, it might not fulfill the needs, but I really feel it's good enough. Okay, that's good to know. I disagree. I use, so everybody I work with in machine learning uses MATLAB. And I thought, well, fair enough, I'm gonna switch to Octave. There's very little support for sparse data matrices. And if you're working on huge, huge data sets with like several million features, you need to use the sparse matrix toolkit, which is there in MATLAB, but isn't there an Octave yet? People are supposed to be working on it, but in the year and a half that I've been looking at this, nothing has been done. There are various other functions that exist in MATLAB that I try to use that exist in MATLAB, but don't exist in Octave. So for instance, when somebody gives me a large MATLAB program that they've written, I can't run it in Octave. Without going back, editing it, hopefully stuff will actually be there in Octave, maybe having to rewrite functions that are built in in MATLAB. Octave for many years was much slower than MATLAB. Recently, it's been not so bad. You can do some really cunning stuff with Atlas and stuff like that to make it run faster. And then it does run very fast, but I've never succeeded in getting that to work. Like I put several hours into trying to compile the relevant stuff, and no. Yeah, I've used both the standard Octave stuff and Octave forage, yeah. And I find it incredibly frustrating because everybody I work with uses MATLAB and the compatibility is not there and the speed is not there, the sparse matrix support isn't there, there are many other functions that just aren't there and maybe it's okay for some fields, but certainly for machine learning, it's not. And that's a really big problem, machine learning. So probabilistic modeling, Bayesian probability but for practical tasks basically, yeah. So we need a lot of finding probability distributions, we need a lot of sparse matrices, that kind of stuff. Yeah, did you want to say? Well, yeah, I mean, the basic problem is just this, that there are these big standard libraries of functions for machine learning and most people just don't care or have never tested whether they work with Octave. So it's maybe not that difficult to fix each individual bit of program to work, but you'll then end up spending your whole research time fixing other people's code and they probably won't even want to take the changes back. Sure, MATLAB is available for Linux, isn't it? Yeah, so it's just a question of forking out for the license. Yeah, I want to agree with the last opinion. I've been used MATLAB a lot in my investigation and so and it doesn't reach the level we need, it lacks a simulink, the libraries lack a lot of functions and work very differently and it doesn't reach the level. All the people I know that uses Linux at the university for this are using MATLAB for Linux, not Octave. So maybe we really need to warn people that if their research is very dependent on MATLAB, they're gonna have a problem with Octave before they waste weeks trying. Great if there was a document around actually detailing the differences between MATLAB and Octave and saying which things aren't supported and Octave. But the problem is the people who actually have that knowledge are too busy publishing papers to actually write such a document. So I think there is definitely a need for that but I don't know anybody who has time to do it. Yeah, so going on, for me, I have problems in my research because I need to use proprietary software. I spoke earlier about digital micrograph which is only available for Windows and for Mac. That's the small bane of my life. That means I need a dual booting laptop. It's not the end of the world. Cross-compiling my code so that my colleagues can run it on their Windows machines has been much more entertaining. And I think I've gotten it beat now but it was difficult and I think that we could probably do better at, again, maybe producing, maybe I should write a how-to, producing documentation so that people who need to do this for their research and I think it can't be that uncommon. It can get some information about whether it's likely to be practical or whether it's likely to be easy for them to do or if it's going to take weeks or whatever. Does anyone else here cross-compile? No, am I the only person insane enough to agree to do that? All right, well, okay, so let's grab that. Another problem that I think is possibly more significant is that you go to the IT person in your department and say, oh, I'm running Debian and they say what? Actually, I was very impressed when I started at Monash recently that the IT guy was able to point me to a webpage which claimed to give instructions for setting up the email stuff in the email system for Linux. It didn't work, but at least there was something there. But actually, in practice, not all scientists want to support their own systems, especially not their system and the system of the five students under them which is likely to get broken because students mess around with computers. Well, yeah, you may not even be allowed to, even if you'd like to. So this is a problem. I was just gonna say, I mean, certainly in the department where I work, this is kind of an interesting one because, I mean, their Windows is unsupported as well, but in the standard system is some kind of weird forked stuff from Red Hat. But that means again, in that situation, people are actually less likely to want to go for a Debian system just because they've already got this mostly working system. And even though people with, I mean, well, I don't know whether it's the problems with the Red Hat or the problems with the stuff they've done to it, but people waste a huge amount of time with things that I know would work fine on a basic Debian system. But of course, yeah, if you're gonna throw away the fact that it's supported, it's a big decision to actually bother doing that and running it yourself. Yeah. Of course, the other side of it is that universities tend to be fairly permissive about things and you can quite often get away with doing whatever you want to do, which you might not be able to do in a different work environment. And what other problems do people have? It occurred to me after watching Margot's talk about our users and are we really prioritizing them that internationalization of software would be a problem for lots of researchers. It's okay for me, but probably other people that's all right. Anything else? You're gonna be boring again, I don't know. So where do we go with this? One thing that I think would be, yeah. I think of Debian compared to Red Hat is that it's harder to make packages. Of course in Red Hat, it's very easy to make crappy packages, but at least where I work, they can make Red Hat packages of the software they use and put them in some local app repository. There's in no way they may be able to do it with Debs. Maybe there's a case for Debian people to put their names on a list to volunteer time to a will package your thing for you so that you can use it with Debian. Because if it's simple, it's a trivial matter for many of us to create a working Debian package, which might require hours of work for someone who doesn't have that experience. So one of the things that I thought would be of great use to people trying to decide to use Debian or people using Debian for research would be to have a list of the stuff that you really want installed so that you don't have to muck around comparing different packages that do roughly the same thing. So I wrote down the things that I use. I use, the first thing I use is the development environment and I use KDevelop. Plus the compiler for whatever languages you're using. I use Labplot to draw my graphs. I use LaTeX and a GUI editor for it. And I use OpenOffice. Beyond that, things that I might use in future include Octave, though maybe not based on what your people are saying. And Python I've been told is very good for numerical analysis and I don't know about that. I've been told. GNU-R. Sorry, GNU. GNU-R. Okay. I've been using the GNU-R for... MCMC packages. Okay, sure. You were saying that GNU, is it R? R, okay. I'm not familiar with it, but... Okay, the website... I don't. So the website is... Okay, great. Okay. Yes, Brandon. I think we need a mic. Yeah, it's simple. It's a stop. It's simple. You can repeat it. Okay. You are stopped. All right. We'll go back to Enrico, because I'm sure he'll like what I said. I was just wondering about Kyle. I actually hadn't heard of it and I've been a Lix user for some time. If you could use Lix, could you give me some insight as to the differences between the two? I haven't used Lix. I think I might have looked at it once. Kyle is an editor. Give it a go. I find it easy, but I'm heavily into GUI things. If I can press a button to say compile and run, I will rather than having to type something in. So that's my take. Yeah. It's not WYSIWYG. Is Lix WYSIWYG? There is... Okay, Kyle is not. Kyle is you're actually writing. Yeah. Look, tell me what you think. Yeah, I'm installing it now. Sure. Do you know if there's a GNOME equivalent to KDevelop? I assume that there is, but I don't use GNOME. Can anyone answer that? Oh, okay. Is it good? What's the name of the packet? Anjuta. Anjuta, okay. I haven't tried it. I'd expect there'd be something as good as KDevelop or, you know, broadly similar. Okay. Cool. Anything else? Can we have a small party for the final introduction of Fortran 95 in Debian? Yes. Yes, there's that. I think... Yeah, okay. That sort of comes up on my next page, which is this question of how does Debian software compare with the alternatives? And I think we've already discussed Octave and Matlab. And again, I haven't tried the Debian Fortran 95, but I'm interested actually in switching back to using Fortran for my physics code. Because it has some real advantages that I'm sure lots of you would be aware of. So can anyone comment on the new compiler? Does it work? I mean, is it easy? Or am I going to spend hours or weeks getting it to work? Has anyone tried it? Yeah. It's still pretty... It's still pretty raw. I don't use Fortran myself, but judging from the man on this, it's still pretty raw. It's very actively being developed upstream. I think once GCC hits 4.1, it'll be a lot better. OK, so I should not leap into it right away. Well, don't, like, you know, make your job depend on it. OK. I think lack of Fortran is a real problem. Has been, possibly, obviously still is a real problem for a lot of researchers. Certainly in physics, loads of people use Fortran. That one is a problem. And again, for me, again, we talked about it briefly. The office suites or word processors that sometimes you're going to have to use to submit papers. Has anyone else got experience of major problems with Debian packages for producing word documents or PowerPoint presentations, being the main ones? Open Office Calc tends to kind of blow up when you try to connect it to databases to pull in data sets. Mm-hmm. I know I've had trouble with Excel spreadsheets opening them in Open Office, and they died because Excel isn't case-sensitive. Open Office was, is by default. That's about the only trouble I've had with Calc. I was just going to say, I mean, in machine learning, the journals still tend to ask people to do stuff in latex, and then, in a way, it's easy for us there, and it's actually the windows user who will have a who end up having problems. But on the other hand, if we want that situation to stay the same, rather than people all to switch to Microsoft Word, then we as, well, the free software community needs to make sure that the facilities for writing latex and so on continue to improve. I mean, for example, the GUI editors, which I think, I mean, I think are kind of that, but it's still people end up having problems. I mean, again, even if you're just using latex, even if people are happy writing raw latex, you can still have problems with things like ending up with PDF files with bad fonts in them and so on, and all these kinds of technical issues that really should be solvable. Yep, yep, I agree. Actually, speaking of PDF, I think OpenOffice is saved to PDF is one of the best things they've ever done, and that's one serious advantage it's got over Word. As far as I know. And again, the other thing that I use is the plotting programs. I used Sigma Plot in Windows for several years. It's good. It's really, it's a bit slow, but it's good. In my opinion, Psygrafica used to be better. Does anyone use Psygrafica now? I've had a lot of trouble with it in the last couple of years. As far as I know, it's not being maintained upstream, and it seems to have died, at least for me. Maybe it just hates me and my, however the Python stuff is set up on my computer. And that was why I ended up packaging Labplot, which is being very actively developed upstream, but it's still lacking some of the features that Sigma Plot and Psygrafica had. Notably, I think really easy rotating on the axes to get a look at your three-dimensional data from whatever direction you want, stuff like that. So I have hopes of Labplot in the future that I think it will well overtake the competition, but I don't think it has yet. Does anyone else know of good plotting programs that we should know about? GNU Plot. Everybody I know uses GNU Plot. I don't use it. It may suck, but everybody uses it. Yeah. People have a lot of respect for it, even though it is apparently horrible. So it's probably worth at least looking into. I'm sure I did look at it, because I've heard of it, and I went through everything I could find in Debian, but I must have had so much trouble that I've blanked it out, because I don't remember exactly why I didn't like it, but I didn't like it enough to... So if you use Octave, then all of the plotting stuff. So Matlab has a plotting environment built in, whereas Octave uses GNU Plot. So that's another Octave Matlab difference. Yeah. Well, that's a good thing. If Octave compares to Matlab in its plotting abilities, then that's a good possibility. Well, only kind of. Matlab's pretty good for plots. Yeah. Hello. Well, I used POR++. I don't know if anything could be useful, but it's quite powerful. It's available in Debian. Okay. I don't know. And the last thing is the question... Yeah. ...of making it easy for scientists to use Debian, or to get the help they need with using Debian. I've wondered whether it would be worth setting up a mailing list, Debian Science, something. And, yeah. I was just going to suggest an idea which has just been pinging in my mind over and over again since your talk began, is this sounds like a great opportunity for a custom Debian distribution, like Debian Med or Debian EDU. You know, have Debian Psy, or whatever a real scientist would want to call it. There's something more clever. Yeah. As an unlettered person, I wouldn't presume to pick a name for the project, but that's something I might personally be interested in getting involved in is kind of the Debian... It's an interesting idea. Debian experienced half. You know, as a person who can probably package something quickly, you know, you maintain the X window system for a few years, you learn the ins and outs of packaging, at least up to a certain point. I think that's something that we should probably try to move on towards. You have not been the first person, I believe, on Debian Day, which was a week ago, Saturday, there was a talk given relating the... Actually, it was Gunnar Wolf's talk relating the scientific tradition to the principles of free software. It seems a very logical place to have a kind of marriage. Absolutely. That's a really good idea. I hadn't thought of that. I think that's a really interesting idea. We should talk about it. Well, that would go a little bit further than making a wiki for people to write down their experiences. But... Yeah. I think it would be great if people could search on, you know, Linux science software or something like that and come up with a page listing, well, this is good and this is crap and this one's changing. Well, of course, because Debian is distro. Anyway, don't... You know what I mean. And the other thing that we can do as scientists is let our colleagues and our students know about Debian and say, hey, I've got this great piece of software that I'm using that I can use because it's Debian and I can just say, I've got to install this and it's there and all its dependencies are there and you Windows people don't have a hope and you Red Hat people are going to be struggling for hours. Let people know why it's good. Yeah. I know there are some kind of knopics-like life CDs for different kind of scientists and maybe you can have a look at the software selection they used to give a free environment for this kind of scientists. Yes, that's the good idea. It would be great to be able to hand new students. Here's your CD. You can just install this or whatever. About CDDs, this seems one of the fields for simple CDD which is a script that takes a package selection and a list of pre-seeds and builds a DI image for it and it would be nice to expand it to make a live CD. Yeah. Well, the problem of making a CDD for science sounds like the problem of making a CDD for non-profits. Science is really broad. I'm in the weather environment and things are totally different than in physics than in electronics and so on. So you need to have a large variety of stuff which you can do when you have a simple tool and you allow everyone to make their own one and at least they would work 100% inside Debian. They would be motivated to package for Debian having it inside of the BTS and so on. And that was my command for CDDs so I guess simple CDDs is a good way to try. A good thing to try for that. Okay. About other things that could help DevTags has now really shitty support for science tags because I haven't been into the science world for long and I'm no way expert and I could really use some help in finding out categories from people who actually know science software inside Debian. If I try to have a look I get lost between quantum physics and things that I can't understand if it's math or physics because they refer some mathematical thing that I totally don't know but that could be a nice effort because I am pretty sure we could come out with a lot of categories because we have a lot of software which will show that Debian actually contains a lot of useful stuff we are not used to think of. That's a great idea. I think that we're just about out of time. I think, well, we've done that bit really. So if there are no more comments I think. Last two. I have a small request for help if anyone has worked with Starlink software, among other things they contain possibly the best library for binding Fortran and C together which can do the bindings for any possible architecture and it's GPL software and I can't get it into Debian because they are the packaging equivalent. They are British and they are the packaging equivalent of driving on the left. There's no way to take their upstream tarble and turn it into something useful and it's like that's one major packaging problem I'd like to talk with others about. Okay. I was just thinking about an idea that was suggested before and it's opening some kind of project in which scientists or developers who do not have a clue on how to package for Debian would put their programs, asking for help. I mean, I do have this problem I want this package and everybody could have a look at it and say, well, I'm interested in this one. It would be nice to open such a project inside Debian. Yes, I agree that would be great. So I think that's the end. Thank you for telling me what you think now I know more than I did.