Get more Science into your Distribution

[Posted November 8, 2006 by ris]

There are a number of specialty distributions with scientific applications. Generally they come as live CDs with a few handpicked applications, like the Quantian Scientific Computing Environment, or BioBrew Linux. What if you want a broader range of scientific applications? The packages are there, maybe in your distribution's repositories already; but if you just want to get going without spending a lot of time with {emerge, yum, apt-get, conary, etc.} there are two distribution specific efforts to make it easy.

Mandriva is joining with Scilab, a consortium managed by INRIA (French National Institute for Research in Computer Science and Control). According this announcement, the Scilab open source numerical computation platform will be integrated into Mandriva Linux 2007 (Discovery, Powerpack and Powerpack+) editions.

Ubuntu 6.06 LTS users looking for more scientific applications will want to take a look at Scibuntu. Scibuntu is a script that adds scientific programs and other tools to your Ubuntu desktop. Most of these programs are already out there in the Ubuntu repositories, but Scibuntu collects them for you and puts them into your computer and fetches a few more from other places.

Scilab is not open-source

Posted Nov 9, 2006 10:00 UTC (Thu) by ballombe (subscriber, #9523) [Link] (7 responses)

Despite Scilab claiming to to be open source since at least 2001,
it is not the case. Sections 4.c) and 4.d) of the license restrict
distribution of modified version to non-commercial purpose.

Scilab is not open-source

Posted Nov 9, 2006 20:00 UTC (Thu) by mightyduck (guest, #23760) [Link] (6 responses)

So what? It's still open source, just not GPL or "free" (as in speech)
software.

Scilab is not open-source

Posted Nov 10, 2006 7:19 UTC (Fri) by njs (subscriber, #40338) [Link] (5 responses)

Man, and ESR is always complaining that the problem with the word "free" is that people misinterpret it...

Please see the Open Source Definition, which begins "Open source doesn't just mean access to the source code": http://opensource.org/docs/definition.php

In common parlance (modulo this kind of misunderstanding), "open source" and "free" refer to exactly the same sorts of licenses; the differences are at the level of motivations and philosophy. If people want to be more precise, they say things like "OSI-approved" or "DFSG-free", but this level of specification is not really necessary in situations like this.

Scilab is not open-source

Posted Nov 10, 2006 15:35 UTC (Fri) by mightyduck (guest, #23760) [Link] (4 responses)

Sorry, but I don't give a flying f*ck about this ideological crap. To me,
open source means, I have access to the source code and can modify it to
fit my needs or fix bugs etc., period. I'm not a distributor and I don't
care about the other ideological bullsh*t.

Scilab is not open-source

Posted Nov 10, 2006 18:14 UTC (Fri) by kmccarty (subscriber, #12085) [Link] (1 responses)

Sorry, but I don't give a flying f*ck about this ideological crap [...] I'm not a distributor and I don't care about the other ideological bullsh*t.

Wow, defensive much?!

Without "this ideological crap," it's highly unlikely that Linux would ever have become competitive with high-end proprietary UNIXes, and you would have to spend $$$ to even have a platform on which to run Scilab, free (as in beer) though it might be. (I guess you could run it on Windows, for which you only have to spend $. But if you would rather use that than Linux, I assume you wouldn't be posting on LWN. Right? :-)

No one is saying "Don't use Scilab"; the OP is only cautioning readers that Scilab doesn't come with some freedoms that most people using a Linux platform routinely expect of their software. For many, this is an important piece of information.

Also, distributors' needs for software to be Free (libre) are not in fact irrelevant to you! If Scilab were actually Free software, distributors would be able to support its use and integrate it with other parts of their operating systems to a much greater extent. As just one example, in the present situation no one may distribute Scilab binaries that use GPL'ed readline.

Scilab is not open-source

Posted Nov 10, 2006 18:36 UTC (Fri) by mightyduck (guest, #23760) [Link]

The OP claimed that it's not "open-source" for some definition of
"open-source" and I don't agree with THAT! I know that it's not free as
in "free speech" but it's still open-source for me because I have access
to the source and can modify it. That is how I define open-source. But
that whole discussion is all about wording and leads to nowhere. For me
Scilab is "open-source", "free as in beer" but not "free as in free
speech". I (or better, my employer) doesn't care so much about the "free
as in beer" thing and would be willing to spend $$$ if necessary, but I
care a lot about access to the source because it makes finding and fixing
bugs or adapting the program to your needs a lot easier. You are no
longer at the mercy of some company and even if the vendor goes out of
business you can still sort of use the program and port it to a new OS
for instance (happened to me already). With closed-source, even if it's
"free as in beer" that wouldn't be possible. And I don't care whether I
can distribute the modified program or not as I'm not even allowed
(according to my employment contract) to distribute substantial changes I
developed to open-source programs.

Scilab is not open-source

Posted Nov 10, 2006 19:42 UTC (Fri) by njs (subscriber, #40338) [Link]

If you read again, you may notice that I said nothing at all about ideology (except to note that it was irrelevant to the discussion at hand).

"When I use a word," Humpty Dumpty said, in rather a scornful tone, "it means just what I choose it to mean -- neither more nor less."

"The question is," said Alice, "whether you can make words mean so many different things."

"The question is," said Humpty Dumpty, "which is to be master -- that's all."

If OSS to you means "a nice knock-down argument" (and from your response, perhaps it does!), then great, but, eh, don't expect to be able to actually y'know _communicate_ with other people like that.

Scilab is not open-source

Posted Nov 30, 2006 16:48 UTC (Thu) by lysse (guest, #3190) [Link]

Presumably you'll never, ever share the bugs you fixed, or the program you compiled that includes your bugfixes, with your friends or work colleagues. That would make you a distributor, after all...

If you don't care about the arguments, fine, but don't make the mistake of believing that your ignorance makes them irrelevant to you.

Scibuntu - one line of apt-get plus non-free stuff...

Posted Nov 9, 2006 10:17 UTC (Thu) by DeletedUser32991 ((unknown), #32991) [Link] (4 responses)

I'm afraid that the Scibuntu recommendation is a bit below LWN's usual high standards (thanks a lot for keeping them high!):

The difference between scibuntu and

apt-get install libgl1-mesa-swrast emacs21 build-essential g77 gfortran sharutils binutils sysutils tetex-base tetex-bin tetex-extra gs gv grace octave gnuplot qalculate-gtk labplot gromacs gromacs-doc rasmol pymol t-coffee r-base openbabel ncbi-tools-x11 ncbi-tools-bin python-biopython blast2 bioperl clustalw clustalx tree-puzzle treewiewx

It installs some additional non-free software and some whitespace. Thanks a lot.

Scibuntu - one line of apt-get plus non-free stuff...

Posted Nov 9, 2006 17:12 UTC (Thu) by Kluge (subscriber, #2881) [Link] (3 responses)

Now that gfortran is available, is there any need for g77?

Scibuntu - one line of apt-get plus non-free stuff...

Posted Nov 9, 2006 22:28 UTC (Thu) by kmccarty (subscriber, #12085) [Link] (1 responses)

Now that gfortran is available, is there any need for g77?

Last I checked, gfortran was still missing some FORTRAN 77 intrinsics. In part for that reason, all the FORTRAN-using libraries, at least in Debian (and so I imagine also in Ubuntu) are built with g77.

Also, the default ABI of code generated by g77 and gfortran is slightly different, so one can't always compile code with gfortran and link it against libraries built using g77. (Look up the -ff2c and -fno-f2c options in the g77 and gfortran info manuals for details.)

Presumably when the first shortcoming above is fixed, Debian-based distributions will switch to compiling FORTRAN libraries with gfortran. This will necessitate an ABI transition similar to the C++ transitions required when the default g++ gets updated to a new major version.

Scibuntu - one line of apt-get plus non-free stuff...

Posted Nov 10, 2006 3:52 UTC (Fri) by Kluge (subscriber, #2881) [Link]

Thanks for the info!

g77

Posted Nov 12, 2006 13:54 UTC (Sun) by vblum (guest, #1151) [Link]

Last time I checked, g95 was vastly preferable over gfortran. The gfortran developers made the choice to fork g95 and get gfortran into the gcc mainline as soon as possible. The price was, in my opinion, a focus on infrastructure over actually making the thing work. Thus, when I tried an official gfortran I simply could not get it to work whereas g95 works reliably and out of the box.

In my opinion, g95 can fully replace g77 at this point.

Get more Science into your Distribution, switch to Debian

Posted Nov 9, 2006 18:10 UTC (Thu) by ballombe (subscriber, #9523) [Link] (3 responses)

Seeing the tremendous amount of scientific applications packaging work made by Debian developers not being mentionned is a bit painful to me. Providing acceptable quality packages for most scientific applications is a challenge, and scripts like scibuntu mostly draw from this pool.

Nowadays, Quantian is much more than a live CD, it has grown to the size of a live hard-disk image and include more than a "few hand-picked applications".

Get more Science into your Distribution, switch to Debian

Posted Nov 9, 2006 22:32 UTC (Thu) by kmccarty (subscriber, #12085) [Link] (2 responses)

Very much agreed. I hope it isn't too self-propagandizing to mention a short article I wrote related to this problem: Physics Software Rant

Get more Science into your Distribution, switch to Debian

Posted Nov 11, 2006 4:23 UTC (Sat) by roelofs (guest, #2599) [Link] (1 responses)

Physics Software Rant

Note that you (indirectly) mention Geant4 as a "not a license" example, but they very clearly link a real, BSDish license off the page to which you link. I can't tell when you wrote your rant, however, so maybe that's a "new" development (30 June).

Greg

Get more Science into your Distribution, switch to Debian

Posted Nov 11, 2006 6:10 UTC (Sat) by kmccarty (subscriber, #12085) [Link]

Thanks for the notice; now fixed. (I was actually aware of the new license, but forgot that Geant4 was mentioned in this context on the web page.)

Get more Science into your Distribution

Posted Nov 9, 2006 18:39 UTC (Thu) by wookey (guest, #5501) [Link]

And here is another scientific distro. This one is really an engineering distro but that implies many of the same apps and great piles of documentation:

Cambridge-MIT MDP Project Resource CD

It is a knoppix bootable CD that also lets you run it under Wine on Windows. It sticks to leightweight X11-type filemanager/windowing stuff in order to squeeze it all onto a CD, but has an impressive array of useful software like CAD, PCB layout, gEDA schamtic capture, latex or openoffice doc-processing, octave (mathcad-a-like) which connects nicely to latex to produce rendered equations and Firefox to fill in the values for an equation, then gnuplot to draw the results. There is a 500page book on wood technology, assorted useful things like periodic table info and so on. I was impressed.