Jacques Pitrat was one of my intellectual masters, and the pioneer of French AI (his PhD thesis mentioned artificial intelligence in 1966). He was the president of the jury of my PhD defense (1990). I probably owe him my PhD title (at least partly). He has a page on French Wikipedia and a very interesting blog (in English) on http://bootstrappingartificialintelligence.fr/WordPress3/ and interacted in the mid 1960s with John Mc Carthy

I was in good relation with J.Pitrat, and visited him a few times each year, we spent delicious moments speaking of Artificial Intelligence. I believe to be respected by his family.

He spent nearly 30 years full time working on his CAIA system, a Quine artificial intelligence system in GPLv3+ license for Linux and Unix systems.

He gave me a tar sourceball of CAIA, and verbally the permission to publish it on my web page as http://starynkevitch.net/Basile/caia-su-24feb2016.tar.bz2 and he published a previous version of that software (as MALICE) on http://jacques.pitrat.pagesperso-orange.fr/MALICE.html

Jacques Pitrat passed away on October 14, 2019. I did attend his funerals

I believe to have kept his trust (even if we had some scientific disagreement: mostly I believe that artificial intelligence is part of computer science, he thought it is a different science).

I respect a lot Pitrat's vision and work. See http://refpersys.org/ for more.

Am I legally allowed to put the source code of that tarball on github.com ?

(obviously, under my name, but keeping the generated copyright notices mentioning his name).

The tarball contains many binary data files (e.g. _10527, _11065, _11603, _14) and many generated C files (such as dx.h and ATOME84T0.c, EQMPOSX0.c, PLUSC2.c, Z52Z0.c) essential for operation. The C files carry a comment mentioning GPLv3+, but not the data files). There are 3833 generated *.c files and 5972 generated binary files _*. CAIA is able, in 30 minutes of CPU time, to regenerate all the 38 34 C files it is made of (totalizing 522586 lines of generated C), and all the 5966 data files (totalizing 32 megabytes, serializing objects) it needs to work.

  • 1
    The verbal permission does not seem to matter in this case since he already released it under the GPLv3 as well. Unless the software he gave you permission to post on your website was somehow not also GPLv3 -- I did not inspect it to find out.
    – Brandin
    Feb 26, 2020 at 9:41

1 Answer 1


The GPL license is quite clear on redistribution: you may do so, under the condition that you retain the original copyright and indicate your own changes.

As such you are of course allowed to post the code on github, if you recieved it under the GPL license.

If you received it with another or no license - that's different. Technically you need explicitly the license with each change made by an author. If the license is part of the source package, it's usually no issue. If you just receive the license verbally, it will be hard to prove, if someone questions the legality.

However in your case: all(?) files carry in their header a copyright notice and statement that they are released under GPL. So I see little way someone can question the source code license, even when there's no license file.

  • Data files are essential and binary, without GPL notice. Feb 19, 2020 at 11:38
  • 2
    Yes, you are right. And they are not necessarily an integral part of the code itself. That's the grey area. Strictly technically it might then required to request permission from the heirs or former employer of the deceased (or whatever the legeslation requires). It boils down to what you can argue is the intention of the original author. Feb 19, 2020 at 11:40
  • Most of that work was done while Pitrat was retired. Before that he was at CNRS. Feb 19, 2020 at 11:45
  • 2
    @BasileStarynkevitch, for data files, there is also the question if any human creativity was involved in putting the data together. If it is just a compilation of facts put together in the right format for the software, then it most likely is not protected by copyright at all (similar to how a list of telephone numbers isn't protected by copyright). And for generated files, you should look at the source for them if there is any indication of a copyright license. Feb 19, 2020 at 15:05
  • 2
    @BasileStarynkevitch, if the bare bones system is capable of regenerating all (binaries and generated code), you should be in the clear. As a courtesy to users, you can provide them in the git repository. I'd suggest you mark them some way as "generated" (different directory perhaps?) with instructions on how to build. Or keep them someplace else.
    – vonbrand
    Feb 20, 2020 at 17:45

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.