[R] How can you buy R?

[Berwin A Turlach]

G'day Spencer,

>>>>> "SG" == Spencer Graves <spencer.graves at pdf.com> writes:

    SG> I'm not an attorney, but it would seem to me any code written
    SG> in R is arguably "derived from" R.
IANAL either, and I long since stopped reading gnu.misc.discuss in
which the interpretation of the various licences are regularly
discussed.

    SG> Even if R code were not "derived from" R, I don't see how it
    SG> could "reasonably be considered independent" of R.
R is one implementation of the S language.  If the R code works
without modification under S-PLUS (another implementation), then I
believe you can argue that it is independent of R.  

On the user level, it might well be the case that most commands work
in R and S-PLUS, but on the package developer lever there are enough
differences that typically the same code does not work on both, R and
S-PLUS, and that you have to make small adjustments depending on the
package.  If all R specific code is within "if(is.R())" constructs
(and likewise for all S-PLUS specific code), then you can probably
still argue independence.

It might become trickier if you handle the R/S-PLUS specific code
externally via Perl/Python/??? scripts and provide files with
(slightly) different code for the R package and the S-PLUS package.
In this case the R code for the R package is presumably derived from R
and has to be put under the GPL.

A question that always interested me was whether you can used GPL'd
code in S-PLUS.  At some point, I got the impression that according to
the GPL the user would violate the GPL if a package contained GPL code
(in particular C and/or FORTRAN code) that was dynamically linked into
S-PLUS by the R code.  My understanding was that in that moment a
product was created that would have to be wholly under the GPL, so the
user was violating the GPL and lost the write to use your package.
For this reason I started to use the LGPL for S-PLUS packages that I
put on statlib.  I noticed that when these packages were ported to R,
the licence was changed to GPL, but that is o.k. and allowed by the
LGPL.  I guess this question will soon become more interesting again
since there have been e-mails to this mailing list that S-PLUS wants
to become more compatible to R so that packages developed for R can be
easily used (ported?) to S-PLUS.  I guess the guys in Insightful have
to be very careful on how they do that... :)

    SG> A "boundary" case is provided by the "glmmADMB" package.  As I
    SG> read the GPL, this package must operate under GPL.
According to the DESCRIPTION file (at least the 0.3 version for linux)
it does.

    SG> This means that if anyone wants their source code, the authors
    SG> of that package are required to give it to them.
I agree, but isn't it all there?  Or are you talking about the files
in the (inst/)admb directory?  The authors of that package should
probably write somewhere tht the files in that directory are not under
the GPL and everything would be fine.

    SG> I just noticed that the version of "glmmADMB" that I
    SG> downloaded 3/14/2006 does NOT contain a "src" subdirectory.
    SG> This surprises me, given the comment on
    SG> "http://cran.fhcrc.org/banner.shtml" that "we generally do not
    SG> accept submissions of precompiled binaries".
But from where die you download it?  I cannot see it on CRAN.  I found
it on the web-site of Otter Research and, presumably, they are free to
distribute packages that contain precompiled binaries.

    SG> That is, however, not required by the GPL, as I understand it.
    SG> Rather, it seems to say that Otter Research
    SG> (http://www.otter-rsch.com/), who distribute more general "AD
    SG> Model Builder" software, could be required to make freely
    SG> available source code for all the binaries they use.  This
    SG> should be fairly easy for them, because their "AD Model
    SG> Builder" produces C++ code, which they could easily include in
    SG> a "src" subdirectory of their package.  The GPL would NOT
    SG> require them to distribute source code for the "AD Model
    SG> Builder" itself, since that has an independent existence.
I definitely agree to the latter.  But from a quick look at the R code
it seems to me that this packge does not dynamically link any code
into R.  Rather, it seems that the communication with the precompiled
binaries are via calls to "system()" and communications via files
written into a temporary directory.  So while the C++ code could be
made available in the "src" subdirectory, I don't see why the GPL
would require them to do so.  Those binaries seem to be also
stand-alone and independent.  You can probably reverse-engineer the R
code to see how you could use them without R.

    SG> If anyone has any evidence contradicting the above, I'd like
    SG> to know.
I guess the above indicates that I have partly a different
interpretation than you have.  But, as I said,  I am not a lawyer.
And as the German proverb goes "Wo kein Klaeger ist, ist auch kein
Richter"---which means that you will probably only get definite
answers to these licences questions by testing them in court.  Good
luck if you choose to do so. :-)

Cheers,

        Berwin

========================== Full address ============================
Berwin A Turlach                      Tel.: +61 (8) 6488 3338 (secr)   
School of Mathematics and Statistics        +61 (8) 6488 3383 (self)      
The University of Western Australia   FAX : +61 (8) 6488 1028
35 Stirling Highway                   
Crawley WA 6009                e-mail: berwin at maths.uwa.edu.au
Australia                        http://www.maths.uwa.edu.au/~berwin