[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: randlib licence
From: |
David Bateman |
Subject: |
Re: randlib licence |
Date: |
Wed, 11 Jul 2007 15:49:18 +0200 |
User-agent: |
Thunderbird 1.5.0.7 (X11/20060921) |
Petr Gajdos wrote:
> Hi David,
>
> first, I must say, that I haven't another wish than leave octave as is,
> really. As I stated before, I am packager of octave only and law is far away
> from my mind :). But, can you define notion "believe", which you use two
> times before? Your answer has some presumptions, doesn't it?
>
I'm an engineer and not a lawyer.. I therefore use "believe" in the
context of legal advice on the source code in that is my opinion. Your
lawyers might see differently
> SUSE layers've asked me, and I've asked you. And I've asked authors of
> randlib
> themself. After few weeks, there's no reply at all (btw. thank you for your
> replies). I am really sorry, but I need *yes or no*, not something in the
> mindst of. Patch I send you I rate as temporary solution, I preffer remove it
> as soon as posible from factory. Let's solve it before 10.3!
> Petr
>
Hey the code is really old, it doesn't surprise me that no one responded..
The published code in ACM is under the ACM license. However, the authors
have every right to republish this code under any other license they
like. They have done so in that they RANDLIB license states..
"We place the Randlib code that we have written in the public domain"
So the randlib code that the authors wrote is public domain.. So what
didn't they write? This was the crux of my previous doubts.. So I dug a
little more, and in the randlib/README it clearly identifies the code
from other sources as
<quote>
SOURCES
The following routines, which were written by others and lightly
modified for consistency in packaging, are included in RANDLIB.
Bottom Level Routines
These routines are a transliteration of the Pascal in the reference to
Fortran.
L'Ecuyer, P. and Cote, S. "Implementing a Random Number Package with
Splitting Facilities." ACM Transactions on Mathematical Software,
17:98-111 (1991)
Exponential
This code was obtained from Netlib.
Ahrens, J.H. and Dieter, U. Computer Methods for Sampling From the
Exponential and Normal Distributions. Comm. ACM, 15,10 (Oct. 1972),
873 - 882.
Gamma
(Case R >= 1.0)
Ahrens, J.H. and Dieter, U. Generating Gamma Variates by a Modified
Rejection Technique. Comm. ACM, 25,1 (Jan. 1982), 47 - 54.
Algorithm GD
(Case 0.0 <= R <= 1.0)
Ahrens, J.H. and Dieter, U. Computer Methods for Sampling from Gamma,
Beta, Poisson and Binomial Distributions. Computing, 12 (1974),
223-246. Adaptation of algorithm GS.
Normal
This code was obtained from netlib.
Ahrens, J.H. and Dieter, U. Extensions of Forsythe's Method for
Random Sampling from the Normal Distribution. Math. Comput., 27,124
(Oct. 1973), 927 - 937.
Binomial
This code was kindly sent me by Dr. Kachitvichyanukul.
Kachitvichyanukul, V. and Schmeiser, B. W. Binomial Random Variate
Generation. Communications of the ACM, 31, 2 (February, 1988) 216.
Poisson
This code was obtained from netlib.
Ahrens, J.H. and Dieter, U. Computer Generation of Poisson Deviates
>From Modified Normal Distributions. ACM Trans. Math. Software, 8, 2
(June 1982),163-179
Beta
This code was written by us following the recipe in the following.
R. C. H. Cheng Generating Beta Variables with Nonintegral Shape
Parameters. Communications of the ACM, 21:317-322 (1978) (Algorithms
BB and BC)
Linpack
Routines SPOFA and SDOT are used to perform the Cholesky decomposition
of the covariance matrix in SETGMN (used for the generation of
multivariate normal deviates).
Dongarra, J. J., Moler, C. B., Bunch, J. R. and Stewart, G. W.
Linpack User's Guide. SIAM Press, Philadelphia. (1979)
</quote>
Linpack isn't an issue, however we use the exponential, normal, gamma,
poisson routines as well as the base generators. So the problematic
references are
L'Ecuyer, P. and Cote, S. "Implementing a Random Number Package with
Splitting Facilities." ACM Transactions on Mathematical Software,
17:98-111 (1991)
Ahrens, J.H. and Dieter, U. Computer Methods for Sampling From the
Exponential and Normal Distributions. Comm. ACM, 15,10 (Oct. 1972),
873 - 882.
Ahrens, J.H. and Dieter, U. Generating Gamma Variates by a Modified
Rejection Technique. Comm. ACM, 25,1 (Jan. 1982), 47 - 54.
Algorithm GD
Ahrens, J.H. and Dieter, U. Computer Methods for Sampling from Gamma,
Beta, Poisson and Binomial Distributions. Computing, 12 (1974),
223-246. Adaptation of algorithm GS.
Ahrens, J.H. and Dieter, U. Extensions of Forsythe's Method for
Random Sampling from the Normal Distribution. Math. Comput., 27,124
(Oct. 1973), 927 - 937.
Ahrens, J.H. and Dieter, U. Computer Generation of Poisson Deviates
>From Modified Normal Distributions. ACM Trans. Math. Software, 8, 2
(June 1982),163-179
So yes there is a case that randlib is not GPL compatible. The current
ACM license is dated June 1998, and all of these references are prior to
that. I checked the ACM website and unfortunately they only have the
articles themselves and not the legal material for the journals
online.. So someone who had a paper copy of the ACM transactions from
1991 would have to check the licensing terms at that date... If there is
nothing basic copyright applies and its not GPL compatible either..
Ok, I think I've convinced myself that randlib has to go... Damn..
D.
> Dne středa 11 červenec 2007 13:58 jste napsal(a):
>
>> Petr Gajdos wrote:
>>
>>> Hello,
>>> I have tryed to remove randlib from octave 2.9.12. Can please someone
>>> look at patch attached, if it's ok?
>>>
>>> Have a nice day
>>> Petr
>>>
>> Petr,
>>
>> Does SUSE require this to distribute Octave?
>>
>>
>>> Dne středa 13 červen 2007 00:20 jste napsal(a):
>>>
>>>> On 12-Jun-2007, David Bateman wrote:
>>>> | which makes the whole question moot. If the authors, as owners of the
>>>> | copyright on the code, place the code in the public domain, I don't
>>>> | see how an ACM restriction can apply. The only question I see is in
>>>> | the statement "that we have written". If there is any code in randlib
>>>> | that was not written by the authors, then it would default to the ACM
>>>> | license as the only one under which you can distribute.. I don't
>>>> | believe this is the case, so I think randlib is in the clear for
>>>> | incorporation in GPLed software..
>>>>
>> Basically as I stated above I believe the code is not constrained by the
>> ACM license as the code is explicitly placed in the public domain by the
>> authors. So I believe the code is GPL compatible. However if SUSE
>> lawyers think otherwise, then that is something else and the patch you
>> supplied might be a way out.
>>
>> However, the state/seed means of switching between the generators is
>> compatible with what matlab does and it would be a shame to loose this
>> immediately, as it will allow existing Octave code that explicitly sets
>> the key to continue to have the same behavior. Yes in the long term the
>> randlib generators should go, but if SUSE can live with randlib in
>> Octave perhaps it would be better to keep the old generators around till
>> say the stable release after the 3.0 series (3.2?).
>>
>> Regards
>> David
>>
>
>
--
David Bateman address@hidden
Motorola Labs - Paris +33 1 69 35 48 04 (Ph)
Parc Les Algorithmes, Commune de St Aubin +33 6 72 01 06 33 (Mob)
91193 Gif-Sur-Yvette FRANCE +33 1 69 35 77 01 (Fax)
The information contained in this communication has been classified as:
[x] General Business Information
[ ] Motorola Internal Use Only
[ ] Motorola Confidential Proprietary