Re: eigs and Octave.app

[David Bateman]

John W. Eaton wrote:
> On 12-Nov-2007, David Bateman wrote:
>
> | Mark McClure wrote:
> | >
> | >  2) My impression is that eigs is unlikely to every be included
> | >     in the standard Octave distribution due to license concerns
> | >     involving ARPACK.  Is this correct?
> | >
> | > I ask the second question more for my students than for myself.
> | > I'll be teaching a class in numerical linear algebra next
> | > semester.  Our students have some access to Matlab on campus, but
> | > I'd like to let them know about Octave as a viable alternative to
> | > potentially save some of them some money and to expose them to
> | > open source software.  Many of our applications of interest
> | > involve eigensystems of large sparse matrices, however, so a
> | > working eigs function is really essential.
> | 
> | Well, Sorensen isn't necessarily against a GPL distribution of ARPACK
> | from what he said to me. However, he would be required to negotiate with
> | Rice Universities Lawyers to allow this. The license terms of ARPACK
> | allow full distribution in both source and binary form of ARPACK with
> | Octave.
>
> If this were true, then we would be distributing it.
>   
It is true, but ARPACK is not GPL compatible as you have to contact Rice
University to state that you are using ARPACK if you use it in another
project than Octave. Therefore someone can't extract Octave's eigs
function and use it in another project with an additional step that is
excluded under the GPL.

Note ARPACK is a fedora package for example as it has been cleared by
their lawyers for distribution in Fedora..

> | The restriction of ARPACK is trivial in that Rice University
> | must be informed when ARPACK is used in a different project.
>
> It may be trivial, but it is still an additional restriction not
> allowed by the GPL, so as far as I can tell, it makes the ARPACK
> license incompatible with the GPL.  The clause in question is
>
>   Redistribution and use in source and binary forms, with or without
>   modification, are permitted provided that the following conditions
>   are met:
>
>     [...]
>
>     Written notification is provided to the developers of intent to
>     use this software.
>
>
> The ARPACK authors may have said "we don't care", but the license
> clearly states that redistribution requires written notification.
> Since the GPL requires that *everyone* must be permitted to
> redistribute the software without asking permission to do so, this
> causes trouble.  It is also not clear what "use" means, so narrowly
> it seems that this means that even someone who just wants to run the
> software must provide written notification.
>
> Also (as a separate issue) it is unfortunate that the ARPACK source
> distribution (arpack96.tar.Z) doesn't even include a copy of the
> license and the words copyright and license don't appear anywhere.
>
> If the "written notification" clause were removed from the license,
> then Octave could include an interface to ARPACK.  It would also be
> nice if the arpack distribution itself included the license statement.
>   
I'd like nothing better given the effort I put into writing the eigs
function. I'm just not willing to chase Sorensen at this point, but
would endorse anyone else who did.

D.




> jwe
>
>   


-- 
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