Ahhhh. I get the (significantly more complicated) picture. The questions to ask in that case...
1) Is the copyright individually held or is the copyright held by the consortium as an entity? I am assuming the former from your post. If so, you would need the permission of any copyright holder whose work ends up in the program in order to alter the terms of the GPL. In theory you could pare the code down to just what you need in order to reduce the number of authors, but that might not solve the problem. However, the problem could be worse. I would guess most of the code is probably not even owned by the authors. Anything I write at work is owned by the UC Regents, although thus far they have been kind enough to allow me to set my own licensing terms (within very general guidelines.) 2) Is IMKL or ACML covered by the GPL v2 clause "However, as a special exception, the source code distributed need not include anything that is normally distributed (in either source or binary form) with the major components (compiler, kernel, and so on) of the operating system on which the executable runs, unless that component itself accompanies the executable." or the GPL v3 clause I posted earlier (if the license is "v2 or later")? If the Intel compiler is required in order to use IMKL and an AMD compiler is required to use ACML, you could probably make the claim that it is allowed, especially if those libraries are included by default in any builds with that compiler. If you can link to these libraries using a stock GCC or Visual-C++ compiler, then that claim gets more difficult to make. If the consortium has access to all the authors via email, you might be able to get a consensus on an exception. Every 7 days for two months send a message: "Some member of the consortium would like to add an exception to the license that explicitly allows binaries to be distributed when linked with the proprietary optimized math libraries, IMKL and ACML. Source for all non-IMKL and ACML modules would still be require to follow the original license terms. Please send comments or objections to the consortium mailing list within the next N days." It might be that universal concurrence is easier to get than we think. I would guess a some of the other authors want to be able to use those libraries, too. _______________________________________________ boinc_dev mailing list [email protected] http://lists.ssl.berkeley.edu/mailman/listinfo/boinc_dev To unsubscribe, visit the above URL and (near bottom of page) enter your email address.
