Aleksej Saushev <asau%inbox.ru@localhost> writes: > Taylor R Campbell <campbell%mumble.net@localhost> writes: > >> By default, math/octave has the option hdf5 enabled, which makes it >> depend on devel/hdf5, which in turn has the option szip enabled, which >> makes it depend on the non-free software archivers/szip. >> >> I wish that nothing that is itself free software in pkgsrc would >> depend on non-free software by default. (I'd rather not see non-free >> software in pkgsrc at all, but that's a separate issue.) It seems >> especially silly that for a GNU program such as GNU Octave, pkgsrc >> would do that! > > I strongly object the policy of shipping underpowered software just > because of questionable political reasons. If a user expects legal > problems, he can turn the option off. The basic LICENSE framework is not a political issue; it's instead the notion that unless one makes a conscious decision one will only get software that is at least one of Free or Open Source. So far, no one has made a case for changing that. The issue here is really about binary packages, because it's easy enough to change options when building from source. But, we currently don't provide binary packages for packages that depend on non-redistributable packages, and szip can't be distributed at all. There's a separate issue of people not liking non-free software in pkgsrc, and by pkgsrc has long been agnostic - only making sure that people don't get non-free software by accident. In the szip case, there is no permission to redistribute. So octave binaries with szip can't be redistributed because of that, and they also can't be redistributed because szip's license is GPL-incompatible. So I agree with Taylor; the right thing to do is either to turn off the hdf5 option by default, or to make hdf5 not use szip by default. Is hdf5 without szip generally useful? If so, we should change hdf5, and if not change octave.
Description: PGP signature