On Tue, Sep 13, 2005 at 05:25:47PM -0400, michael chang wrote: > On 9/13/05, Stu Teasdale <[EMAIL PROTECTED]> wrote: > > > The killer is the GPL code linked into libmetrics. AIUI there's no legal > > way to use GPL code in a BSD licenced project, as there is no allowance > > in GPL for "derivative works" to have anything other than a GPL licence. > > > > This rather wrecks libmetrics, which is statically linked even in my > > packages. The only real option is to pull > > srclib/libmetrics/linux/fsusage.c and fsusage.h or get > > the original author to relicence it. Given that the original author is > > listed as the FSF I suspect this is unlikely to be possible :). > > Which means that theoretically linux support is (at least for whatever > metrics are provided by fsusage.c/h are) impossible without GPLing the > program. Oh dear.
fsusage.c is just a wrapper around statfs. Someone just needs to sit down and implement device_space directly in terms of statfs without looking at fsusage.c. The obvious thing to do would be to start with the FreeBSD code since (from the Linux statfs manpage): CONFORMING TO The Linux statfs was inspired by the 4.4BSD one (but they do not use the same structure) This should take a competent reader of man pages no more than a day. > Do we know when Matt will be available to talk about his options? Or > is he forced to BSD Ganglia because of some other condition when he > created the program (or because of e.g. his official job position or > where it's being used)? I'd prefer to keep it BSD licensed, but I'm a BSD user. :) Note that the FreeBSD code is currently under a four line BSD license and thus can't be GPL'd. The third clause can probably be removed since it's all UCB code and they rescinded that clause a while back. -- Brooks -- Any statement of the form "X is the one, true Y" is FALSE. PGP fingerprint 655D 519C 26A7 82E7 2529 9BF0 5D8E 8BE9 F238 1AD4
pgpsgAGaIjaCb.pgp
Description: PGP signature