MOSIX

["Bill Rugolsky Jr." <rugolsky@ead.dsa.com>]

Miguel de Icaza,

I'm CC'ing this to linux-legal, because it seems the appropriate place
to get a serious discussion going without flooding lkml.

I'm glad to see that you started a firestorm at /. over MOSIX.  When I
went to the MOSIX site after the MO6 release, the language left a bad
taste in my mouth, though not knowing the details, I assumed the binary
module language would allow this.  I hope that you are correct.  You
seem to have gotten Alan and others interested, and that is a
Good Thing.

After the long lkml thread regarding breaking binary compatibility and
AFS, and now this, I'm increasingly in favor of getting rid of
binary-only modules.  (I certainly don't use any, and have no intention
of giving my money to vendors that require it.) The GPL is difficult
enough to defend without the additional complications of holes like
this, and it *will* get worse.  If the MOSIX precendent stands, what is
to keep the major database vendors from implementing proprietary (and
different) raw I/O, say, to use another lkml thread as an example? The
binary-only module exception could end right now if the kernel
developers just designate all new patches, and thus all future
releases, as strictly GPL.

I have long believed that if folks want to write proprietary code for
Linux, they should do it in userland.  If they need access to kernel
facilities, then they should write a GPL'd driver to provide the userland
interface.  If they don't like that option, for performance or complexity
reasons, they should simply use BSD.

Anyone care to comment?

   Bill Rugolsky
   rugolsky@ead.dsa.com

-
Linux-legal: discussing the legal aspects of Linux and Open Source
http://humbolt.nl.linux.org/lists/