Linuxant adds native 64-bit (x86_64) support to DriverLoader

Posted Jan 8, 2005 21:27 UTC (Sat) by Duncan (guest, #6647) [Link]

IIRC, both sides moved. They claimed they were only doing it to prevent
multiple "kernel tainted" notices, and let the first one display. The
kernel devs, after some discussion, decided they /did/ have a point there,
that displaying a single notice that the kernel was tainted, then avoiding
displaying the notice (but still logging it) when further kernel-tainting
non-open modules were loaded, /was/ better than displaying the notice
possibly many times each boot. Thus, the kernel code was modified to only
display the notice the first time a tainting module was loaded, at the
same time detection of such lying modules (well, a blacklist, not real
detection) was added. Linuxant agreed that now that the notice was only
displayed once, they'd remove the offending code, since that's the only
reason it was there in the first place, they said, while not admitting
that what they did was actually a violation.

Actually, from the LWN report, an argument could be made that they weren't
technically in violation, or were technically but not in spirit, depending
how one chose to look at it. There are functions exported to all modules,
and functions exported with a gpl-only tag. They made use of a gpl-only
tagged function in a non-gpl module. That clearly violates the intent of
the function tagging, but a narrow legal argument could be made that due
to the wording at the time, allowing limited linking for non-gpl modules,
and not specifically disallowing it for gpl-only tagged functions in the
license wording that allowed the limited linking in the first place, it
wasn't technically illegal.

Further, they claimed that their intent wasn't to infringe or hide the
fact that they weren't GPL, but rather to avoid the display of multiple
tainted warnings. Since the kernel devs ultimately agreed that the latter
was better, they claimed that even if it was a technical violation, it
wasn't a violation in spirit, even if they did go about the solution in
the wrong way, just doing what was clearly not intended to be allowed,
rather than asking nicely that it be changed.

Most on the kernel team and those with a libre perspective believe there
was a violation. However, the compromise was made and Linuxant agreed not
to do it any more, so those with copyright privs that had legal standing
to sue in the case decided that was a mutually agreeable solution and
decided to let it be.

It should be noted that technically, because the kernel /is/ open source,
while it could include measures to prevent current modules from linking,
said modules could just be written around that, if the intent was to
deliberately violate the terms both intended and as written. While that
could/would ultimately lead to some sort of court challenge to enforce the
license, until an injunction specifically preventing it, this could lead
to a measure/counter-measure increasingly escalated contest. Further, due
to differing local laws, a deliberate violator could simply move
development from nation to nation for some time. While they'd ultimately
be stopped and their code declared illegal in most of the world, this sort
of outcome is undesirable for many libreware authors as well, for the
international cooperative precedents it could set for other code,
particularly that which offends the US DMCA, or similar statutes
elsewhere. That's in addition to the costs of pursuing the case. Thus,
as has been the common case with GPL violations, once the alleged
violators agreed to stop and appeared to do so, a negotiated settlement
was in the interest of all parties and therefore took place.

IANAL and all that, and I'm not specifically familiar with this case,
either. This is just based on the coverage I've read here on LWN and
elsewhere, as I understand it. Therefore, best verify the above info
before using or quoting it, and corrections welcome.

Duncan