Re: [gentoo-dev] udev-ng? (Was: Summary Council meeting Tuesday 13 November 2012)

[Rich Freeman]

Wow, that's some kind of thread you started...  :)  I'll respond in
general to a bunch of stuff on this list by topic.


COUNCIL MEETING

On Sat, Nov 17, 2012 at 10:29 PM, Greg KH <gregkh@g.o> wrote:
>
> So, that's a nice summary, but, what is the end result here?
>

Speaking as somebody who was there, but not for the council, the
summary was the end result OF THE COUNCIL MEETING.

The council was asked to set a deadline for everybody with a separate
/usr to adopt one of the proposed mitigation solutions, like using a
script, initramfs, or whatever.  That is ALL that it was asked to
decide on, and that was all it did decide on.  The whole business
about some devs wanting to fork udev came out about a day in advance,
and speaking personally it only had a little influence on my vote.

The reason I agree with chainsaw's proposal to defer the decision one
month was that there seemed to be enough blockers on this that nothing
was going to happen for almost another month anyway (best-case), and
getting people to move to initramfs or mdev or
[nu/eu]dev[-ng]/whatever wasn't actually going to be holding anything
up for a while.  I'd also have been willing to approve a plan to set a
target for something like 90 days after all the necessary tools (like
genkernel) were stable and news was sent out.  Based on my questions
for williamh I did not get the sense that delaying a month was
actually hindering the udev project (the established udev).  They were
encouraged to continue working on their blockers, preparing news
items, and so on - everything but having a deadline/go-ahead to break
systems that didn't follow the news.

So, a bunch of ideas were floating around in the meeting, and I
embraced the wait a month option since that seemed to have the most
support of any of the options out there.  If williamh had identified
some actual impact of delay on the udev team I'd have probably pushed
for setting the deadline now, but just putting it far enough out there
(90 days from genkernel/etc being ready) that all the various teams
would have a shot at it.  If the udev team gets their news items all
worked out and perhaps even sent out (sans deadline) and all the
blockers cleared before the next meeting I'd be supportive of setting
the deadline around 60 days, but that would be just moral support
since I'm not on the council.


OFFICIAL UDEV PROJECT

I have nothing to do with the new udev project, but I did pass the
staff quiz with much help from calchan.  :)

Read the GLEPs - any Gentoo developer can start a project at any time.
 That's how things work around here.  If I wanted to start a linux
kernel fork as an official Gentoo project I could do so tomorrow.

That doesn't mean that the new udev will become the default udev, any
more than Gentoo hardened will ever become the default experience for
new Gentoo users.  Gentoo is about choice, and if we have devs
interested in maintaining something new then we'll offer that choice
to our users for as long as somebody takes care of it.

If anybody wants to change the defaults/etc, I'd expect that to get a
lot of discussion, and almost certainly a council vote.


COPYRIGHT

I think this issue is best dealt with on the side - it has no bearing
on any of the really contentious points here.

I note that the owners of the copyright on udev have announced to the
world that (emphasis mine):
You may modify your copy or copies of the Library or ANY PORTION OF
IT, thus forming a work based on the Library, and copy and distribute
such modifications or work under the terms of Section 1 above,
provided that you also meet all of these conditions...

None of those conditions included keeping the copyright line intact.

Anybody can therefore alter the copyright line as they wish, as they
have been given explicit permission to do so.  They need only comply
with the other terms in the LGPL to do so (the most important being
licensing it under the LGPL and making the source available.

In fact, (L)GPL v3 has an optional attribution clause, and the fact
that they made this explicit is because some projects might not want
to give out this authorization.

So, if you want an official ruling from the trustees we would need to
meet/vote on it and perhaps discuss with counsel, but my thinking is
that anybody distributing work under the (L)GPL has waived their right
to be named on the copyright line of any copies distributed by others,
and as far as I can tell I have found nothing to the contrary from any
authoritative source.  The only way I think you could argue that
removing copyright notices for a (L)GPL work is illegal is if you
argue that an author doesn't have the legal power to license that
right to another.  However, I'd still think that promissory estoppel
would probably interfere with any kind of recourse - you can't give
somebody permission to do something, and then sue them for actually
doing it.  So, legal or not anybody with standing to sue over this has
likely given up their rights to do so.

Again, that's my two cents and not a license for anybody to do
anything.  This topic did come up recently with regard to accepting
some other kind of outside work into Gentoo, and as I recall there was
some debate over whether the copyright notices could be changed.  I'd
have to dig up the details - I think the issue might have been mooted
before any kind of formal decision was reached...


IS THE NEW UDEV A GOOD/BAD/UGLY IDEA

Seems like this is the main point of this whole thread, and I don't
find it terribly useful to harp on.  If people want to start a udev
fork more power to them.  I'm supportive of that, just as I'm
supportive of having systemd in the tree and unit files for as many
packages as possible.  As projects mature I'd be all for offering them
as options in the handbook.  Gentoo is about choice.

Ditto for a /usr move or whatever else.  I think we should have a
reasonable default behavior, and as others have pointed out we could
use profiles to control a bunch of these behaviors globally (like
library install location, and so on).  Again, offer the user the
choice, and generally be conservative with the defaults.

Will all these projects go the distance?  Hard to say.  I agree that
projects inspired by "hate" or whatever often fizzle out, but several
high profile forks have stuck around - usually because of conflict
over the major goals of the project.  The next few years should be
interesting, as the amount of vertical integration seems to be
creeping up, and if a major Gnome release is systemd-only or whatever
that could really bring things to a head.  This almost seems like an
androidification of the traditional linux distro - where choice still
exists but the ability to swap out layers starts to go away unless you
stick with a lighter desktop environment.  Should make things fun for
the toolkit developers when a system might or might not have
dbus/systemd/udev/X11/wayland/linux/bsd and who knows what else.


Rich