[gentoo-amd64] Re: Sun and GPL

[Duncan]

Isidore Ducasse <ducasse.isidore@×××××.com> posted
20070528024149.4f6d918c@Bazaar, excerpted below, on  Mon, 28 May 2007
02:41:49 +0200:

> le Sun, 27 May 2007 23:32:49 +0000 (UTC) Duncan <1i5t5.duncan@×××.net> a
> écrit:
>
>> They ARE considering dual-licensing Solaris under GPLv3, however, which
>> they've been working closely with the FSF on.  Of course that's not a
>> given until it's out, but it'd definitely widen the interest base (I
>> for one may well be interested, especially if Linux stays GPLv2 only).
>
> You mean the bare kernel, right? Solaris' kernel could be an alternative
> to linux? Is the latter really different from the *BSD's? I've installed
> a NetBSD on my machine "for fun" recently (tho I switched back to using
> my good'ol gentoo, can't get used to anything else now. pkgsrc looks
> like a sympathetic old auntie); it appears to practice monolithic
> kernel. What would be different in running a GPLv3 kernel? I've read
> about the anti-DRM part of it; is there some other reason you/we could
> be interested in it?

BSS Jr's response covered much of what I would have covered, but I've a 
bit to add in places.

What I think Sun may be angling for in their leaning towards releasing 
OpenSolaris under the GPLv3, particularly since earlier the Linux kernel 
devs were nearly unanimous in saying they weren't interested in moving to 
the GPLv3 (as I said, based on the early drafts), is ideally to "Out-
Linux Linux", so to speak.

First, you may or may not have heard of Nexenta, or the Gentoo on 
OpenSolaris port, etc.  You also may or may not be aware that Gentoo/FBSD 
is one of the other Gentoo Alt projects, from what I read coming along 
reasonably well, too. =8^)

Basically, what many of these do is a variant of GNU/Linux, only in this 
case, GNU/BSD or GNU/Solaris or GNU/whatever, with the GNU toolchain and 
GNU based userland running on whatever other *ix kernel, be it FBSD, 
Solaris, whatever.  Debian does it.  Gentoo does it.  That's not new.

In fact, to some extent it's older than GNU/Linux, or at least older than 
the popularization of GNU/Linux, back when Linux wasn't half the kernel 
it is today, and was way under-featured and under-speced compared to the 
Unixes of the time.  Back then, while GNU had a relatively mature 
userland, it lacked a good kernel.  At the same time, many of the Unixes 
(Solaris included, this was back around Solaris 2, thru Solaris 4 or so, 
tho Linux was coming on strong by then) had solid kernels but aging and 
hard to work with userlands.  It therefore wasn't uncommon for people to 
buy a Sun box, and essentially replace most of the Solaris userland tools 
with GNU tools.  If you read about the time, many people tell how the 
first thing they did after they got Solaris up and running was install 
the GNU tools, and pretty much never use the Solaris tools again.

So there's really some history to GNU/Solaris, and it's not as strange a 
thought to /either/ side as it might appear to some of us newbies to the 
scene.

Now that Sun seems to be "seeing the light" in terms of free and open 
source software (note that they have a lot of code in a typical Linux 
install already, Open/Star Office, Java, particularly on servers, they 
are huge GNOME sponsors), and have already opened much of their Solaris 
code under the CDDL as OpenSolaris, were they to go GPLv3, with the GNU 
code ALSO licensed GPLv3 (after the official license comes out, of 
course), /especially/ if the Linux kernel remains GPLv2 only, it's /
quite/ possible Stallman and the FSF might officially bless GNU/Solaris 
and deemphasize GNU/HURD AND Linux.  He/they might then see that as one 
of the ways to encourage the use of GPLv3, in the face-off with Linux 
staying GPLv2.  Obviously, that could put a whole new twist on the way we 
see the Free Software community.

That's the Solaris side.  Now examine the side staying with GPLv2, if 
Linux indeed does so.  What's the future look like?  Well, we have the 
likes of Tivo, already making it impossible to run the so-called "open" 
code on their hardware, due to code signing and not releasing the keys 
necessary to run any user modifications on that hardware.  Many people 
predict that's the way DRM may head, the way of Intel/MS Palladium, aka 
"Trusted Computing", as well.  As another example, we have the HDMI 
digital audio/video interface, designed to only run what is properly 
licensed to run, or at least only allow it access to "privileged" data 
such as media content.

Then we have the whole Novell/MS patent deal thing, where MS licenses its 
technology to certain preferred Novell users, but not others, and not 
those using other distributions.  Further, MS says they won't sue 
hobbyist developers as long as they only use the changes they make 
themselves, not distributing them.  Of course, that breaks the back of 
the whole idea of Free Software.  The GPLv2 doesn't have any direct 
protection against such things, but the GPLv3 has been engineered in such 
a way that if you use GPLv3 code covered by your patents, if you let 
anybody else use it, any customers, etc, well then, it applies to all, 
customer and non-customer alike.

So, the future looks like it could be pretty dark for freedom, if we 
continue to depend on GPLv2.  The license will practically be little 
different than the 3-clause BSD license, as people will be able to 
effectively close their code, with patent agreements exploiting the MS/
Novell loophole and hardware signed code verification Tivoizing things so 
even with source, modified code won't run, even if they can't directly 
close it.  That kills the dynamic that has made Linux and a lot of GPLv2 
code what it is.  Without that dynamic, it could easily be headed for the 
relatively quiet backwaters neighboring the BSDs, open source still, but 
not (practically) forcing folks to contribute their changes back, thus 
slowing down development.

Put those two together and you have what I think Sun is hoping for.  If 
the OpenSolaris kernel can become /the/ blessed GNU kernel, and Linux 
says GPLv2 and gets BSDed, OpenSolaris could eventually eclipse Linux.  
While it wouldn't be proprietary, and it's likely they'll have to open up 
development even further in ordered for it to take off and become really 
dominant, so other companies would contribute and could distribute it 
just as they distribute Linux today, Sun would still be in a /very/ good 
position.

I think that's the game they are playing, the ultimate goal they have in 
mind.  And yes, if they go GPLv3 with the Solaris kernel, and Linux stays 
GPLv2, IMO it's quite possible it'll happen that way.  

IMO, the Linux devs will ultimately realize this too, and have to choose 
between marginalization and going GPLv3.  In fact, from Linus' recent 
comments, he seems to already be giving himself room to do that.  Yes, 
many of the issues he raised have been addressed, definitely making it 
easier to come around, but he may be seeing this same game being played 
out in his head too, and not particularly like its result.

> BTW isn't there a technical issue licensing a single version of a soft
> against two incompatible licenses? Or did you mean dual-licensing GPLv2
> and GPLv3?

It depends on who holds the copyrights.  The copyright holder can license 
however they please, and in fact distribute under a license that makes no 
sense, if they wish to.  It's the other people that have to live with the 
legal uncertainty, and that uncertainty exists ONLY because if the 
license isn't consistent, the copyright holder can yank permissions to 
distribute or even continue to run the software at all (because at least 
in the US, the act of loading into memory from disk or other permanent 
storage has been held to be an act of copying, thus subject to copyright 
restrictions and permissions).

However, this wouldn't ultimately be inconsistent.  The idea is similar 
to the business model used by Trolltech for Qt and by the MySQL guys for 
it.  In both cases, they dual (or more, triple...) license their software 
GPL, and proprietary.  The developer/user/distributor gets to choose 
which of the two licenses they agree to.  If they are going to free the 
code of anything they build on it, great, the GPL license works just 
fine.  However, if they want to build proprietary tools on the dual-
licensed software, they can't use the GPL, and must pay the company in 
question for a proprietary-commercial license, which generally costs a 
significant amount of money.

This has in fact been a QUITE successful business model for Trolltech.  
The open source guys develop stuff like KDE on Qt, which works as a 
pretty convincing demonstration of the capacities of the toolkit, as well 
as providing feedback and new features and bugfixes from the community.  
Other companies see how effective Qt is, and how it could shorten and 
improve their development process, and not willing to release their own 
code, they must pay to buy a commercial license from Trolltech.  Yet they 
are happy to do so, because the return is far more than what they pay.  
This in turn funds Trolltech to pay developers to continue to advance the 
product, benefiting both their paying customers and the Free Software 
side.

This model has in fact been SO successful for Trolltech that with Qt4, 
they opened up the GPL licensing to apply to Qt on MSWindows as well -- 
it formerly applied only to the *ix platform.  They'd not dare open up 
the possibility of a free version on MS if the model wasn't already 
demonstratedly working very well for them.  In doing so, they've also 
opened up the possibility of KDE on MS, and in fact, much of KDE 4 is 
indeed going to run on and be available for MS Windows as well as Linux 
and the other *ix platforms.  (Not the entire thing.  Most general KDE4 
apps will run in MS, they say, but KDE as a unified environment is going 
to remain *ix only, for both practical/technical and political reasons.  
For example MSWindows already has a windowing system, so KDE's isn't 
necessary.  Thus, the full experience will remain a 'nix thing.  That 
said, Konqueror for example, could give Firefox some serious competition, 
with its KHTML engine already being the basis of Apple Safari, so it'll 
now run on all three platforms just as Firefox does, and people could 
start with Konqueror and other KDE apps on MS, and continue using the 
same things when they switch to Linux -- or OpenSolaris. =8^)

So anyway, as long as Sun holds its own copyrights, and/or has gotten 
appropriate permission from the other owners where Sun doesn't hold them, 
they can license however they please.

>> Of course Linus and the other kernel devs were originally very much
>> against early GPLv3 drafts.
>
> Is it a matter of diverging positions towards industrial partners/users?

I think Boyd covered that pretty well.

>> The Gentoo Java devs are working on it, but as I said, I don't believe
>> enough of the entire Java infrastructure has been released as GPL yet
>> to do the entire thing as sources.  Even after it has, it'll take
>> several months as experimental ebuilds in the Java overlay (emerge
>> layman and read up on using it, if interested)
>
> Ok! Does anyone know the difference between the java-overlay and the
> java-gcj-overlay?

GCJ is GCC's Java compiler.  Generally, it'd be for compiling Java 
sources to arch-native code, not to the traditional VM targeted Java 
bytecode.  Thus, while it might be useful for someone wishing to compile 
their Java app just as they would a C/C++/whatever app, to native binary 
code to directly execute on their CPU, it's not particularly interesting 
for someone primarily interested in Java as a browser VM.

Since you specifically mentioned Java as a browser VM, I therefore assume 
you will be more interested in the standard java-overlay.

One word of caution, just in case you hadn't figured this out from what I 
and others have already said.  The Gentoo devs (and contributing users, 
overlays give the flexibility to allow non-Gentoo-dev users more direct 
access, if the devs in charge of the overlay trust them of course, 
without the user having to go thru the entire Gentoo dev process) use the 
java-overlay as a staging ground for working stuff up to standard Gentoo 
tree quality.  Some major changes go on there.  It was used to work out 
the switch to the new java-config arrangement before it hit the tree, for 
instance.  However, as the staging ground, it won't always work like the 
unmasked stuff in the tree should work.  At times, parts of it will be 
broken, and you'll have to do some things manually in ordered to get 
stuff to work, or unmerge it and go back to the stuff in the tree, if 
it's too broken.  It's there for users to use if they feel up to it, 
hopefully to test and pitch in and help if they find stuff broken.  
However, don't expect it to all just work all the time, because it's a 
development overlay, and development is what happens there, including 
breakage at times.  If you are prepared to deal with that, well, go for 
it! =8^)

-- 
Duncan - List replies preferred.   No HTML msgs.
"Every nonfree program has a lord, a master --
and if you use the program, he is your master."  Richard Stallman

-- 
gentoo-amd64@g.o mailing list