Torvalds vs Schwartz GPL Wars 335
javipas writes "The controversial message published by Linus Torvalds (mirrored) in the Linux Kernel Mailing List was from the beginning to the end an open attack to Sun and its Open Source strategy. Linus criticized Sun's real position on GPL, and claimed that Linux could be dangerous to Sun. Upon his words, "they may be talking a lot more [about Open Source] than they are or ever will be doing." Jonathan Schwartz's blog has been updated today with a post that is a direct response to Linus claims, but in a much more elegant and coherent way. Sun's CEO notes that "Companies compete, communities simply fracture", and tries to explain why using GPL licenses is taking so long."
Not a bad Linus message (Score:5, Interesting)
Also of note is Theo's de Raadt message in Sun's blog: "Jonathan, I wish the above was true. 15 years ago I was the biggest Sun fan. Today I speak as the project leader for another set of open source projects -- OpenBSD and OpenSSH. OpenSSH will be better known to your audience, as it is what they use daily to connect securely to and from their Solaris (or Linux) machines. OpenSSH killed telnet and rlogin, for those who still remember those mechanisms. We give our software completely freely to the world, without even the standard encumberances people see in the GPL or CDDL. Yet when we turn around and ask Sun to give us documentation for the chips on their machines -- chips Sun themselves designed, not via contractors -- Sun drags their feet. Recently we tried to reopen these 10-year-old repeated requests, and once again nothing positive happened. You may remember, because you and David Yen were in an email conversation with us. Lots of nice open words were exchanged, but no action. However, let me give an example of the duplicity of Sun. (I wish I could use a lighter word). Two operating systems run on Sun's latest PCI-e based (smallish) Ultrasparc-III machines, the v215/v245 -- Solaris and OpenBSD. The latter system runs on those machines because the code to support the non-processor chips on the board had to be written after painstaking reverse engineering, because Sun refuses to make available documentation for how these chips are programmed. Now we will readily admit that not every programmer in the world needs to know how to program these chips. But does every programmer in the world need to know how to program every little detail on Sun's processors, in system mode? Sun gets great press out of UltraSPARC being all "open", but what use is supervisor-mode documentation when the rest of the chips that the supervisor-mode code has to communicate with are entirely undocumented??? The press does not spot this problem, but Jonathan, you should clearly understand this is a fallacy. There are two operating systems which surprisingly do not run on the Sun v215/v245 -- Linux and OpenSolaris. OpenSolaris?? Yes -- Sun isn't even open enough to give the OpenSolaris community enough documentation to support their new machines. So I think that Linus is right, and Sun has a long road ahead."
I tend to listen to Theo's opinion carefully on this subjects. I'm an "FSF fanboy" to the bone, card carrying and all, which curiously is one of the reasons I tend to view Theo's opinion on this subjects with interest, more so than Linus. When it's not a GPL vs BSD thing (which is a fait-diver discussion in my sense of priorities) the fact remains that he seems to see the problems with licencing with a greater depth and in general is more "idealistic" than "pragmatic".
It's skewers time (Score:1, Interesting)
GNOME pissed of (Score:1, Interesting)
http://www.placenet.org/benoit/index.php/post/200
http://0pointer.de/blog/projects/project-indiana.
Not really a war (Score:2, Interesting)
Re:License changes take a loooong time (Score:3, Interesting)
I think you're partly right: for small machines, I've found Linux perfomance excellent, and they do have a whole bunch of good ideas. Building an ltrace (shared-library call-tracer) that can jump in and trace a running process was cool, and clearly better than Solaris apptrace (of which I was one of the three authors).
I mostly work with large data centres, and personally run SPARC Solaris except on one machine, and that one's Linux. I find them very interoperable, and I enjoy watching both Linus and the Solarii compete with each other for quality, elegance and speed (:-))
--dave
Re:Linus needs to stop speaking for Linux (Score:3, Interesting)
I think Linus is right and you are wrong on Java:
1. Sun still retains "control" over Java-the-platform through the JSR/committee process. GPL'ing the reference implementation doesn't affect their control of the trademark.
2. The Microsoft lawsuit was settled for a LONG time before Sun started talking seriously about GPL. In the meantime MS was committed to
3. The 'it factor' was in danger of permanently moving away from Java. F/OSS was picking up Mono, Ruby, and Python instead. Java's reputation as "the new COBOL" was turning it into a platform that pays the bills but is otherwise very uninteresting.
4. Once Kaffe/Sabre/Classpath/etc. were about to run Eclipse, Sun got very serious about GPL'ing the JDK.
All in all this leads me to conclude that Linus was dead on right: Sun would prefer to have a non-GPL Java, but they ran out of options. It was either GPL Java and breathe new life/freshness into the platform and deal a blow to
This is what Theo de Raadt has to say about it... (Score:3, Interesting)
Working the way RMS intended (Score:4, Interesting)
That may be the major reason for Linus's striking change of heart on GPLv3.
You have to wonder whether RMS talked to Sun at all about this. We do know that he has praised the company for the decision to GPL Java. If RMS wanted to strongarm Linux into a license change, what better way to do it than through ZFS?
Re:Schwartz's bullshit (Score:2, Interesting)
How about:
1. DTrace
2. Zones/Containers
3. User and process rights management (way better than sudo, IMHO)
4. How about binary compatibility?
I can't stress how hard it is to deliver application binaries on Linux because of incompatibilities between libstdc++.so and libc.so -- and that's on different versions of the same distribution. Try delivering those binaries to a different distro for a real nightmare. This is _never_ a problem on Solaris.
5. Stability -- Linux is stable, but I have to say that Solaris is even more so.
The file systems on our linux machines are much more susceptible to corruption during power outage than UFS or ZFS on Solaris.
6. SMF (Services Management Framework)
A lot of people don't realize how far Solaris on x86 has come in the last 3 years. It's the real deal. I encourage you to find out more, and see for yourself.
Re:Schwartz's bullshit (Score:3, Interesting)
As for binary compatibility, all major distributions support backwards compatibility for different versions of major standard libraries. If it's really an issue, just ship your own shared libraries or link statically.
I don't think anybody has numbers to back up claims about stability. As a nearly 20 year SunOS/Solaris user, I have to say, I have no confidence in Sun's ability to maintain data integrity, and Sun kernel and system bugs have caused me enormous headaches and lost work
Finally, SMF-like self-healing has been around for many years. Recently, it has become popular to throw out the old frameworks and develop new ones. Linux has done that just like Apple and Sun. So, nothing new here.
Re:Linus is not the god you think he is. (Score:3, Interesting)
I would have thought that Dijkstra's work on concurrency was directly relevant to OS design.
Given that Linus originally intended to produce a better kernel for Minix I would say that he achieved that goal. I don't think that AST would give Minix a passing grade if it was turned in for his OS class either. But at the time Minix was missing various features, and newer hardware support which AST did not want to add himself. If you consider the original goal to be "write a better kernel for Minix" then Linus succeeded. If you consider the goal to be "Write a new 386 Kernel from scratch" then Linus also succeeded. If you consider the goal to be "Write a new OS with new features and without just copying previous designs" then Linus failed.
I personally believe that the original goal was a combination of 1 and 2. I do not believe that it was 3 in any way.
Does this make Linus a god? No. Does it make him an extraordinary software engineer? No. Does it mean that he succeeded at his goal? Yes.
Z.