Slashdot Mirror
SCO vs. IBM Battle Over Linux May Finally Be Over (networkworld.com)
JG0LD writes with this news from Network World: A breach-of-contract and copyright lawsuit filed nearly 13 years ago by a successor company to business Linux vendor Caldera International against IBM may be drawing to a close at last, after a U.S. District Court judge issued an order in favor of the latter company earlier this week.
Here's the decision itself (PDF). Also at The Register.
Here's the decision itself (PDF). Also at The Register.
Have you ever used SCO?
I have. It wasn't a bad system. I didn't like it as well as Solaris, but it was stable and reliable and pretty well documented. For a long time, they had a good product and supported it pretty well.
Yeah, the company sucked, but all that work, good programming, is now going to waste. What I'd like to see is IBM take ownership and open source all of it, have it relicensed under GPL and MIT licenses. Ultimately, I'd like to see a lot of that code legally incorporated into Linux.
Why? Just to make the people responsible for the fiasco the lawyers and executives of the company SCO weep.
PJ and Groklaw would be a huge boon for slashdot if you could somehow reach out to her and bring her back.
Jesus... even Duke Nukem Forever was finished before this lawsuit was!
SYS V needs to go open next, not that overloaded slowlaris, but lean mean SYS V
I was under the impression that the entire POINT of SYS V was for the major UNIX vendors to re-implement the guts of Unix as a clearly, enforceably, proprietary product (after the CONTU recommendations and the resulting copyright law changes explicitly extended copyright to software), then move to it and orphan the original development thread. (This might make opening it a hard sell to the members of the consortium.)
There were at least a couple issues with the proprietary status of the AT&T code:
One issue was that AT&T was still a government-regulated utility monopoly and there were some requirements about disclosing and releasing non-telephone-related inventions they came up with.
The big issue was that, before copyright applied and before software patents were hacked up (by recasting software as one embodiment of, or a component of, a patentable machine or process), the only protection was trade secret and the related contract law. Trade secrets generally stop being enforceable when the secret out of the bag (with some details about whether the claimant contributed to the leak). Bell Labs had shipped code to a LOT of educational institutions. When the U of New South Wales used the System 6 kernel code and an explanation of it as the two-volume text for an Operating System class, the textbooks became an underground classic. This, along with AT&T's benign-neglect licensing policies, led to the burst of little, cheap, generic UNIX boxes, as this was also when microcomputer chips were just becoming powerful enough to do the job.
Up to then a big barrier to entry was that every new machine needed a custom O.S. to deploy, and these were enormous, machine specific, and mostly in assembler. That made it an expensive, undertaking, suitable only for financial giants. But all but under 2,000 lines of Unix was in C, and the entire kernel, which included essentially all the platform-specific code as a subset, was well under 10,000 lines of code. If you had a C compiler and assembler for your new machine, it was a matter of a few man-months to port it and get it up and running. Essentially ALL the utilities and applications came right over. You didn't have to train users, either, because they all worked pretty much just like what they'd used in college.
The game was:
1. Grab a bootleg copy of the code.
2. Port it to your machine and get it working.
3. Go to AT&T and ask for a license "to port Unix to our new machine and sell it."
4. AT&T, as a matter of policy, completely ignores any "violations" you may have committed during the porting phase and cuts you a license at a very reasonable price.
5. You "port Unix in an AMAZINGLY short time" (like the ten minutes it takes to tell Sales to go to market) and you're in business.
6. You (with your new business) and AT&T (with their small cut) slap each other on the back and laugh all the way to the bank. PROFIT! for you. (profit) for AT&T.
7. Because of the policy in 4., everybody ELSE manearly everbody's king a new machine knows they can do the same thing. So many do. AT&T gets a rakeoff from ALL of them. PROFIT! for AT&T. Far more than if they went dog-in-the-manger, held up the first few for all the traffic would bear, and got no more customers for Unix.
And because of this, it was in nearly everbody's interest to NOT challenge the AT&T-proprietary status of Unix. And it stayed this way until SCO's management screwed up and altered step 4. (Even then the case turned on other issues, so it never did come to the point of attacking AT&T's claim that Unix code was proprietary.)
Bantam Dominique roosters crow a four-note song. Once you've heard it as "Happy BIRTHday" you can't NOT hear it that way
Some Australians gave it a go because if it succeeded then SCO employees in Australia could be charged with "demanding money with menaces". SCO refused to deal.
Slashdot headline: "May Finally Be Over"
Referenced source: The Register.
from The Register article: "the case isn't over"
I'm really confused as to why people hate systemd so much. Based on the negative reactions on Slashdot, I expected systemd to be unstable and bloated, and was thinking about perhaps migrating back to Gentoo or FreeBSD if the rumors were true. However, during the past year I've tested systemd with first Kubuntu 15.10 and then Arch Linux, I've had no trouble with it at all. On the contrary: the bootup was lightning-fast compared to previous systems, and everything just worked out of the box. Taking a look "under the hood", everything looked neat and clean as well: system services are configured through readable config files, that are much shorter and tidier than the typical SysV init scripts I've gotten used to. Most of the design choices make sense as well: I see no reason to keep daemons like e.g. initd and inetd separate on a modern system.
I've also read that systemd apparently saves a lot of work for e.g. the distribution maintainers and desktop environment programmers, in the first case since it is much easier to maintain a systemd service file than a SysV init script, and in the latter case because a lot of work that previously had to be redone for every Linux distribution can now be easily shared or ported between distributions. I don't think some homogeneity among base systems are a bad thing if it makes it much easier to make software work across distributions. For instance, almost nobody's complaining that using the Linux system with the GNU base system and X11 display server is bad because it makes the Linux ecosystem too homogeneous. Sure, you do have legitimate usecases for the BusyBox base system and framebuffer applications, but that's not the majority of Linux desktop systems. There will of course be legitimate usecases for other things than systemd, but I don't believe that holds for the majority of Linux desktop systems either.
The only criticism of systemd that I agree with, is that plaintext log files are a good thing. I think I understand the reason for having binary logs (making it easier to parse for programs and scripts without a making a regex-hell), but in that case it would be much saner if journald automatically transcribed the logs it generated to plaintext files as well. Apparently it is not too difficult to set this up yourself, but I still think human-readable logs should be default.
I miss the simplicity of the bsd-like init config scripts sitting on top of SysV in Arch, before they adopted systemd. So much could be configured from rc.conf, the daemon commands were simple, and I never had problems booting. gah
Yielding the power of UNIX has always laid in creating your systems inittab file, I thought everyone did that. I used to look upon rc scripts as an unnecessary complication of the system and wondered why they were there. If a service needs to be up, init makes sure it's up. If you want to take the service down, tell init to take it down. vim /etc/inittab && kill -1 1 then get on with the rest of your day.
Network services, is good example, a shell script handled by rc, is a prime candidate for an init controlled service. Getting init to kick of printer services after a short delay so that CPU time is focus on providing a GUI to a user could be another. Messaging system is a perfect example.
What about using runlevel 4 for your customised system state, 3 for shell level maintenance, 5 for GUI level maintenance? How about an ondemand runlevel?
Just learn how to use init *actions*, which is a lot simpler than systemd. Simple, scarily powerful and totally under utilised in Linux.
After spending some time with systemd writing unit files and playing around with jounalctl my sense is that this entire situation would be resolved with a set of small tools that manipulate inittab files properly that could support a GUI based inittab editor, thus complementing/completing the original design philosophy with a small maintainable set of tools that rpm, yast, apt-get could utilize. I wonder if people would be interested in such a thing? Perhaps it's time to contact the Devuan people.
I can agree with systemd supporters that the rc system is crap, however that still isn't init and systemd is as monolithic as the rc system, except it's compiled. I think the main objection to the idea of systemd is init is a core idea of the UNIX Operating System that is powerful. I've never seen a Linux distribution that uses init properly and essentially the argument is to replace a core idea of a stable operating system platform because people just don't understand how to use UNIX's most powerful concept one step removed from the kernel. Fast and lean!
The funny thing is, after all these years, I still haven't got everything I can get out of init. Do you understand what you are destroying systemd guys?
Has anyone got a use case yet?
My ism, it's full of beliefs.
It's not just systemd that is destroying SysV. It's also Solaris, OS X, PC-BSD, and FreeBSD -- we can probably say well over 90% of the major Unix platforms. For better or worse, OpenRC is also making many of the same choices as systemd, including heavy dependence on C libraries, dependency resolution, parallel startup, and cgroup support. The critical failure of SysV init is the pidfile. It was always a bad hack — perhaps necessary for cross-platform support in a world without real process tracking, but now there isn't a hell of a lot of competition in any given market segment, and cross-platform support is being seen as less important than being able to accurately manage services. Yes, pidfiles almost always match the service they are supposed to, but this is not something that should ever have been left to userland. Similarly, daemonizing a process should never have been left up to script writers, given that glibc doesn't even do it correctly.
So now we're putting process tracking back into the kernel and it's a breaking change. If cgroups had been part of the POSIX standard years ago, systemd would have attracted no more attention than upstart when it was released. It *is* init, it has a superset of the same features, only with an event-driven model, and a slightly more sensible approach to dependency resolution than upstart. You may not have reached the limits of inittab, but other people with different use-cases have, and there's nothing in particular wrong with either case. The argument with systemd isn't an argument about how best to manage services, it's about technical debt so entrenched that people think that's the way it's supposed to be.
Those who advocate genocide deserve every protection afforded by law, and none afforded by common human decency.