Can GNU Ever Be Unix?

An anonymous reader writes "The question isn't whether Linux can be certified as Unix. At least some distributions no doubt can. But who would pay for it? And is it worth the trouble? Jem Matzan asks these questions on NewsForge, and reminds us that the Open Group, not SCO, owns the Unix trademark,"

Comments

The *real* question is ...

[ggvaidya]

can Unix ever be *nix?

Seriously, for all practical purposes, GNU + Linux is setting the trend now. Ask IBM, Novell, SCO ...

Why?

[lphuberdeau]

Is there really a good reason why would GNU be considered as UNIX officially? GNU has it's own credibility. What is UNIX anyway? Does anyone have a concrete definition of what UNIX is right now (no historical reasons, not the fact that the filesystem starts with /).

Re:Why?

[k4_pacific]

The Unix trademark is allowed on anything that confroms to several standards laid out by the Open Group who owns the Unix trademark. Linux on X86 won't comply becuase some of the errno codes are incorrect, being based on Minix, which also uses incorrect values. GNU/Linux for other platforms could qualify as they are, but again, GNU/Linux seems to be evolving as its own standard which seems to be more widely supported because of the freeness and wide availability of Linux.

Who cares?

[Ianoo]

If it's "close enough", surely big business are going to do more research than just look at whether it's been certified by The Open Group just so the Linux community can use its trademark?

The problem, as well, is what to certify. There are so many combinations of kernel, drivers, libc, userspace utilities and windowing systems that any certificate could well be rendered useless.

For example, if IBM paid for SuSE to get certified, would that apply to RHEL or Debian, if they were using slightly different kernel versions or different kernel patches as is often the case?

No. And Yes.

[T-Ranger]

"UNIX" means different things to different people. One definition would be something that contains ATT UNIX code. Another would be something that has a bunch of certifications. Linux has neither.

But, the BSDs, and I believe even Solaris and AIX have a Linux compatability layer, or at the very least, "the GNU toolset", GCC, glibc, etc. Of course you wont beable to run IA32 binaries on a UltraSPARC, regardless of the compatability layer, but you could run IA32/Linux stuff on IA32/*BSD, or SparcLinux stuff on a Solaris box.

I guess Im trying to say, given that lots of things can run Linux binaries, can cleanly compile Linux targeted sources, Linux + GCC + glibc may be a better standard to target then POSIX and whatnot. It is definitly more modern. Or to put things another way, UNIX is irrelevent, the question shoud be: can UNIXes ever be Linux?

Linux doesnt need it.

[nurb432]

Its just for 'brand recognition' anyway, and Linux has that now.

If you say 'Linux' to the general IT population, they already know what you are talking about. ( and they also realize the differences beteeen it and 'unix' ) so why muddy the waters?

It's not important.

[IGnatius T Foobar]

No Linux distribution has bothered to achieve Unix branding because it's simply not important anymore. People who purchase Unix systems know what Linux is and they know it's the best and fastest growing Unix-like system out there. More importantly, they know that the applications they use have been tested on Linux, probably as a top-tier platform, often as the recommended platform (see Oracle). That being the case, why would a Unix certification from The Open Group make any difference?

Meanwhile, commercial Unix vendors are going out of their way to achieve Linux compatibility, at either the source or binary level. Linux is quickly becoming the standard to which other Unices are compared. This means two things:

• The Open Group and its branding are irrelevant
• Richard Stallman is effectively wrong: GNU is Unix. Except in the real world we don't call it GNU, we call it Linux.

Re:It's GNU/Linux!

[kbahey]

Hurd is definitely a good idea, but so far it is only that: an idea.

I have been hearing about Hurd at least since 1992 or so, ever since Linus started his project. This is 12 years now, and nothing concrete has come up yet, that can be adopted by the masses.

Don't get me wrong, I like many of the ideas and design decisions they have. But my gripe is that their model does not allow hordes of programmers to join in and get things out faster, like the Linux model.

UNIX 2003

[taj]

Is the branding alive or not?

What Unix passes Unix 2003? OK. Who passed UNIX 98? Get the picture? Its going to cost ~$0.5M when all said and done. What advantage is there? Some of the 'UNIX' systems out there have not passed a checklist in over a decade.

Linux does not need people who dont code deciding what is right and wrong in expensive ongoing beurocratic processes. Things are decided much faster in open forums which document the process in ample detail.

Linux does deviate but given a coin toss, it goes with the previous 'standards.' If the legacy means does not make sense, its ignored and documented.

The UNIX branding made sense with legacy closed source Unix systems. It provided a level of trust that customers could drop to without even (imagine!) seeing the underlying code.

It was a bandaid on a broken model. The outdated Unix systems deviated but the customer could only read documentation, not code.

So systems like Solaris, AIX, HP-UX ... have two options. Continue down the documentation/standards/branding route increasing customers costs $100's/install or just open up the developement process/source.

If they decide to open up the process, they have to decide wether to join open source projects or try to replicate the efforts.

'UNIX' is dead. Do we need a netcraft survey?

I know people are going to say that wont work. "Look at all the forks in apache and perl and python. It will be anarchy."

Thats proven to not be the case. The problem has always been closed source.

What's in a name?

[cpghost]

My first reaction was: "Why the heck should Linux be Unix-certified? With increasing popularity, Unix will soon have to be Linux-certified if it wants to get any kind of market acceptance."

Well, as amusing as it may be, this thought is flawed. First of all Linux is merely the kernel; it's not even glibc, nor any other GNU tools, or third party packages. BSDs are Unix-like OS, just like Linux(-distros) are Unix-like. Solaris is also _a_ Unix-like OS, just like HP-UX.

Actually Unix has become a generic term which refers to all kind of kernels that expose a POSIX (don't remember the exact number) interface to userland applications. Any kernel (or microkernel + servers) that implements this interface, can be justly called Unix (or at least Unix-like; so as to not feed SCOundrel or Open Group lawyers).

The really interesting thing about the hype around Linux, is when we will move on and replace the Linux kernel with something totally different (be it microkernelized, or whatever). Then, we won't have just a GNU/Linux system anymore, but, say, also a GNU/Hurd/L4, GNU/Hurd/Mach or GNU/BSD, BSD/Linux, BSD/Hurd/*, ... system (terminology being "OS personality"/"OS servers"/"microkernel" or "OS personality"/"monolithic kernel").

It seems silly to use the kernel name only as a brand for all kind of Unix-oid systems, regardless of them using the Linux kernel or something else; but providing the POSIX Unix interface.

To wrap it up: it's just a matter of names and brands. As other posters have commented before, Linux has gained enough popularity and visibility. It doesn't need to be certified to be successful!