% fortune -ae paul murphy

Linux is Unix

Two weeks ago frequent contributors p_msac and bportlock challenged me to see Linux as not Unix and to discuss the consequences of that difference.

The reality here is simple: Linus Torvalds started out by hacking the Minix kernel to use x86 interupts, graduated from there to announce his project as "a simple minix-lookalike" in August of 1991 and followed that with the December 0.11 Freax release and his decision to compete with Gnu in the creation of a "free Unix for the 386".

Since he acknowledged the Unix heritage until just after the SCO lawsuit pushed him solidly into IBM's camp and the thing still looks like Unix, acts like Unix, and runs Unix software, I think he should be given full credit for doing exactly what he said he would: creating a free Unix for the 386.

Again setting aside legal issues afflicting the word, Unix is neither a brand nor a particular product: it is a set of ideas - implemented differently by the people behind Linux, the BSDs and Solaris, but all with the same basic structures expressing the same basic values and ideas, and all useful for running the same utilities and applications on the same hardware.

I believe, however, that the distinction Pedro and others draw isn't really between Linux as Unix or Linux as somehow not Unix, but between Linux as they use it and Solaris as they've seen it misused by corporate data processing managers.

In other words I see their usage of "Unix" as meaning something like "controlled by aliens" where Linux means "free as in mine" - theirs to do with as they wish.

A lot of this difference comes from two closely related factors: history and scaling. Thus one reason a lot of people associate Solaris with repression is that Solaris on SPARC is the only remaining member of the ninties choices open to the people whose mistakes they associate with Unix: System VR3 (and thus AIX) never became sufficiently scaleable, Tru64 on Alpha is functionally dead, and HP-UX on PA-RISC has turned into a spectrum of its former self.

Since then x86 has progressed rapidly - going from 200Mhz for the Pentium Pro in 1996 to the dual core 3.2Ghz, Xeon in just ten years. As a result you can now use a PC running Linux to do jobs that only ten years ago required Solaris on a half million dollar Sun 6500 - or HP-UX on a million dollar K-580.

Unfortunately, many of the data processing managers whose actions are responsible for the poor reputation Unix has among some haven't adapted - and are now happily licensing Red Hat for use in typical Wintel or data processing "server" racks that will, I guarantee, produce comparably adverse reactions in the next generation of systems technologists to come along.

That's sad - but I'd argue that those who continue to draw the distinction as between Linux and Solaris rather than as between idiot managers and IT leaders haven't updated their thinking either - because in those days the combination of Linux on increasingly cheap x86 power seemed to be the best bet for the rebellion, but today there isn't much (if anything) x86/Linux can do that Solaris for x86 can't do better - and Solaris on SPARC can handle high end jobs where x86/Linux doesn't have a look-in.

There are exceptions: IBM's decision to adopt Linux as the key interface technology for Cell has combined with x86 backward compatibility for Linux code to push Linux to the top of every super computer manager's wish list - and the somewhat desperate efforts to cash in on this by what's left of SGI should certainly qualify for some kind of Darwin award - but none of this affects the broad reality: more people do more things with Linux, not because it's better than Unix, but because it's exactly what Torvalds wanted it to become: a free Unix for the cheapest, lowest end, hardware around - the x86 stuff.


Paul Murphy wrote and published The Unix Guide to Defenestration. Murphy is a 25-year veteran of the I.T. consulting industry, specializing in Unix and Unix-related management issues.