You could say that Linux is AT&T's fault

January 16, 2018

Recently on Twitter, I gave in to temptation. It went like this:

@thatcks: Blog post: Linux's glibc monoculture is not a bad thing (tl;dr: it's not a forced monoculture, it's mostly people naturally not needlessly duplicating effort)

@tux0r: Linux is duplicate work (ref.: BSD) and they still don't stop making new ones. :(

@oclsc: But their license isn't restrictive enough to be free! We HAVE to build our own wheel!

@thatcks: I believe you can direct your ire here to AT&T, given the origins and early history of Linux. (Or I suppose you could criticize the x86 BSDs.)

My tweet deserves some elaboration (and it turns out to be a bit exaggerated because I mis-remembered the timing a bit).

If you're looking at how we have multiple free Unixes today, with some descended from 4.x BSD and one written from scratch, it's tempting and easy to say that the people who created Linux should have redirected their efforts to helping develop the 4.x BSDs. Setting aside the licensing issues, this view is ahistorical, because Linux was pretty much there first. If you want to argue that someone was duplicating work, you have a decent claim that it's the BSDs who should have thrown their development effort in with Linux instead of vice versa. And beyond that, there's a decent case to be made that Linux's rise is ultimately AT&T's fault.

The short version of the history is that at the start of the 1990s, it became clear that you could make x86 PCs into acceptable inexpensive Unix machines. However, you needed a Unix OS in order to make this work, and there was no good inexpensive (or free) option in 1991. So, famously, Linus Torvalds wrote his own Unix kernel in mid 1991. This predated the initial releases of 386BSD, which came in 1992. Since 386BSD came from the 4.3BSD Net/2 release it's likely that it was more functional than the initial versions of Linux. If things had proceeded unimpeded, perhaps it would have taken the lead from Linux and became the clear winner.

Unfortunately this is where AT&T comes in. At the same time as 386BSD was coming out, BSDI, a commercial company, was selling their own Unix derived from 4.3BSD Net/2 without having a license from AT&T (on the grounds that Net/2 didn't contain any code with AT&T copyrights). BSDI was in fact being somewhat cheeky about it; their 1-800 sales number was '1-800-ITS-UNIX', for example. So AT&T sued them, later extending the lawsuit to UCB itself over the distribution of Net/2. Since the lawsuit alleged that 4.3BSD Net/2 contained AT&T proprietary code, it cast an obvious cloud over everything derived from Net/2, 386BSD included.

The lawsuit was famous (and infamous) in the Unix community at the time, and there was real uncertainty over how it would be resolved for several crucial years. The Wikipedia page is careful to note that 386BSD was never a party to the lawsuit, but I'm pretty sure this was only because AT&T didn't feel the need to drag them in. Had AT&T won, I have no doubt that there would have been some cease & desist letters going to 386BSD and that would have been that.

(While Dr Dobb's Journal published 386BSD Release 1.0 in 1994, they did so after the lawsuit was settled.)

I don't know for sure if the AT&T lawsuit deterred people from working on 386BSD and tilted them toward working on Linux (and putting together various early distributions). There were a number of things going on at the time beyond the lawsuit, including politics in 386BSD itself (see eg the FreeBSD early history). Perhaps 386BSD would have lost out to Linux even without the shadow of the lawsuit looming over it, simply because it was just enough behind Linux's development and excitement. But I do think that you can say AT&T caused Linux and have a decent case.

(AT&T didn't literally cause Linux to be written, because the lawsuit was only filed in 1992, after Torvalds had written the first version of his kernel. You can imagine what-if scenarios about an earlier release of Net/2, but given the very early history of Linux I'm not sure it would have made much of a difference.)


Comments on this page:

I think it bears mentioning that 4.3BSD Tahoe was the reason that 386BSD became possible, as that was when the code base was split into machine independent/machine dependent code (without that, BSD probably would've stayed on the VAX) - in fact it's demonstrably provable in two ways. So it's not entirely fair to say that Linux predates 386BSD (especially considering Linus has said something to the effect of "If BSD had been available when I started on Linux, Linux would probably never had happened.", and all he would've had to do was add x86 code to 4.3BSD NET/2 as that already contained MI/MD seperation and virtual memory or use BSD/386 Alpha, which were released on 1991-06-28 and in December, 1991 respectively).

I'm also not sure I agree that Linux would've been unaffected by USL/Novells IP claims (it was filed by USL, settled under Novell), if it hadn't been settled out of court or had been tried in a Texas court rather than a Californian one - but it's a moot point, since that's thankfully not how history turned out.

I seem to recall that I've heard McKusick say, in one of his BSD History talks, that during the preliminaries at which he was present, it turned out that quite a lot of BSD-original code had found its way, stripped of any BSD licenses, back into UNIX.

By Christopher Barts at 2018-01-16 13:23:36:

If you have the right kind of mind, you can see SCO's attempted assault on Linux as being a bizarre parody of AT&T's assault on the BSDs, with the difference being that nobody really took that lawsuit seriously in terms of casting doubt on Linux's provenance and so there was no concomitant cloud cast on Linux's existence or continued rise in the server world.

By Jim at 2018-01-16 20:25:02:

Any what-iffery about the advantages of Linus Torvalds not launching Linux should also consider the loss of git, which he arguably would not have been motivated to code without a major distributed project driving it.

Of course, the debate over git vs. Mercurial (with BitKeeper playing the "AT&T" villain and CVS being ... Minix?) also swims around such religious topics as licensing, reinventing wheels, and so forth, so perhaps that's not a productive line of thought.

Written on 16 January 2018.
« Meltdown and the temptation of switching to Ryzen for my new home machine
My new Ryzen desktop is causing Linux to hang (and it's frustrating) »

Page tools: View Source, View Normal, Add Comment.
Search:
Login: Password:
Atom Syndication: Recent Comments.

Last modified: Tue Jan 16 00:07:10 2018
This dinky wiki is brought to you by the Insane Hackers Guild, Python sub-branch.