Stories
Slash Boxes
Comments
NOTE: use Perl; is on undef hiatus. You can read content, but you can't post it. More info will be forthcoming forthcomingly.

All the Perl that's Practical to Extract and Report

use Perl Log In

Log In

[ Create a new account ]

scrottie (4167)

scrottie
  scott@slowass.net
http://slowass.net/

My email address is scott@slowass.net. Spam me harder! *moan*

Journal of scrottie (4167)

Sunday April 23, 2006
01:19 AM

Unix vendors as we know them exist thanks to the Motorola68k

[ #29415 ]

Skip this article if you don't know who HP, Sun, and NCR are.

Okay, good.

Here's an old-fart story you might not have heard: Unix exists as we know it today because a company called Motorola introduced an inexpensive, high performance chip capable of running Unix. Oh, of course, someone else might have done the same thing if they hadn't, but here's what happened.

Unix machines were big and expensive -- VAXen, specifically, and this is long before the microVax. Maybe there were ports at this early point in time -- if so, stories are welcome. And of course, when the Motorola 68000 came out, Intel still had only the 286 or 8086 or something lacking a linear, non-segmented address space. But then again, Microsoft had their Xenix, and they had ports of that to the 8086 and 80286. The 286 port actually had 16 processes, each with their own private 1m (max) of memory, done with segmentation tricks. Of course, an int was only 16 bits and you couldn't malloc more than 64k at a time. Back in those days, most hosts on the 'net weren't Unix systems -- Vaxen running VMS were quite popular, and you had MULTICS still, and a random pile of other things. Honeywell had systems running some horrible thing.

So, anyway, the 68000 comes out, and a lightbulb goes over a bunch of peoples heads. Mini computers with 68000s lurking inside come out first, doing industrial control work like disk duplication, and running cash registers for larger department stores. Departmental style Unix minis are created. Scientific computers. Then workstations. Sun Microsystems uses one. The Sun 2 would feature the 68010 and then later the 68020, and the Sun 3 the 68030, but the 68000, which lacked an MMU and couldn't accept an external one (the 68010 had a few minor optimizations -- in particular, it had a few word cache so small loops could run without hitting the memory bus -- but it's main point was to add glue to wire in an external MMU). Early 68k Unix systems lacked memory protection. NCR, the cash register company, built serveral 68k systems. These things were wired to the dumb cach registers with really long serial cables, and the machines had 16 or more serial ports on them. Of course, they'd run full scale terminals, too. Apollo started making BSD 4.2 workstations based on the 68020 (perhaps earlier), and HP bought them up and made "HP Apollo" machines for years before introducing their own PA-RISC architecture (which has since bowed to Merced and then the Xeon). Even Radio Shack, who made the popular Model I, Model II, Model III series (big grey box with 8" drives, monitor, and keyboard all in one unit), introduced a 68000 card and SysIII Unix (I think it was) for their "home"/small office computers. With two 10meg external hard drives (one for the OS, the other for user data) and SASSY (direct ancestor to SCSI) interface from Shugart. These would run small offices like summer camps (where I saw one) and newspapers and control dumb terminals plugged in over serial connections, as was the way back then. If you got to sit in front of a Unix machine, you must have been a very important research scientist, and that was exactly Sun's market right up until the UltraSparc series. Even Apple introduced A/UX, which ran on their existing machines that shipped MacOS7, and it was awesome. It ran MacOS 7 programs (OS7 was implemented as an API, but it controlled the screen, not X, and Unix was the kernel), and OS7 programs were written to do Unix sysadmin and config chores. This was, of course, 68k. Or maybe it was earlier than OS7... but it was 68k. I promise.

Oh, and how could I forget! Even Commodore -- and even Atari -- got on the Unix on 68k bandwagon. Commodore shipped a version of SysV on tape for use with their 68030 based Amiga 3000 system. I don't know if it had any special features like A/UX did, but it didn't have shared libraries, which is one of many Sun inventions. And then Atari did basically exactly the same thing a bit later for the TT040, 68040 machine. I don't remember what Atari called their version. Neither attained the legendary status of A/UX.

Oooh! I could just die with self hated for my memory loss -- how could I forget NeXT? I think they were original 68020 machines starting with the cube but were later 68030s. The cube was of course the first machine to ship with a CD-ROM drive -- back when that was a really big deal. And, of course, they ran Mach, and, of course, OS X is directly descended from NeXT's operating system, NeXTStep. That was another Unix vender that started with the 68k series, albiet ultimately a failed one.

Right about the time Sun had their Sun 3, and HP was using the Apollo name, the whole RISC storm was brewing. HP wasn't the first to market by any stretch of the imagination. IBM had their IBM PC RT, which was a 4mhz RISC spread across two chips. This was perhaps the first generally available RISC workstation machine. Sun later came out with their Sparc architecture, but Digital, the makers of the VAX, hooked up with some of the RISC pioneers and their MIPS, the "microprocessor without an interlocking pipeline system". Rather than the chip doing real-time data dependency checking to keep pipelined operations from stomping on each other, the compiler knew enough about the chip's timing details that it could insert no-operation instructions at strategic points to prevent clobbered data. Likewise, the first instruction or two (I forget) after a branch gets executed as if it were before the branch -- it's easier than invalidating parts of the pipeline. And this chip appeared in the DEC (Digital Electronics Coproration) 2100 and 3100, nick-named p-max and p-min. Silicon Graphics would hop on the MIPS bandwagon and use that chip, then later buy up the MIPS company itself and later be forced to sell it. IBM eventually made their POWER line, long before partnering with Motorola for the consumer PowerPC version of the architecture (which changed things and added things but didn't break the basic instruction set too badly, I don't think). Digital would later do Alpha, then sell out to HP, who would bury them.

It's fun to note that NeXT as a company eventually started selling NeXTStep for 386+ based PC clones, Digital had a "Rainbow" system that was x86 based, Sun has repeatedly dabbled with x86 systems including a 386 based Unix system way back when besides their more recent AMD offerings. Many vendors experimentally kicked out PC clones running Unix, DOS, or Windows, especially in the wake of the announcement of NT -- NT 3.51 was originally supposed to put Unix out of business with it's affordability, stability, security, and ease of use -- remember?

Okay, who am I forgetting? Who else had 68k machines or early RISC architectures actually used in machines? If you look at compilations of microprocessor data sheets from the time, there were piles of 32 bit RISC chips hitting the market trying to capture the Unix market share -- but they all fizzled in favor of ones developed in-house -- except for Intel's offerings. Intel made the i960 and i860, which would see some use, but the i960 would wind up in dedicated X terminals (dedicated machines with ethernet that did nothing but act as an X workstation for programs run on a Unix server elsewhere on the network). The Clipper (the first of three times the name has been used for a chip, so far -- second was in reference to a government mandated encryption chip that you had to buy if you wanted to do encryption -- thank god that didn't last -- no, thank the EFF that didn't last). Maybe my mind is playing tricks on me, but I remember Intersil doing a Prism chip. Intersil would later do a wifi chipset called Prism and change their companies name to whatever it is now (something lame with lots of X'es an crap).

After RISC took off basically without an Intel chip (despite Intel's efforts) in any major brand machine, Intel introduced the 386, which gave the x86 architecture a linear 32-bit address space -- and memory management! Another bread of companies, such as Sequent and Unisys, would build and sell Unix machines based on the 386. All in all, these companies were far less user friendly and treated Unix as a very serious games with lots of contracts and secrecy, and both have since all but vanished, fitting with Unix going mainstream. SCO also made a version of SysV for PC compatable machines sold by other companies (PC clones) and this proved quite popular -- for quite a while.

Then there's the near history. I won't even start down that road. Okay, I'll start down that road, but that's it. Minix happened, and it got pirated like crazy, and eventually went not quote open source (the code was distributed in conjunction with the Operating System Concepts book) but kinda free in that you could copy it without buying the book, and the demand for Unix for the hobbyist was demonstrated to be great. BSD was ported to the 386 by the Jolitzes and that forked. Linux came out, which was absolute garbage, but people saw it as being "thiers", and "from scratch" they thought was a good thing (they were idiots, of course, and still are), and BSD being mired in suits with AT&T over the original AT&T Unix code still thinly scattered here and there didn't help. Everyone went RISC, as Unix vendors, and RISC and Unix became almost synonymous, until more recently where the PC architecture and Intel chips have all but surplanted high-end workstations. Unix system integrators and system manufacturers have taken a hell of a beating, and now the PC platform is the norm -- especially in light of Apple hopping on the Intel bandwagon. Few Unix vendors joined the fray since the original rise of RISC.

And a bunch of other stuff...

-scott

The Fine Print: The following comments are owned by whoever posted them. We are not responsible for them in any way.
 Full
 Abbreviated
 Hidden
More | Login | Reply
Loading... please wait.
  • Good article, but a couple of things are out of sync with what I remember.

    Radio Shack Model II ran Unix and had 8 inch drives; the III ran TRS-DOS and had 5.25 drives. The I didn't have any drives at all (and hence no DOS) unless you ponied up and bought the external interface. I had a Model I and my best friend had a III, both of these computers used Zilog Z80 CPUs.

    John

    • Thanks John. I'm pretty fuzzy on these as I've only had limited exposure and I was young. I'm pretty sure they were all Z80 machines unless you bought the 68k card, and then I have no idea which models exactly actually took the card.

      Glad you enjoyed it.

      Cheers,
      -scott
  • Moto 68k systems were awesome. The 68k built the Mac and the Sun empires. And lots of me-too systems that everyone's forgotten. For lazy folks who didn't like segment registers, they were simple to deal with. Those of us who thought operating system should reject writes to the wrong parts of memory always thought flat address spaces were a problem. History has alas proved us right far too late.

    But ... back in the day, the 68000 was a me-too chip, just http://www.jargon.net/jargonfile/b/bagontheside.html [perl.org]"h

    --
    Bill
    # I had a sig when sigs were cool
    use Sig;
    • Holy cow!

      Thanks very much, Bill, for contributing this -- complete with references. Of course, a whole article could be written on Xenux, but I'm not qualified, and Wikipedia probably already has it ;)

      Cheers,
      -scott