MS-DOS was *not* an illegal clone of CP/M

[liam_on_linux]

"MS-DOS" is the MS version of what IBM sold as PC DOS. Microsoft produced that on very short notice by licensing (note, not buying) 86-DOS from Seattle Computer Products. That was originally called QDOS, for Quick’n’Dirty OS.

Tim Paterson wrote QDOS based on studying the docs for CP/M and CP/M-86. It was API compatible, but used a different disk filesystem: Paterson used the FAT format of MS’ standalone disk BASIC.

It was wholly-new code, but written to be closely compatible with DR’s published info about CP/M.

That is not even reverse-engineering. Indeed CP/M-86 was released late and it didn’t even exist to be reverse-engineered yet, AFAIK. QDOS was written for and sold with SCP’s 8086 cards in 1979; CP/M-86 did not ship until 1981.

Writing compatible code to a published API is what APIs are for. That’s why the info is published.

QDOS wasn’t a clone of CP/M-86; in fact, it is older than and predated CP/M-86.

It was a compatible OS written to info DR published. That is entirely legal. DR published the APIs intending this for app writers, not for people writing OSes compatible with DR OSes, but it’s not breaking any rules.

In fact in the late 1970s there were lots of CP/M clones out there, such as CPN and Cromemco CDOS and many others. Later MSX-DOS was a much-enhanced CP/M clone.

The difference is, most other companies cloned CP/M on 8080 or Z80. SCP did it on 8088/8086.

But while yes, it’s arguably something like a clone (for different hardware, with a different file system), it was just one of many and didn’t use anything illegal or violate any licenses.

The key thing is that QDOS ran on then-modern hardware with a future. Most of the others ran on what was rapidly becoming obsolete hardware. SCP QDOS became 86-DOS became PC DOS and MS-DOS, and sold in the tens of millions of copies, and made MS huge amounts of money.

DR and IBM made big bad mistakes and it cost them dominance of their industries and lots of money. MS was smart and got lucky and got very very rich.

Later on, MS abused that power repeatedly, stole code, copied ideas, unfairly pushed rivals out of business, and generally became a bully and a criminal. MS effectively killed Be, Netscape, and Central Point Software; it crippled Aldus and STAC; and many more.

But DR survived and briefly it staged a successful comeback, before being bought by Novell.

I entirely understand how angry Dr Gary Kildall was. It was justified. But he did make mistakes. Sadly some of them are only clear in hindsight. DR should have rushed to make CP/M-86 quickly for IBM, and reserved the rights to sell it to others, as Microsoft did. DR should have sold single-user single-tasking CP/M-86 cheaply, building the market, and made Concurrent CP/M the premium product. It should have sold GEM cheaply to get wide adoption. It should have made standalone single-user multitasking CP/M a desirable power-user OS, rather than aiming at the multiuser market, which was on the way out as PCs got cheaper and cheaper.

But as little as I personally like MS, in how it cornered the market and became rich, it did it by being clever, and fast, and outmaneuvering bigger, slower rivals, and there’s nothing wrong with that.

Comments

[history_monk]

You've never done any GEM programming, have you? When you use the GEM desktop, you're using almost all of GEM's capabilities. It's really quite remarkable how little is there, and it isn't at all surprising that the few successful GEM-based applications (notably the early versions of Ventura Publisher) came with their own extended versions of GEM.

The world is actually better off with Windows as a dominant GUI than it would have been with GEM. It would be better yet had it been X11, but you can't have everything.

[liam_on_linux]

Oh, I'm sure you are absolutely right!

No, I only played around with GEM, but oddly enough, for a little personal side-project, I've been playing with it again recently.

I completely agree re us being better off with Windows. TBH, that applies to OS/2, too, and I really liked OS/2 2 back in the day. But even the unfinished beta of Win95 was so much better it was hard to believe. Far easier to set up, far better UI, far better compatibility... and TBH the multitasking wasn't as far behind as OS/2 zealots claimed. I could crash OS/2 2.0 easily but Win95 was very good for its time and for its technical limitations.

I'd love to read reports from alternate universes where some stuff I've idly speculated about on here came true.

Where IBM targetted OS/2 1 at the new 386 chip and every DOS power used wanted it. Where Windows never happened, but maybe OS/2 NT did happen.

Where DR sold CP/M-86 for $40 and MS-DOS remained niche. Where up against late-1980s OS/2 was a multitasking Personal CP/M Plus with X/GEM-based tiling DOS windows.

Emboldened by this it also did a proper multitasking Concurrent CP/M-68K which Atari sold in a cheap multitasking VME-slot 68030 workstation.

Where the GNU Project adopted the BSD-Lite kernel and had a practical, working, and Free xNix OS by 1989 or so, so Linux and the HURD and Free/Open/Net/Dragonflt BSD never happened.

(But Andy Tanenbaum pressed on with microkernels anyway and did amazing stuff.)

Where, after Apple bought NeXT, Be went to Acorn, and the merged result started selling silent, passively-cooled multiprocessor ARM workstations with PCI slots and a pre-emptive SMP OS at the end of the 1990s... and slim and light ARM laptops with no need for extra cooling and battery lives of many hours.

Down at the low end, Sinclair adopted Timex-Sinclair's better ULA for the Spectrum 128 and later hired Alan Miles and Bruce Gordon from MGT, reinvigorating the Spectrum world and successfully selling fancy but cheap 8-bit games machines for half the price of all the American 68000 boxes.

Dream dream dream, dreeaaammminnnnggg...

[history_monk]

"Where IBM targeted OS/2 1 at the new 386 chip and every DOS power used wanted it. Where Windows never happened, but maybe OS/2 NT did happen."

That requires IBM to have been willing to make the PS/2 models 50 and 60 significantly more expensive, by giving them 386 processors (rather than 286) and more RAM (4MB rather than 2MB). They promised large, lawsuit-prone customers that those machines would run OS/2, so they had to deliver on that. Even if they did that, a falling-out between MS and IBM was more or less inevitable, because the company cultures were so different.

"Emboldened by this it also did a proper multitasking Concurrent CP/M-68K which Atari sold in a cheap multitasking VME-slot 68030 workstation."

That requires Jack Tramiel to have died or become incapable of running Atari in about 1986. He was very focussed on consumer systems, and wasn't very interested in flexibility. I worked for Perihelion Hardware while they were trying to turn the Atari Transputer Workstation into something saleable, as opposed to demonstrable, and Jack did not seem interested.

"Where the GNU Project adopted the BSD-Lite kernel and had a practical, working, and Free xNix OS by 1989 or so, so Linux and the HURD and Free/Open/Net/Dragonfly BSD never happened."

Not all that different from today, really. This OS would have got established earlier than Linux did (and would have killed off Xenix and Solaris/i386), but it could not become dominant until x86-64 hardware became available, and Linux was ready for that when it happened.

My favourite alternate was something I suggested to my Intel customer engineer when they announced that they were going to soft-pedal Itanium for a couple of years "until the manufacturing technology caught up with it. But this was not the end! Itanium would be back, and this time it would rule the world!"

What they needed, I suggested, was a 64-bit processor that could be built with today's technology, and had room for growth: "Itanium Pro," formerly known as Alpha EV7. Intel owned all the Alpha IP, because Compaq had sold it to them. My engineer reckoned this was funny, but he wasn't going to suggest it to marketing.

[liam_on_linux]

That requires IBM to have been willing to make the PS/2 models 50 and 60 significantly more expensive, by giving them 386 processors (rather than 286) and more RAM (4MB rather than 2MB)

The ones I worked on -- quite a few of 'em -- came with only 1MB.

I have idly speculated on this before. The 50 & 60 sold well: they were relatively affordable, for PS/2s, but they were proper ones, not like the Model 30 & 30-286, which weren't real PS/2s at all, being ISA-based.

None of the customers I dealt with wanted OS/2 at all. In the longer run, IBM could have given free planar upgrades to every PS/2 50 & 60 owner who bought OS/2 & a RAM upgrade. It would have made more of a profit from the more successful OS.

(There are some wrinkles to work out here: how to limit it so that freeloading chancers don't order OS/2 just so as to get a free upgrade. But IBM was smart. They could have done it.)

IBM crippled the product, and in the end destroyed its own PC business and handed the OS market to MS, just to honour a pledge its customers didn't care about.

In the IT press at the time, there was speculation as to which partner wanted it to run on the 386, but in hindsight, the answer is clear.

Because MS was already working on 386 versions of the OS. Decades later, they leaked.

The original MS-DOS 4, the multitasking version, released only in Europe: https://www.os2museum.com/wp/multitasking-ms-dos-4-0-lives/

A Goupil version: https://www.os2museum.com/wp/multitasking-ms-dos-4-0-goupil-oem/

A prototype, "Sizzle", with 386 code: https://www.os2museum.com/wp/before-os2-was-os2/

Which became the 386-specific "Football": https://www.os2museum.com/wp/playing-football/

It's out there.

MS was working on adding 386 enhancements and support for 386 V86 multitasking to DOS, using prototype Compaq kit.

In the light of that, I think it's easy to deduce which of the 2 partners wanted the new OS to run on the 286.

To be fair it is only in hindsight that it is so clear which was the right direction. At the time, it wasn't so clear. The 286 was called a superchip when it came out, as I reference here: https://www.theregister.com/2022/08/04/the_many_derivatives_of_cpm/

Direct link to InfoWorld: https://books.google.cz/books?id=ZjAEAAAAMBAJ&lpg=PA4&pg=PA4#v=onepage&q&f=false

(Reversing order for clarity)

That requires Jack Tramiel to have died or become incapable of running Atari in about 1986. He was very focussed on consumer systems, and wasn't very interested in flexibility.

Well that might be so, but I had a specific machine in mind, and it was the 1990 Atari TT: https://en.wikipedia.org/wiki/Atari_TT030

The OS that I was thinking of was MiNT: https://en.wikipedia.org/wiki/MiNT

Which later shipped with the TT replacement, the 1992 Falcon: https://en.wikipedia.org/wiki/Atari_Falcon

... which I submit shows that Atari had some awareness it needed a better OS.

I worked for Perihelion Hardware while they were trying to turn the Atari Transputer Workstation into something saleable, as opposed to demonstrable, and Jack did not seem interested.

If I knew that, I had forgotten! Did you see my article about Parhelion HeliOS?

https://www.theregister.com/2021/12/06/heliosng/