It seems that every version of MS-DOS from v2.0 onward was actually developed by Compaq. I had no idea.

> That relationship had been established in late 1982. Back then, Gates had contacted Canion and asked, with some concern, if Compaq was trying to get into the operating system business. Surprised, Canion denied it. Gates told him that Microsoft was hearing worrying reports from the dealer network. People were buying copies of Compaq DOS, rather than Microsoft DOS, without buying a Compaq PC.

> Both men knew why: Microsoft DOS had never been a true copy of PC DOS, as Gates had admitted to Canion during the development of Compaq’s first machine. The differences had only increased over time, as Microsoft’s deal with IBM prohibited the same developers working on both versions. Compaq had made its own version of DOS since the beginning. With its singular focus on 100 percent compatibility, the result was a product that was more compatible with PC DOS than Microsoft’s own product.

> Word was spreading among computer buyers that Compaq DOS was better. Even people who owned other PC clones were choosing to buy that instead of Microsoft’s own public version. This could have created friction between Compaq and Microsoft. Instead, Canion did something extraordinary. Compaq withdrew Compaq DOS from sale unless it was specifically bundled with a Compaq computer. He then licensed Compaq DOS back to Microsoft.

> From Gates’s perspective, this was an incredible deal. He was able to halt all internal development on Microsoft DOS, saving time and money. From this point onward, every version of Microsoft DOS he sold was, in fact, Compaq DOS, with the digital equivalent of its serial numbers filed off. All Canion asked in return was that Microsoft never release the very latest version of DOS that Compaq provided it until after a few months’ delay. This was to make sure that Compaq always had a slight advantage in compatibility over its rivals.

> Canion even agreed to Gates’s request that they keep the entire arrangement secret, to avoid souring Microsoft’s relationships with the other clone companies. It would remain secret for almost 40 years.

I wonder if this is slightly mangled over the years?

MS-DOS wasn’t a specific IBM PC OS, it was designed to be generic 8086 OS. It was up to the OEM to adapt the machine specific code (mostly in IO.SYS / IBMBIO.COM) to their system. IBM owned IBMBIO.COM, and some of the utilities like MODE and FDISK, and early on Microsoft didn’t have its own implementations to offer for people building generic PC clones. You had to write your own, and hope they were compatible. Microsoft did eventually offer a generic MS DOS with an IBM PC type IO.SYS and reimplementation of the utilities, so perhaps those are descended from Compaq’s versions?

This is largely untrue. MS-DOS and PC DOS were both built out of the same codebase for most of the time MS-DOS was on the market. They didn't diverge until some point after DOS 5 (1991). If you look at the MS-DOS source code, you'll see an IBMVER define that controls whether to build IBM PC DOS or generic MS-DOS. The option was there in the first place because Microsoft initially sold MS-DOS to OEMs as a generic 8086 operating system. In the early 80s there were a bunch of computers that ran MS-DOS, but weren't PC compatible. These largely died out as time went on. DOS was barely an operating system, so there were a lot of programs that bypassed it and accessed the hardware directly. None of those programs worked on non-PC clone MS-DOS computers, so most consumers chose PC clones for the far greater software library. What is true is that unlike most early 80s OEMs, Compaq built their version of MS-DOS with IBMVER set, so Compaq DOS was closer to PC DOS than many other early DOSs.

As time went on and Microsoft started selling more and more copies of MS-DOS to manufacturers making PC clones, MS-DOS and PC DOS grew closer together, not further apart. MS-DOS 3.2 (1986) was the first version that Microsoft made available in a packaged form to OEMs that were shipping PC clones (it was still available as source code to OEMs who wanted more customization). For previous versions, OEMs were required to write their own versions of several hardware specific utilities (such as FDISK.COM and MODE.COM) that were originally written by IBM. MS-DOS 3.2 contained Microsoft-written clones of these utilities. MS-DOS 3.3 (1987) was written entirely by IBM (not Compaq!), as most of the key members of the MS-DOS team were busy working on OS/2. Because it came out after the Microsoft-IBM joint development agreement, Microsoft gained the rights to the IBM-written utilities and the packaged versions of MS-DOS 3.3 and all later versions shipped with the same utilities as their PC DOS counterparts.

The closest thing to the claims this article makes is that Compaq did in fact maintain their source code license to MS-DOS and made enhancements to their version. The most notable example of this is that Compaq DOS 3.31 introduced a modified version of FAT16 that supported partitions larger than 32MB. I assume Compaq licensed this functionality back to Microsoft, as there's versions of DOS 3.31 branded for other OEMs, and support for Compaq's modified FAT16 (known as FAT16B) was included in MS-DOS 4.0.

> It seems that every version of MS-DOS from v2.0 onward was actually developed by Compaq. I had no idea.

The article is wrong about when this occurred—Compaq DOS wouldn't have been in stores in 1982; 1983 is likely the correct year—but regardless, this is an astounding revelation.

This is too funny. I got down voted to hell for talking about how Gates never really made anything and was a lucky conman who managed to make his cons a reality by the skin of his teeth, and here we have further proof of just that.

Licensing Compaq DOS back to Microsoft was a mistake. It gave Microsoft’s OS incumbency. The PC industry has been suffering from that decision ever since.