Well that’s funny, because ftp://ifctfvax.harhan.org/pub/micro/msdos/above640k/TrueDPMI/README says that the document you referenced is not the “True DPMI” spec. And that he (Michael Sokolov) never found it.

The “MS-DOS Extensions for DPMI Hosts” document only confirms what I said. It is not a complete 32-bit DOS environment; one of the biggest missing pieces is the ability to load a 32-bit protected-mode application. In other words, a DOS extender is still needed, even if it does not have to do the job of actually extending the DOS INT 21h API. And because this capability never became part of DOS itself, it’s completely moot — application writers still needed DOS extenders. Unless they thought requiring Windows to run a DOS application was smart. It should also be obvious that however evil and shortsighted the DOS extender vendors were, building “True DPMI” support into DOS was not in their power.

Yes, Microsoft didn’t want people to have to get out of Windows, and that’s why they came up with the idea of DPMI. But Microsoft also didn’t really want people to run or write 32-bit DOS applications (or DOS applications at all). Which is why there was no 32-bit application support built into DOS. It would have been easy for Microsoft, but there wasn’t even an optional DPMI server like QEMM and 386MAX (and Novell DOS, I think) had. Which is why Microsoft C/C++ 7.0 had to be shipped with 386MAX just so that DOS users could run it.

Microsoft only needed a basic interoperability interface, sort of VCPI 2.0. DPMI as it was standardized is actually quite powerful, but not necessarily terribly easy to use. That was part of why DOS extender vendors stayed in business (for a while) — DPMI was not a complete software environment, and I don’t think it was ever meant to be.

It’s possible that for a while, Ralph Lipe was thinking about a real 32-bit DOS which would directly load and run 32-bit applications, although there’s no solid evidence I’m aware of. It’s pretty clear that Microsoft did not actually want that, because they had a lot more control over Windows than they had over DOS.

Well, according to that article they _didn’t_ let the committe dictate anything.. they just let them go through their motions of defining their standard, and then Microsoft went ahead and implemented their own version (Ralph Lipe’s version) anyway. Business as usual then, the MS way! 🙂

I wonder if the author found that True Spec.. I’ve not followed the dos extender market since way back before Windows (I never really moved to Windows, any version).

Although it still had some life in it, there was no doubt that the dos extender market was kind of cooked if Windows could run stuff all by itself so I could see DPMI being a crippled base to keep the extender people in business.

If Windows could run EXP’s then why would anyone buy Pharlap 386|Dos Extender / TNT.

And the bit about DOS extenders being bad because they’re built into every app is total nonsense. It completely ignores the history–early DOS extenders had to have everything built in because they simply had no other choice. The companies which could create an extended DOS standard (IBM and Microsoft) were very much not interested in doing so, because it went against their strategic goals.