October 18, 2018

The problem child (4/7)

It there was a single piece of technology that marked where things started to go wrong between IBM and Microsoft, the final straw that convinced Bill Gates, "It's not you, it's—no, it is you," it was DOS 4.0. DOS 4.0 was released by IBM in 1988. Microsoft walked away from the relationship two years later.

DOS 4.0 was at first greeted with great acclaim. Paul Somerson stated in the 27 September 1988 PC Magazine, "DOS 4.0 answers just about every major complaint about prior versions." But the glow faded fast. DOS 4.0 was soon causing more problems than it solved.

Asked John Dvorak in his 15 November 1988 Inside Track column,

Can't IBM do anything right? [DOS 4] is their baby, and it has so many bugs that we're told that we can expect to see 4.1 sooner than expected. I'd wait for 4.3 the way they are going.

In the 17 January 1989 issue, Ray Duncan rose to the defense of the OS, arguing that it was a victim of inflated expectations.

When IBM's DOS 4 first appeared, analysts and pundits hailed it as a major evolutionary step. A few weeks later, the same analysts and pundits came to their senses and there was a severe backlash. DOS 4 was subjected to a torrent of abuse for a handful of bugs no worse than those that accompanied the release of DOS 2 or 3.

But a little over a year later, Duncan took another opportunity to analyze the role of DOS 4.0 in what he saw now as the systematic undoing of IBM. In his 16 October 1990 Power Programming column, he succinctly summarized the beginning of the IBM's declining influence in the personal computer arena.

DOS 4 will probably merit a footnote in the history books as one of personal computing's major operating-system fiascos. The changes that appeared in PC-DOS 4 were entirely implemented by IBM, leaving Microsoft in the uncomfortable position of having to reverse-engineer the system in order to come up with a "generic" MS-DOS 4 that could be licensed to other OEMs. Users stayed away from DOS 4 in droves. As I am writing this, DOS 3.3 is still outselling DOS 4 by a significant margin.

As a result, Duncan reported later in the same issue,

By 1990, Microsoft had awakened from its preoccupation with OS/2, realized that DOS was still a cash cow, wrested control of DOS's development back from Boca Raton, and deployed famous software guru Gordon Letwin to recoup the damage done to DOS's reputation by DOS 4.

Microsoft and IBM parted ways that year. Once there was no longer any need to pretend they liked each other in public, the gloves came off. PC Magazine editor-in-chief Bill Machrone reported in the 28 May 1991 issue that Microsoft's message at the System Strategy Seminar earlier that year was hard to miss:

Forget DOS. Forget OS/2. Forget LAN Manager. Forget NetWare. Forget Unix. But don't forget Windows. That's what Microsoft will ask you to do in the coming months and years.

The pieces puzzled over by the prognosticators for years now fell into place.

The next OS/2 from Microsoft will bear little resemblance to the versions we've seen to date. OS/2 3.0 will feature a "New Technology" (NT) kernel, full Windows support, portability, distributed processing, and POSIX support. The development team is headed by Dave Cutler, the guy formerly responsible for VMS at DEC.

OS/2 optimist William Zachmann heard the same message—"Forget about OS/2 2.0 and stick with Windows"—and fretted that "this Windows-centric outlook has led many to assert—incorrectly—that OS/2 is dead."

Well, those "many" were asserting correctly. As far as the PC consumer was concerned, OS/2 was dead. The death certificate was delivered a year later (though the undead OS/2 would wander the Earth for another decade).

A month later, John Dvorak observed in his 11 June 1991 Inside Track column that the feud between IBM and Microsoft "over the direction of the industry, over Microsoft's emphasis of Windows, and over DOS 4"

finally erupted in a Forbes article where the two companies clearly stated that they were in disagreement about direction. Gates said that he and Microsoft were largely responsible for the success of the IBM PC and today's hot computer market. He made it clear that IBM's contribution to the leadership was a mirage.

In his 15 October 1991 column, Dvorak pointed to the "infamous Bill Gates memo" first leaked by the San Jose Mercury News. "The memo criticizes IBM for providing Microsoft with lousy code and intimates that Microsoft is better off without IBM."

IBM had hamstrung the DOS operating system long before DOS 4. In an truly ironic juxtaposition, the same 28 April 1992 issue that previewed Windows 3.1 in a cover story (and mentioned OS/2 2.0 in a much smaller font) also ran a letter from one Tim Paterson of Redmond, Washington.

Titled "IBM was the problem, not MS-DOS," Paterson objected to the "common misconception" that the designers of MS-DOS had "divided the first megabyte of memory into 640K RAM for the operating system and application and 384K reserved for hardware," creating the long-derided "barrier" that restricted memory access in MS-DOS to 640K RAM.

And who was this Tim Paterson to make such a claim?

As the sole designer of MS-DOS, Version 1.0, I did no such thing. The first computer that ran DOS could have a full megabyte of memory of DOS and applications. The system ROM occupied only the last 2K of the address space, but even that could be switched out after boot-up.

When Microsoft purchased DOS from Seattle Computer Products, Tim Paterson came along as part of the deal.

Those of us who lived with the 64K address space of the 8080/Z80 had learned our lesson. IBM, unfortunately, did not. They alone decided to build memory-mapped hardware right into the middle of our precious address space. I am mystified why anyone would consider this poor hardware design to be an aspect of DOS. DOS uses as much contiguous memory as can be made available.

Four years later, in the documentary Triumph of the Nerds, Steve Ballmer ruefully admitted that

Windows today is probably four years behind, three years behind where it would have been had we not danced with IBM for so long. Because the amount of split energy, split works, split IQ in the company cost our end customer real innovation in our product line.

Microsoft's successor to DOS 4.0 was a case in point. First reviewed in the July 1991 issue, PC Magazine would give DOS 5.0 a Product of the Year award.

Between 1990 and 1995, Windows 3.0 was followed by DOS 5.0, Windows 3.1, DOS 6.0, Windows for Workgroups 3.11, Windows NT, and Windows 95. Microsoft was on a roll. IBM became an afterthought as a supplier of PC operating systems.

Though IBM did engineer some of the best laptops ever, before selling its entire PC division to Lenovo.

Paul Allen (1953-2018)
Co-founder of Microsoft
Requiescat in pace

Related posts

The future that wasn't
The future that wasn't (1/7)
The future that wasn't (2/7)
The future that wasn't (3/7)
The accidental standard
The grandfathers of DOS

Labels: , , , , ,