30 years on, how Bill Gates saw MS-DOS
It will soon be the 30th anniversary of the launch of the world-changing IBM PC in August 1981, and technically it's also the 30th anniversary of MS-DOS, which was about to appear as IBM's PC DOS. This dominated the operating system business until Microsoft launched Windows 95. At that point, Windows changed from being a graphical interface that sat on top of DOS into an operating system that used DOS as a loader (though, of course, DOS was still integral to the Windows 95/98/Me line until it was replaced by the Windows NT/XP/Vista line).
David Bunnell launched PC Magazine to cover the brave new PC world, and the first issue included an interview with Microsoft co-founder Bill Gates, which PCMag.com has now republished on the web: MS-DOS Turns 30: PCMag's Original Interview With Bill Gates.
At the time IBM's Jack Sams visited Microsoft in Seattle in 1980, its staff and total turnover was smaller than an IBM rounding error, with IBM having roughly 10,000 employees for each Softie. The disconnect is visible when, having started late on its PC, IBM tries to speed up the process. Gates says:
And then they said, "We have a lot of things to do, we'll have our technical team meet with your technical team so let's do them in parallel. We'll have our legal team meet with your legal team, we'll have the purchasing team meet with your purchasing team, we'll have our technical team meet with your technical team, so we can do four or five things at once." Well, that is fine, but that's me who is going to do those things and I can do only maybe two things at once, so we're not going to be able to have five simultaneous meetings.
One of the things Gates tried to do was push IBM towards using a 16-bit processor, to provide a whole megabyte of address space, rather than using an 8-bit processor with only 64K. Gates argues that Microsoft will be able to focus on improving the user interface and software integration rather than cramming as much as possible into too small a space. He expands on this point later:
In five years the cost of computation will really be effectively decreased. We'll be able to put on somebody's desk, for an incredibly low cost, a processor with far more capability than you could ever take advantage of. Hardware in effect will become a lot less interesting. The total job will be in the software, and we'll be able to write big fat programs. We can let them run somewhat inefficiently because there will be so much horsepower that just sits there. The real focus won't be who can cram it down it, or who can do it in the machine language. It will be who can define the right end-user interface and properly integrate the main packages. I expect over the next five years between us and others a heck of a job will get done. You'll be able to sit at your desk and do whatever it is you want to do with information or presenting data or interchanging data incredibly effectively. In other words, we will have changed the way people work.
The IBM PC had a floppy drive and a cassette port -- no hard drive -- but Brunnell asks him: "What kind of mass storage device will machines have in five years?" Gates replies:
Well, you'll probably still have local floppies in a lot of cases, but most of the storage size-wise will be in shared file servers — and although optical disk may have had an impact, even at present prices and capacities large (magnetic) disks would suffice. There are 300-megabyte disks down in the $10,000 to $15,000 range now. If you can spread it across 20 users — that is, with a good networking scheme — you could justify it. So, while there ought to be some improvement there, I don't think that we've got any bottleneck even today. Networking is probably one of the big challenges.
I'm reminded that the first PC hard drive in the company I worked for way back then was only 5MB, and it was shared by four users. Happily, I soon had an IBM PC/XT with a whole 10MB hard drive all to myself. Reed even shelled out £999 for an extra 512K of memory so we could run Microsoft Xenix (Unix) as well.
Finally, with hindsight we know that the IBM PC established a platform that has sold well over a billion PCs. However, when Bunnell asks Gates "How many machines do you think IBM will sell in 1981?", he says: "My guess is not based upon any inside information whatsoever, but I think it will be not far from 200,000."
If they can deliver them, the potential is there. I've heard numbers ranging anywhere from 100,000 to 150,000, so I am an optimist beyond the median point of that scale. They'll have to open up more distribution, though. I don't think Computerland can push through that many. And they may run into some production bottlenecks. There are a lot of outside vended parts on the machine and they are not going to compromise quality. Certainly at this point the machine is incredibly short, you know, we've got a ton on order and it is going to take a few months before they come in.
Some things don't change, though the numbers get bigger every decade.