No, its true. You might think, well sure, Microsoft is proof of that buggy, poorly designed software pays, but actually, they pay the price for quality software as often as anyone else. If a company produces a product that works reliably, does everything that the customer needs, and doesn't have any annoying little bugs, what is the incentive for the customer to continue to purchase newer versions? If they don't purchase newer versions, the software developer's revenue dries up and they either go out of business or move on to greener pastures, often leaving the customer with a dead product. Everybody loses.

There are lots of examples of this phenomenon:

• Where would all those Linux companies be if any monkey could install Linux and set up an Apache server? They'd be looking for something else to do, that's what. Instead you have a mishmash of distributions, difficult to configure software, and enough fixes, patches and upgrades to take up a full-time position just keeping up to date. On the plus side, at least once you get it running it is reliable.
• OS/2 hit the quality sweet spot in 1994 and IBM's revenues from it have suffered ever since. OS/2 1.3 was a great product, but it lacked DOS support. 2.0 was a great leap over Windows 3.1 and had good support for DOS, but it crashed pretty regularly. 2.1 was better, as was 2.11, but they both still crashed pretty often. Then came Warp 3. Reliable, pretty, fast. It had everything going for it and it sold really well. And the damage was done. Warp 4 came along, initially it was buggier, upgrade rates dropped, and as a result of this and other influences, IBM got less interested. Banks, large companies and lots of individuals had settled on Warp 3, were happy with it, and run it to this day.
• Watcom C++ (and most of the other C++ compiler vendors as well - Borland, Symantec, IBM to name a few) ran into the same problem. Watcom C++ 10.6 was a wonderful product. Reliable, produced good code, got good reviews, sold well. 11.0 came out and nobody really cared. It had more bugs and didn't offer enough compelling features to get people to give up the compiler they knew. 11.0 was out for over a year before I upgraded, solely because I needed support for 64 bit integers for a Windows app I was working on. There was no similar incentive for OS/2 or Netware developers or for that matter, most Windows developers. Just this week I heard someone recommend 10.6 over 11 and 10.6 hasn't been for sale for years. This has got to be a pretty big factor in the current malaise in the C++ compiler market.

When you start talking about crappy software, people immediately think of Microsoft, its continuous upgrade cycle, and its fortunes, yet Microsoft has made two potentially fatal missteps itself: Windows 2000 and Office 95. An article I read this week stated that the upgrade rate for Office 2000 was just 20%. With Windows 3.1 now down in the single digits for marketshare, that means that the bulk of Office users are still running Office 95 or 97 and Microsoft hasn't seen a dime from them in as many as 6 years. Windows 2000 is another such product. It works well, can run reliably for weeks on end, it runs all the software you might want to run on any moderately modern hardware and it doesn't have any built-in limits to its lifespan like Windows XP. Some people might argue that the same was true of Windows NT 4, however it was always too fat and slow and remains so today, even on modern hardware.

These two products represent the large bulk of Microsoft's revenue stream. According to the popular press, even with some heavy-handed pricing incentives, companies don't seem to be interested in upgrading to Windows XP or Office XP. Microsoft of course has seen its revenue declining the last while and is doing something about it. And I'm not talking about their license compliance effort either, this is just an indication of their level of desperation rather than an effective strategy for improving the company's bottom line and continuing their growth.

The real solution to their problems is their new subscription scheme for Office and likely soon for Windows itself. If they can get users to accept a moderate upgrade fee every year (they've based their pricing on a 3 year upgrade cycle) then it is pretty likely they can move that 20% upgrade rate up to 100%. The beauty of this is that with this new scheme they could lose 80% of their customers before they are worse off than they are now. After all, they're only interested in paying customers, and having your software quit when its subscription is up is a pretty healthy incentive to purchase that upgrade. If they can get any part of that other 80% to upgrade, that's gravy. Plus they'll start to see a huge new revenue stream from all those preloads they have. How many people do you know that had Windows and Office 95 preinstalled on their computer and run them to this day? Microsoft would sure like to see that phenomenon come to an end.

The one exception to this quality vs profit rule is content companies. When you are making content, upgrades are generally a liability to be avoided, quality assurance is king, and as soon as the title is out the door, you're on to something new. Microsoft seems to have realized this a few years ago when it started moving heavily into entertainment production and games which have a limited shelf-life and little upgradability. Electronic Arts is certainly doing well in this space.

Looking at this from the point of view of an OS/2 user looking for good software solutions, this makes for pretty dim prospects for new OS/2 applications. Unfortunately, most OS/2 ISV's are small and don't have the resources like Microsoft to work on alternate revenue streams. To be sure, producing content is a tough business in a small market like ours. As OS/2 users, and often naive startup software companies, OS/2 ISV's for the most part take pride in their products and design them to be reliable and as feature complete as possible, not taking into consideration that they might want to have an income down the road. So the next time you hear about a new version of an OS/2 product you use, think twice before you go and spend that money on the latest Windows game instead. After all, who do you think is more likely to develop more OS/2 software for you, Microsoft and EA, or your favourite OS/2 ISV?