I decided that I won’t bother developing for the aging Windows 9X operating systems. My software may work on those systems since the .NET framework insulates much of the differences among the operating systems, but I don’t plan any heavy testing on those systems. I may need to do a cursory walkthrough to catch any showstopping bugs just in case anyone runs on those earlier systems but my software won’t be advertised to work with them.
Even though I killed off support for Windows ME, my software will still work with Windows 2000, which is six months older. The differences between Windows 2000 (version “5.0”) and Windows XP (version “5.1”) are mostly cosmetic, and Windows 2000, surprisingly, still has 25% more users than XP. Window ME wasn’t very popular and had only one year of air before it was finally choked off by XP.
I just looked at an AssetMatrix statistics on the market share for various Windows operating system. Windows 2000 is the most popular operating system with a 48% market share, followed by Windows XP with 38%. Windows 2000 computers—no surprise—dominate in businesses, especially large businesses, although Windows XP does have a majority in small businesses with less than 250 employees. These two operating systems have a combined 86% market share, which is only getting larger.
Windows 95 and 98 are both under 5% of the user base, which is still millions of people, but these people are technology “laggards,” who are unlikely to buy additional software anyway. Their systems probably wouldn’t have enough memory, at least 64 megabytes, to run my software anyway. That report leaves remaining 9% broken down between Windows NT, Windows ME, and Windows 2003 Server.
(These facts make me wonder: By initially planning to make all new technologies Longhorn-only, Microsoft may have created the perception in businesses that Windows XP is a deadend incremental OS for consumers and slowed its adoption even as Longhorn continuously slipped further away. Assuming many companies don't buy subscription packages from Microsoft, there's probably money to be made just from converting existing Windows 2000 computers to XP. Recent moves to backport the Longhorn feature set to XP may be an attempt to hasten this conversion especially in light of security holes. Basing Longhorn more on the XP code also reduces dependencies and development uncertaintities that could impact the ship date while providing a stable base for building newer APIs like Avalon and so on.)