Joe Wilcox over at Microsoft Watch today wrote a piece entitled "Operating Systems Don't Matter".
As I wrote in the comments to that post, I was a partcipant in a concrete example of his thesis: Digital Research, Inc.
DRI wrote and distributed operating systems starting in 1977, beginning with CP/M and then MP/M, CP/M-86, MP/M-86, CP/M-68k and Concurrent CP/M-86. I was the director of Research and Development at DRI from 1981-1984.
Each of these were state-of-the-art for the time and contained more functionality than the competition (PC-DOS/MS-DOS). However, as Joe points out, end-users don't buy operating systems, they buy tools that help them have fun, do their job or save time (at best all three!).
Digital Research and Microsoft were of approximately equal size (~200 employees) in 1983 with Digital Research leading the way with a broad range of "ISVs" or independent software vendors. We even held our own trade show CP/M-83 in January 1983.
Gary Kildall who founded DRI had a strong philosophical stance that would, eventually, be the downfall of DRI: "We write operating systems and languages but we do not write applications. We don't want to take market share from our ISVs".
Microsoft, on the other hand, started writing applications shortly after PC-DOS was shipped and that made all the difference. The pull through for MS-DOS, and eventually Windows, was driven, of course, by applications and Microsoft wasn't afraid to participate in the application side of the market.
That, in my humble opinion, made all the difference. Digital Research never went public and by 1991 had been bought by Novell and eventually faded into computing history. Microsoft went public, kept buidling applications and, to my continuing bemusement, I now work there (not building applications mind you but creating tools for those that do).