Suppose I'm right: the mass media have largely abandoned their enthusiasm for Linux as counter culture and withdrawn the push that drove adoption in the 1997-2002 period. So what? by the time their support waned, there were millions of highly successful installs, hundreds of thousands of competent Linux people, and thousands of high quality applications. With this base in place, why didn't growth continue to accellerate?
True, some minor upgrades by Microsoft were being heralded as revolutionary by their friends in the computer press, but these didn't actually amount to much and the rest of the computer press continued to trumpet the joys of Linux uber alles while Apple was the only major Unix player to get its competitive response to Linux, the X-serve/X-RAID combination with OS X licensing, into the field.
In other words, Linux growth didn't slow because of competition - something else must have caused it and we need to understand what that was before we can work up a plan to do something about it.
What I think happened is that when the external presures driving growth virtually disappeared, the internal impediments to growth became more important. Specifically:
Something similar appears to have happened in Europe where national government endorsement of Linux as the anti-Bush candidate for their schools and lower level governments triggered a slowdown in acceptance among the more clear headed of the academics and researchers driving continental R&D - something that seems likely to cause the bottom to drop out of Linux acceptance there in a few years.
Basically, legal issues, or the threat of legal issues, caused some key applications developers to back off Linux while the general negativism of Linux marketing caused many of the individuals whose innovations should have been driving Linux adoption to hang fire until MacOS X and Solaris for x86 under the CDDL came along.
Notice that in saying this I'm opting for chickens over eggs: saying that some application suites aren't considered competitive because key developers, including Microsoft, stayed out of the Linux market. In effect, I'm saying that Joe average may not want to use Linux today because other people tell him it's too hard, but since the reality is that the Linux OS isn't any harder to use than NT 5.1, the difference they're concerned about have to be in the applications and the device drivers many of them depend on.
At a deeper level what I think is going on is that the transition from growth driven by the political views of the technically uninformed to growth driven by the intrinsic value of the product has turned off people who might otherwise be creating the kind of innovations that lead to major market success. If so, there's a positive corolary: if Linux can survive both its 15 minutes and IBM, growth should pick up again.