Torvalds' view of desktops seems equally casual. While curious enough to try different ones, his main criteria when choosing seems to be the ability to focus on what matters to him -- and, evidently, simple curiosity about what different projects are doing.
Torvalds' outspokenness and celebrity always make him good copy, but, Segio is also right to point out the obvious: "Linus is precisely one user. For every Linux Torvalds (there's exactly one of them), we have 10s of millions of other KDE users."
The danger, as Seigo suggests, is that undue weight will be given to that one person, precisely because of who that person is. "I don't like losing any user," Seigo says, "and such a happening can be deflating and make one second guess what they are doing." That sort of reflection "isn't an entirely bad thing," Seigo acknowledges, but he points out that the pressure of reacting to such celebrity criticism can "result in bad decision making or paralysis."
Seigo's particular concern is that undue attention to Torvalds' opinion -- or, to be precise, the media-distortion masquerading as Torvalds opinion -- will be a "fear of innovation. 'Don't do anything too big, because it'll cost you and cost you ...' is the lesson some are taking away from all this."
The reaction to Torvalds' off-hand remarks is evidence that the associational fallacy -- the mistaken belief something is good or bad because of the person who endorses it (or seems to) -- is alive and ranting in the free and open source software community. Although Torvald's hard work and expert knowledge of operating systems and Intel architecture deserve respect, he is not a usability expert.
Nor do his brief comments, even when taken in context, suggest that he has thought deeply about his reactions. His usability comments deserve as much consideration by KDE developers as those of any other user -- but no more.
Torvalds' knowledge of usability seems limited to his personal experience. However, after managing a large project like the Linux kernel for eighteen years, Torvalds does know a thing or two when he talks about release management.
Consequently, when he suggests that the release of KDE 4 could have gone better, his words should carry some weight. Exactly what went wrong with the KDE 4 release is a complex question, but the points that Torvalds alludes to in passing -- a release with fewer features than the previous series, the lack of backward compatibility, the rush by distributions to include a release that was not yet ready for the general user -- all seem to have been part of it. These causes all suggest that while the engineering behind KDE 4 has proved to be sound enough, the managerial decisions -- such as what to release as a 4.0 release, and when to enable or add features -- were not made nearly so well.