In reply to AMS:
Julian wasn´t intending to include Windows at all, or even a Linux GUI on the the kernel, so his idea effectively chops the heavy top off the underlying OS and replaces it with the lightweight GUI from RISC OS. OK a thirteen times speed increase is a little optimistic, but how much would you like, five times?
As far as I can see (and I just checked back) you are the only person talking about write caching; if you don´t like it, use a filesystem that doesn't do it. It´s only really a problem if your OS is likely to crash when a program contains an error and Linux is orders of magnitude less likely to do that than RISC OS.
Any alternative system for running RISC OS (or RISC OS-like) programs would be capable of moving back to any ARM processors that become available in the future (I don't think anyone is recommending swapping a reliance on ARM for a reliance on x86). Does the reliance of RISC OS on ARM processors make it more likely that future processors will be more suitable for desktop use? I very much doubt it!
RISC OS needs new users. To get them it needs faster hardware and new applications. Either we can wait around, losing users, until someone produces a faster ARM processor suitable for desktop computers (which no company has said they are going to do), or we can do something to make the best features of the OS available to many more potential customers. The alternatives being suggested here are ways of doing that before there's really only a couple of us still around, and provide at least the chance of more sales of RISC OS applications.
Personally, I believe my suggestion has the potential to be better in most respects than the original, and involve less work than, say, porting Firefox to RISC OS has. I'd welcome arguments to the contrary.