> -----Original Message-----
> From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED]]On
> Behalf Of Dan Minette
[snip]
> Hmm, I'd guess I would be curious to see if Linux took market share away
> from Microsoft or from other Unix. I'd argue that the shares of the
> different flavors of Unix don't matter as much as the summed Unix shares.
The Intel-architecture *nix market was virtually nil before Linux came
along, so it is almost all market share that would have gone to Microsoft,
assuming that these customers were going to IA anyway.
The workstation market only began to change course after Intel shipped the
Itanium, and even that hasn't had much effect yet. But there are already
market researchers asking if workstations will exist as a category in five
years. When it is practical to apply PC architecture to workstation
applications, disruptive cost changes result.
As PC clustering becomes more science than art, it's going to be very hard
to sustain workstation architectures. For example, look at the performance
of NVIDIA's graphics processors v. high-end graphics workstations; a
GeForce4 outperforms an Onyx costing hundreds of thousands of dollars, in
many ways. And even when a company like SGI uses the NVIDIA chips, the
architecture around them drives up the cost dramatically. The manufacturing
volume for PCs is the killer. And guess what operating system is being used
for virtually all major clustering work -- Linux.
Nick
