Power PC with a linux distribution and R

To Stephen's credit, Apple will no longer support PowerPC chips (such as his
G4) in the next operating system (Snow Leopard), so some sort of switchover
will be necessary in order to maintain SOME sort of state-of-the-art
software packaging for PPC-based Mac users in the near future.

Also, it is important to note that Leopard on a G4 iBook would probably run
disgustingly (read: unusably) slow unless the memory is upgraded: It shipped
with 256MB of RAM (and that money would be better saved for a new computer),
and Leopard (on my Macbook Pro) is currently using about 700MB for the
operating system. Switching to linux and X-windows allows for an old system
to be, in a word, functional.

While I love MacOSX and use it on the computers of mine which can run it
usefully (G5 tower, PowerBook G4 w/2GB of ram), I've been much happier with
Linux on my older macs. The machine on my desk is a G4 iMac with 256MB of
RAM which has no trouble fitting mixed models in R while I browse the web on
Firefox 3 while the computer itself serves 80% of the Psychology
Department's web surveys. Under Leopard, these programs would not even run
simultaneously, let alone usably.

Further, while some amount of hand-compiling is necessary (Debian Linux
provides almost all of the software I would need), it's also quite
helpful--using Simon Urbanek's R optimization flags for G5 and G4 (see
https://stat.ethz.ch/pipermail/r-sig-mac/2005-February/001641.html ) and my
own version of ATLAS, the performance of R on that machine is comparable to
the unoptimized internal-BLAS no-effort-necessary .pkg from CRAN running on a 
G4 with 2GB of RAM and 1.5 times the processor speed under
Leopard.

Sure, it took a whole night to compile/optimize ATLAS and most of the next
day to compile R with those flags, but to a grad student like myself, that's
vastly superior to waiting twice as long for my analyses to run...or to the
impossibly high cost of a new computer.

--Adam