recently peter alexander summed up all infos about upcoming GS on the NS-site. the main point, at least for me, was: it will not go above 1.5GB, but the buffersize will decrease to 64 kB/stereo. this seems to be confirmed more or less by nemesys-developers.
no groundbreaking, but even good news - this would give you the option to load 3 times the amount of samples as with 2.5 - not so bad
i don't want to reiterate myself too often and everybody who guesses i'm wrong (and i would love to be) is invited to do his own research: there is no way to go beyond the 2GB-barrier for process-threads on 32-bit operating systems, be it win, mac, whatever.
secondly, systems which can use more than 4GB are in the high-end-segment and priceless (usually in this case the OS comes with a special built and configured server-hardware)
so _if_ nemesys tries and succeeds to build a version (after 3.0) which can be run in two independent instances this will give us 6 times the amount of ourdays, assume the buffer-size can be reduced again to 32 kB/stereo it will be 12 times.
besides then you will have to be very carefull selecting your hardware (has to be a bunch of disks or a high-performance-raid) and reached *the end of the flagpole* for 32bit-systems
but assuming moore's law *) is also valid for VSL, you will _need_ such improvements [;)]
christian
*) Gordon Moore, co-founder of Intel, postulated 1997 data density (on chips, and therefore processor-speed) will double approximately every 18 months, this is expected to hold for at least another two decades
EDIT: i just realized, i took part on hijacking this thread - sorry donimon - next time i'll open a new one