Most of what is written is obviously true, but I am sure that I am not alone to run into memory limitations when stitching large images, or post-processing scanned 4x5... which I do often.
Or if you're doing a manual batch-edit of 40-100 photos.
The theoretical advantage of the OS caching the scratch space is a cop-out. The scratch space will still have to be written to disk, which can slow other computer operations down. Sure, read access may be quicker, but it's not all about read access. And the OS that really provides the theoretical advantage that he touts, is one that isn't supported by Photoshop.
Photoshop users have found their way around this by e.g. creating RAM disks, but unfortunately, this leaves less room for the OS to be efficient in handling memory. (Which may be both good and bad, depending on how you look at it.)
He also overstates the memory bandwidth problem. Yes, it is a problem that memory bandwidth isn't increasing at the same rate as processor computing speed,
but memory bandwidth for standard PC processors has increased by a factor of three since CS2 was released, in addition to gaining significantly better inter-processor memory performance in dual-core systems than in old dual-CPU systems.
That being said, most of his points are perfectly valid, but perhaps not quite in the way he presents them.
I feel that Mr Byer is probably under-estimating the amount of people that would benefit from a lifting of the current memory limitations.
I don't think he's underestimating the amount of people who would benefit today, but he's underestimating the amount of people who will during the lifetime of CS3.
I understand his views regarding compatibility with current users' systems, and the capabilities of
current operating system platforms that Photoshop runs on. Someone is bound to jump in and say something about "what if Adobe had developed PS for Linux instead? Linux is 64-bit ...", but that's not really helpful, as long as the major user base is stuck with a 64-bit OS in betatesting with Microsoft's customers (XP 64-bit) or an incomplete implementation in Mac OS X 10.4.
I'm just a bit disappointed that he ignores the point that CS3 will be the standard Photoshop for the next two or three years, and will have the same limitations even when people are running "real" 64-bit Windows Vista and Mac OS X 10.5 (or 10.6).
What I think would be needed from Adobe, is the willingness to release an update with support for higher memory usage when "the time is right", rather than following this two to three year release cycle. I think it's almost certain that such a release would have merit in early 2008, and if not, at least later that year.