asciilifeform: and build the actual state in fpga out of same.
asciilifeform: if you have the magnetic ram (actually available commercially, no need to wait for hp) you can dispense with the disk.
asciilifeform: only the circuit that marshalls state from the independent modules (e.g. keyboard buffer) to magnetic storage needs to know that it's a disk, and comes in 0...N 'LBA' blocks, etc.
asciilifeform: but if you have the option of doing the last 35 yrs. of hardware again - you can toss it.
asciilifeform: the brain-damage of the 'address space' concept is inescapable on a von neumann arch
asciilifeform: just as you presently cannot dereference a pointer to your disk's sram cache
asciilifeform: that is, there is no reason why, say, garbage collector circuit (yes) needs to be able to request a byte from my keyboard buffer.
asciilifeform: decimation: you don't need an address space as such if anything that might conceivably need to communicate with, e.g., keyboard, has an actual (matrix switched) electrical connection with it.
asciilifeform: decimation: think 'emulation of room-sized fpga' and you're 90 percent there.
asciilifeform: orthogonal persistence is something quite else.
asciilifeform: justusranvier: this is what's on your desk now, essentially.
asciilifeform: (when was the last time you pulled your disk's cache sram out?)
asciilifeform: it's not orthogonally persistent if this actually does something useful
asciilifeform: (no distinction between disk and memory)
asciilifeform: decimation: you must really think about this to understand the implications. a single-address space machine must... actually work. every time - because it can in no sense be rebooted.
