[radiowave]

I was thinking about this just recently, but instead of likening these kind of things to sci-fi, I was thinking more along the lines of non-Euclidian geometry, i.e. what if we take something that is considered to be axiomatic, and change it - a whole different world emerges.

For example, throw out the notion that memory is volatile - or slightly more practically, what if the price we pay for automatic memory management in our programming languages also bought us abstraction over the volatility of memory? How different would our systems look? For one thing, switching things off and back on again wouldn't be the "cure-all" that it mostly is today.

The fact that we can build systems like Smalltalk tells us that much of our current notions of computing are merely convention, not axiomatic at all.

Smalltalk and Forth are definitely "different convention" things, while SICP and CTM are more like detailed examinations of things that might really be axiomatic, giving us the means of combination, and hopefully the means to imagine building things beyond what our mindset of present conventions would allow.

[groby_b]

Quite a bit of memory used to be non-volatile.[1] These days, there's FeRAM[2], but it's not widely used. Power-off not working wasn't really an issue - you'd just manually key in a bootloader of ~30 machine words that you knew by heart :)

[1] http://en.wikipedia.org/wiki/Magnetic-core_memory [2] http://en.wikipedia.org/wiki/Ferroelectric_RAM

[abecedarius]

http://en.wikipedia.org/wiki/KeyKOS was an OS along those lines (but on conventional hardware, treating RAM as cache). It was very different also in being built around capabilities -- a related decision.