Re: memory-mapped I/O (was: Re: Removing ARCNET stuffs)

[Mouse <mouse%Rodents-Montreal.ORG@localhost>]

> Yeah, well a proper from-the-ground-up universally virtual-memory
> operating system has some major advantages over Unix with mmap,

True, but, address space limits prevented this from being done on most
hardware until recently.  For this to be practical, you need enough
address space to contain everything you're working on.  This was
practical in Multics's day (36 address bits and word-addressed, wasn't
it?, and 64 gigawords was "infinite" in that day, even when measuring
file sizes and the like) and is now practical with 64-bit machines, but
in the 16-bit days it was pretty much out of the question.  There was a
brief period in the early 32-bit days when it was realistic, but it
didn't last all that long.  (I have to wonder whether/when we will
reach the point of it being impractical for 64-bit address spaces....)

> When I was a Multics user I struggled constantly with understanding
> what it meant to access all storage as virtual memory [...]

Heh.

Like jkl, my first reaction is to wonder how to append to a file.
Reading your later explanation, my next reaction is to wonder how to
figure out what exists in a file. :-)

I should start experimenting.

/~\ The ASCII				  Mouse
\ / Ribbon Campaign
 X  Against HTML		mouse%rodents-montreal.org@localhost
/ \ Email!	     7D C8 61 52 5D E7 2D 39  4E F1 31 3E E8 B3 27 4B