[ih] First file transfer on ARPANET

[John Day]

I thought cloud was just the ultimate MIMD machine and the move back 
to the centralized datacenter model.  What happens when processors 
get very cheap.

At 16:31 -0500 2012/12/21, Craig Partridge wrote:
>  > John Day <jeanjour at comcast.net> wrote:
>>  > Our biggest problem was that our vision quickly outstripped the
>>  > capability of the computers of the day.  (NLS was a very good example
>>  > of working beyond the ragged edge)  Our model was that the Net should
>>  > be a virtual OS that we could just use.  Too bad those ideas were
>>  > abandoned and we haven't moved very far in the last 35 years.
>>  >
>>  > It really is too bad that the Internet ended up being more like DOS
>>  > than a real OS.
>>
>>  My impression in the few years after I got on the net at the beginning
>>  of 1990 was that distributed computing using the net as you describe
>>  was out there and slowly progressing, and that stuff started dying or
>>  getting blocked in the mid to late 90s due to security problems as the
>>  net got bigger and more public and everyone started reflexively
>>  turning off their RPC services and similar things.  How does that
>>  correlate to the impressions of people who'd seen the 80s on the net?
>
>My impression (I started in 1983) is that distributed computing was
>an idea that for a brief period worked -- the time required for long
>distance communication roughly was in sync with processor times, from
>about the late 1970s until the early 1990s.  That's when we saw distributed
>file systems flourish and experiments with distributed shared memory.
>
>Then the speed of light constrained us.  The cloud (namely clustering a
>bunch of machines very close and limiting our exchange to the start
>and end of the computation) is an interesting approach to circumventing that
>problem.
>
>Thanks!
>
>Craig