[FoRK] Our dispersed, not distributed, computing future
jbone at place.org
Sat Nov 7 19:12:26 PST 2009
> With respect to "connections", if the software is latency-aware, do
> you really mean "failure modes"? ... I suppose even in a dispersed
> environment there will still be potential for fatal failures of the
A "dispersed" environment is one in which the very idea of a
"connection" isn't fundamental or even necessarily sensible.
Consider attempting to establish a TCP connection to a server orbiting
in the outer reaches of the Kuiper belt...
Another interesting use case is the one where the transit time for
some message exceeds the lifetime or uptime of the originating node.
Finally, consider the issue of time protocols and causality over
networks of a scale where relativity starts to be a consideration.
(Nb., such networks aren't very big; consider GPS.)
More information about the FoRK