Hacker News new | ask | show | jobs
by moshezadka 5730 days ago
I would disagree with the central thesis. You can sacrifice P for a weaker version: assume a network which "eventually heals": any live node will answer at least one message in a hundred, say [any node which does not is assumed to be unavailable]. The alternative to P is not "perfect network", it's "bounded from below on the badness thereof network", a significantly more realistic beast.