[e-lang] Content Complete, Ready for Comments
markm at cs.jhu.edu
Mon Jun 27 22:21:33 EDT 2005
David Hopwood wrote:
> I think section 7.2 needs some (brief) discussion explaining causal
> ordering and motivating the need for it. [I prefer "causal ordering" to
> "partial ordering" for this property, since any message ordering is
> necessarily a partial order.] I'll post more about what I understand
> the motivations to be tommorrow.
Actually, E doesn't implement causal ordering. E's message-delivery order
guarantee, which I now generally refer to as "reference-graph order" is
stronger than FIFO and weaker than causal. It can be thought of as FIFO
generalized to dynamic graph topologies, or something.
My original fantasy was to explain the need for reference-graph order in this
document, to explain how Pluribus enforces it among mutually suspicious vats,
and to speculate that causal and stronger orders cannot be enforced among
mutually suspicious machines.
Needless to say, we weren't able to get that far in this document. Instead, in
this document, we explain only the need for FIFO and how we achieve it.
Text by me above is hereby placed in the public domain
More information about the e-lang