Backing up one more step...

[Bryan Ford]

>Bryan:
>
>What you refer to as persistence districts is also known as virtual
>time, and there has been a lot of work on this.  The difficulty is as
>follows:
>[...]

I think you're completely misunderstanding what I mean by
"persistence district".  It's not meant to be any kind of
automatic mechanism to support or improve distributed checkpointing;
it's merely a way to reduce granularity by allowing multiple independent
distributed persistent "subsystems" to be supported on the same
node at one time.  Persistence districts are permanent, essentially
fixed entities - a process that starts out in district A remains forever
in district A over its entire lifetime, even though it may migrate from
node to node arbitrarily.  Creating a persistence district is a
manual, heavyweight, "once-in-a-lifetime" operation,
corresponding to "formatting" a new KeyKOS system.
A persistence district permanently "owns" a fixed (although possibly
growable) set of backing store on disk or somewhere.  For example,
all the data, processes, keys, etc. belonging to a particular user might
be isolated in one persistence district.

Does this make it any clearer?  If so, I suggest rereading my original
message in light of that; I'd like to hear your comments.

				Bryan