Wed Nov 22 10:47:03 EST 2017
Eventual consistency / single writer
Might be useful to understand the principle, so computer admin can be
done in this way.
My intuition so far for the problem I'm facing:
There is a central list of edits. Each node needs to play back the
edits when it comes back online after "missing time" somehow -- can be
benign: laptop powering down, or just no internet connection.
This is simple because there is only a single writer.
Implement it first for /etc/net and ~tom/pool
It could be implemented as a git or darcs repo, or even a nix channel.
Something that will fetch current head from master and instantiate it.
Make it push/pull:
- sync after lost time
- subscribe to notifications
A downside to this is that bad edits that bring down a node will
spread, so maybe keep it manual for now?