Thu Aug 25 10:54:56 EDT 2016
Another round of "while obviously this is fun to work with, does it
have any real practical value?".
And the answer is mainly: simplicity.
What does that really mean in the face of designing distributed
After building a real-life application structured as linux/erlang +
custom ARM nodes, it is clear that the real problem is how to evolve a
design, and how to properly test it.
Evolving: make it easy to upgrade code, possibly without taking down
the whole thing.
Test: it seems almost impossible to avoid an object-oriented
distributed state style because of physical constraints. All that
functional programming mumbo-jumbo is nice on a small scale, but
regarding a full network it's really not applicable.
To try this out, I started building a distributed Erlang system. Next
is to add some sensor/actuator nodes to that using Staapl or some
other low-level state machine OS.