Wed Mar 23 13:37:10 EDT 2016

oo vs. actor

I'm (still?) using a lot of oo techniques to design erlang programs.
E.g. to implement server processes as objects that can be subclassed
by delegating to "base class" handlers.

Not sure if this is such a good idea.  There is always the tension
between storing information in the same place (e.g. a dictionary with
separate tags for separate pieces) and going through the trouble of
separating storage into separate processes that only have a message

The latter involves more protocol, the former gets messy at some
point.  There is a tradeoff..