Tue Sep 20 17:22:06 EDT 2011

Protocol-oriented programming (part 2)

Started here[1].  Recently been thinking about this again because I
ran into some code that is gratuitiously un-streamable, i.e. a byte
stream with non-causal data dependencies across large parts (packets)
that basically requires the use of large buffers.

Is there a way to turn the question around?  How to design a data
protocol such that small buffers can be used?

     In absence of other constraints (i.e. robustness), a good
     protocol is one that is easy to print / parse.

How to formulate this in terms of languages and automata?

[1] entry://20110423-004220