investigate integrating with multiprocessing and mitogen as alternatives to execnet

[RonnyPfannschmidt]

from my pov there is multiple reasons

a) most people are only interested in -n ../auto no need to inflict execnet if its not needed
b) @flub recently pointed me to @dw's https://github.com/dw/mitogen - which to me looks like execnet done right
it also implements a lot of what i wanted to have in execnet but didn't get time to do

[dw]

Hi Ronny :)

Just as a 100 ft overview of execnet vs. Mitogen, execnet is /far/ more mature and the design is generally less adventurous. Of course that means less featureful, but it also, at least in the short term, it means more stable.

There are more notes in docs/comparison.rst .. it's not linked into the main docs because some of the commentary on that page started as private notes, and they're not.. necessarily too kind on others' projects, which needs fixed first.

[edit] Another potential gotcha with Mitogen is file transfer. A solution for it exists already in the Ansible extension, but it's not yet moved into the core library. Ideally rather than using that solution directly, there'd be a cute API for it. This stuff is all very rough around the edges as it's using the "service framework" which is still far from finalized. More unlinked docs here -- http://mitogen.readthedocs.io/en/latest/services.html

[RonnyPfannschmidt]

@dw based on featureset - mitogen does all the things that i wanted of execnet, and just needs a bit of love on the api side and much love on the stability side

that being said however - adding all the things i want to execnet would just make it unstable to begin with and it then would have to be re-stabilized,

also execnet has technically no more manpower behind it, its a zombie

[RonnyPfannschmidt]

closing this off - the multiprocessing experiment may need a repeat, mitogen is no longer maintained