Re: making Linux a hospitable place for TTS engines like Voxin

[илья пащук]

I think such modification to speechd can be only a pro.

modules that can't do that will use an old way, but modules that is able 
will switch.

I think this should be not implemented as full new SpeechOutputMethod in 
the config, because this option is needed to define what method should 
be used at the end of the chain, no matter by speechd or module itself, 
but as fully new option.


about the protocol:

on start modules can start with old text based protocol, but on first 
stage can switch to new binary protocol, or it can be set with help of 
module file extensions. linux allows executables with any extension, why 
don't use it, or by exec parameters, or by the both last methods.

I think new proto should be an binary protocol working by the following way:

instead text lines systems should exchange MessagePack objects (msgpack.org)

this serializing system is very popular and supported in all major 
languages, and it allows for safe transport of raw byte sequences, but 
in the same time it allows to send strings, numbers, untyped arrays and 
maps, allowing to transfer complex data structures.

by this way we will create new, modern protocol for modules that will be 
in some places iwen simpler to use than our current protocol.