[Cryptech Core] Next steps and work packages

[Rob Austein]

> >> This really is smack in the middle of the green/blue
> >> boundary. This actually implies that we probably will have more
> >> than one implementation of the blue/green boundary. One for
> >> tightly interconnected FPGAs and one for less tightly (say USB,
> >> Ethernet). This can naturally be combined using wrappers but
> >> something we need to consider.

Note that Joachim said "more than one implementation" here, not "more
than one API".  This is, I think, the correct way to think about this.

> >  o there is a brutal blue/green api along the lines of which we spoke
> >    some weeks back, a fifo implemented as a ring buffer, with elements
> >    being tlvs.  this would be in shared memory in some way.
> 
> Is protobufs too much baggage for embedded or can we reuse that?
> > 
> >  o for test setups, we would 'tunnel' with some sort of shim adapters.
> >    the blue and green would each be dealing with the same code, but
> >    each would be talking to an emulator which tunneled the abstraction
> >    to the distant partner's end where it would re-simulate that end of
> >    the api.

This is all getting down into implementation details, which is, I
think, premature until we know what we're implementing and what the
real requirements are.  Semantics first, please.