[noise] Protocol Names

[Jake McGinty]

> On Jun 17, 2017, at 4:14 AM, Trevor Perrin <trevp at trevp.net> wrote:
> 
> * Maybe allow "/" in addition to alphanumerics in primitive names?

+1, the “/“ is definitely aesthetically nicer than “-“ for primitive 
names.

> * I could also imagine parenthesized primitive names to allow arguments, e.g.:
> 
> Noise_XX+sig_25519+(RSA_PKCS1_2048)_AESGCM_SHA512
> 
> But I guess we could consider that later?

I intentionally chose a heavily restricted character set in this PR since 
it’s far easier to extend the allowed characters than to reduce it in the 
future. At the moment, all special characters that aren’t “_” or “+” are
reserved for later use as definable tokens.

I guess my question is: do we define the “/“ now even if it’s not being 
used in Rev33 or wait for a use case to be introduced where we can more 
definitively decide it fits our needs? My concern is that if we take a 
stab at defining a “/“ without any primitives that currently use this 
modifier, we’re limiting the possible changes we can make in the future
with few practical gains (unless an extension is proposed and generally
accepted between releasing rev33 and rev34).

~jake