[noise] Comments on the specification

Trevor Perrin trevp at trevp.net
Sun Apr 10 20:25:12 PDT 2016

On Sun, Apr 10, 2016 at 7:06 PM, Rhys Weatherley
<rhys.weatherley at gmail.com> wrote:
> I got a little confused with the description of handshake patterns vs
> message patterns in 5.3. Handshake State.  It got clearer on the third
> reading:
> - A handshake pattern (e.g. XX) maps to a sequence of message patterns, one
> for each handshake packet that is exchanged.
> - Each message pattern is a sequence of tokens indicating what to do in the
> current handshake packet.
> The information is there but my eyes missed it on the first two readings.
> The two different meanings of "pattern" got confused in my brain and I kept
> thinking I was looking at the other one.  Not sure how to clarify.  "Message
> template" instead of "message pattern" perhaps?  Patterns -> templates ->
> tokens.  Different words at each level.

I see your point, it's complicated that we're unpacking "handshake
patterns" into pre-messages and "message patterns", particularly
because we don't do careful definitions of all these until later.

We tried separate words earlier ("descriptors" vs "patterns") and that
felt like too much terminology, and harder to keep track which is

I'll think on this, not sure exactly what to do.

> In section 9.2 in the description of Noise Pipes there is a change over from
> using "..." to "-----" as a separator in the pattern definitions.  Are they
> the same thing, or something different?

Good catch, fixed in draft:


>  The diagrams aren't clear as to
> where the fallback happens.  The text indicates that the responder rejects
> the first IK message and then sends something else with a different packet
> type.  It may help if there were two diagrams for the whole sequence from
> the start, Noise_IK, and Noise_IK_to_XXfallback.  They will both have the
> same prefix, but with a comment saying "if failure happens here, switch to
> the other pattern and continue".

Yeah, probably a flowchart would make this clearer, so you could see
how the abbreviated flow connects back into the full flow.

I'll make a note of that, try to do something more later.


More information about the Noise mailing list