[Cfrg] On terminology (and essence): peer-to-peer or client-server

[Hugo Krawczyk <[e-mail]>]

It is very important that when people propose or discuss specific PAKE
protocols and applications, that they distinguish between those intended
for "peer communication" (where both ends share the same password) and
those for the client-server case (where the server does not store the
password but only a one-way function of the password).

These are different types of protocols, for different settings, different
security goals and  definitions, and quite different designs. In
particular, they cover very different applications (the client-server
setting is where most uses of passwords are found while the peer case seems
of more  restricted applicability, at least in the past).

It would be very helpful if we developed some common terminology for
differentiating these two settings. Unfortunately, there are many different
names people have been using. I like to call them "symmetric" (for the peer
case) and "asymmetric" (for the client-server case), but we have also
balanced and unbalanced, plain PAKE for the peer case and augmented PAKE
for the client server and more (e.g. "Verifier-based" for the asymmetric
case).

It would be good to set some conventions here.

But regardless of names, the most important point is to make clear to what
setting one refers when discussing specific protocols and applications.

Hugo