Now that I'm out of bed and at a device with a real keyboard, I can
elaborate.

I hope Mark will forgive me, but I've envisioned the birth of tcpcrypt as
something like: Student: "Mark, you know that conversation we had two weeks
ago about research projects related to MPTCP?  I figured out how to do low
overhead encryption for TCP."  Mark: "That's great.  We can bundle it with
MPTCP and take it to the IETF. Maybe we can find a way to make it generally
useful."  :-)

There is a trend toward putting capabilities in what I'm going to call the
session layer (if you don't like that I'm glad to call it something else -
that's not the main point), of which QUIC is an example.  Partly this is
because the behavior of the transport layer isn't very consistent anymore.
 The great thing going on in HTTPbis, RTCWeb, etc. is they might get a good
level of behavioral consistency.  And if they do, they will attract even
more feature building.

This leads me to believe that that level/layer is a really good place to
put consistent security capabilities.

Scott
_______________________________________________
perpass mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/perpass

Reply via email to