From: ietf [mailto:ietf-bounces(_at_)ietf(_dot_)org] On Behalf Of Christian
Huitema
Sent: Sunday, April 6, 2014 4:30 PM
To: ietf(_at_)ietf(_dot_)org
Subject: RE: Security for various IETF services
I agree with those who've said a threat analysis is needed before
deciding access is limited to TLS or other secure alternative.
But we have that threat analysis, and the recommended mitigation is
precisely "encrypt everything." The "pervasive monitoring" threat is analyzed
by a number of perpass drafts, and Stephen has merely followed the
conclusions of that analysis. There is no need to repeat that analysis for
each
and every tool that the IETF produces, and there is indeed a need for the
IETF as a whole to "lead by example."
I've been following this thread with some amusement, as it was clear that when
the perpass draft took its unusual journey from personal draft submission to
BCP the fallout would include things like an immediate call for all IETF
services to be encrypted. At the same time, there are an equally well known
number of attacks against TCP itself that make TCP the wrong substrate upon
which to build "secure" communication channels. (Just a couple of well-known
examples: the ease of session hijacking of established connection and SYN flood
attacks against well-known servers, both of which lead to denial of service and
potentially forcing the user to downgrade to an alternative channel for
obtaining the information).
If the same level of urgency were shown towards a viable, and secure,
replacement for TCP itself, then the calls for secure-only access to IETF
services might make sense. Instead, it feels a lot like requiring stronger
deadbolts on glass doors.
Matthew Kaufman