Re: [Tcpcrypt] disabling encryption in the middle of a connection (was Re: Draft charter text

[Joe Touch <[e-mail]>]

On 4/23/2014 3:58 PM, Tony Arcieri wrote:
> On Wed, Apr 23, 2014 at 3:37 PM, Joe Touch <[e-mail]
> <mailto:[e-mail]>> wrote:
>
>     No one needs to measure the power cost to know it's non-zero. If it
>     were, we'd have bigger problems (i.e., violation of the second law
>     of thermodynamics).
>
>
> It's still premature optimization, and will come at a non-zero cost to
> the protocol's security:

Security is a premature optimization too, in that sense. I.e., I already 
care about my cellphone's battery, but don't currently experience 
attacks on TCP connections that a TCP-level solution would be needed to 
secure. You keep claiming that nobody has measured performance impact of 
security, but that's not true (see refs below). The converse is that 
there's very little evidence of attacks at the TCP level of connections 
except between routers.

So if you're core point is that performance is a premature optimization, 
then so is TCP-layer security, and then there's little utility in 
pursuing the solution at all.

Joe

J. Touch, “Performance Analysis of MD5,” in Proc. ACM Sigcomm ’95, pp. 
77-86.

J. Touch, Y. Yang, “Reducing the Impact of DoS Attacks on Endpoint IP 
Security,” Proc. NPSec 2006, in conjunction with ICNP 2006, Nov. 2006.