Re: [TLS] In support of encrypting SNI

On 14/05/14 21:35, Salz, Rich wrote:
> Dan, thanks for writing the passionate and detailed note.
> 
>> . Surely any crucial increase of free expression we here can give
>> others out weighs technicalities.
> 
> I just want to let you know that, somewhat sadly, I disagree with the
> quoted sentence, and I'm not alone.  (Even the author of the "it's an
> attack" RFC has said that barring good technical solutions we're
> unlikely to do it.) Encrypting the handshake will not prevent passive
> surveillance. From a technical view, it's not clear it provides
> enough privacy to justify the non-inconsiderable costs.

I think you've slightly misinterpreted me there Rich, I'll try be
clearer:-)

Yes we do need a good technical solution and we don't yet have
one written down. There is an of course obvious technical solution
for at least passive attacks, but that does have the significant
and real negative aspect that you pointed out, so is perhaps not
good enough by itself.

I'm also less concerned with monetary costs (most who are running
large numbers of TLS instances like you guys could afford some
monetary costs:-) but more with the architectural problem that the
TLS stacks running in different VMs would not be willing to agree
on SNI protection algorithms etc. should a solution require those
to be the same across all VMs. But I'm not at all clear that's
insurmountable really.

So same point I made before: we need folks who want this to put
in the work to write down their solutions. (And the corollary is
that if we don't figure out how to protect SNI then we need to
document why fairly well, not necessarily in an RFC, but in some
kind of referenceable thing.)

Cheers,
S.

PS: And again, wearing no hats for this conversation.

> 
> /r$
> 
> -- Principal Security Engineer Akamai Technologies, Cambridge, MA IM:
> rsalz@jabber.me; Twitter: RichSalz
> 
> _______________________________________________ TLS mailing list 
> TLS@ietf.org https://www.ietf.org/mailman/listinfo/tls
> 