- From: Stefan Eissing <stefan.eissing@greenbytes.de>
- Date: Tue, 18 Nov 2014 11:06:18 +0100
- To: Poul-Henning Kamp <phk@phk.freebsd.dk>
- Cc: Martin Thomson <martin.thomson@gmail.com>, Jason Greene <jason.greene@redhat.com>, Roland Zink <roland@zinks.de>, ietf-http-wg@w3.org
There is no known technology that covers all use cases and security issues of http. Arguing that tech X is bad because of attack vector Y, is always possible for a significant amount (I think almost all) of use cases. A discussion has to take the use case first, investigate threats vs. risks and find the most suitable tech. Therefore, differentiation between "browsing the net" and "connecting to a local ip" is helpful. Discussing mandatory security tech for all uses of http is not. There are more use cases than all of us know combined. If h2 should be a successor of http, forcing tls is not feasible in practise. I think almost all of us know this use case: - I myself need a browser that allows me to accept "invalid" security. By choice. - My mother *should not* have a browser that ever does that, nor should she ever be asked to. - It would be great if my mother could connect to the DSL box and tell me what possible error she sees there. With a browser that she knows. - Installing self-signed certs beforehand on all involved systems/trust stores that do not expire before I need them is mission impossible. //Stefan > Am 18.11.2014 um 09:11 schrieb Poul-Henning Kamp <phk@phk.freebsd.dk>: > > -------- > In message <CABkgnnVWze4YVTfgVc-+9DRTgGdG86xmHbySB=g2uDoyvQ_S=w@mail.gmail.com> > , Martin Thomson writes: > >>> Even better would be to support anonymous ECDH. Why bother requiring all >> of these fake certs to be generated when they have no legit purpose. >> >> That at least is an easy one to answer. If your handshake looks different >> (and any anonymous mode will, unless you use TLS 1.3 and some aggressive >> padding), then you open an invitation to MitM. > > This is exactly the kind of crap-think I tried to warn against in > my FOSDEM keynote: > > The point was *not* to defend against MitM but pervasive monitoring. > > Your attitude there, shared by far too many TLS-heads is like the > parents who forego child immunisations, because their kids might > feel unwell for a couple of days afterwards. > > ECDH would do *wonders* against pervasive monitoring, it would render > almost all of NSAs take worthless to them, and you cannot do a MitM > with a passive splitter. > > It's the same stupid attitude which makes browsers treat self-signed > certs as radioactive waste. > > That attitude, is a BIG part of the problem, and contributes nothing > to the solution for pervasive monitoring. > > -- > Poul-Henning Kamp | UNIX since Zilog Zeus 3.20 > phk@FreeBSD.ORG | TCP/IP since RFC 956 > FreeBSD committer | BSD since 4.3-tahoe > Never attribute to malice what can adequately be explained by incompetence. > <green/>bytes GmbH Hafenweg 16, 48155 Münster, Germany Phone: +49 251 2807760. Amtsgericht Münster: HRB5782
Received on Tuesday, 18 November 2014 10:06:44 UTC