ietf
[Top] [All Lists]

Re: Mandatory encryption as part of HTTP2

2013-11-15 01:14:41
Please follow on the HTTPbis list and contribute there.
We don't need to rehash this again here.
-=R


On Thu, Nov 14, 2013 at 8:41 PM, <l(_dot_)wood(_at_)surrey(_dot_)ac(_dot_)uk> 
wrote:

+1

mandating encryption is unwise.

Make encrytpion attractive. Make it easy to use and to deploy. Make the
risks of not adopting encryption clear.

But NEVER make it mandatory.

Lloyd Wood
http://sat-net.com/L.Wood/

If we had widespread encryption earlier, we wouldn't have an open web.
________________________________________
From: ietf-bounces(_at_)ietf(_dot_)org [ietf-bounces(_at_)ietf(_dot_)org] On 
Behalf Of Iljitsch
van Beijnum [iljitsch(_at_)muada(_dot_)com]
Sent: 14 November 2013 20:42
To: ietf(_at_)ietf(_dot_)org
Cc: iab(_at_)iab(_dot_)org
Subject: Mandatory encryption as part of HTTP2

Forgive me if this has been discussed before, but I haven't been active on
this list for a while and I didn't see subject lines that indicated recent
discussions on this.

Apparently the chair of the httpbis wg is proposing to make encryption a
mandatory part HTTP version 2:

http://lists.w3.org/Archives/Public/ietf-http-wg/2013OctDec/0625.html

I have many medium-sized problems with this, including the issues with
CAs, the additional fragility of depending on certs with limited lifetimes,
performance and energy efficiency issues (both the batteries in mobile
hosts and the power use in datacenters), severely reduced cacheability and
debugging which are reasonable tradeoffs when privacy and authentication
are needed, but are wasteful when they're not, which is still very often
the case.

But a more fundamental problem with this approach is that it ties HTTP2 to
TLS, while TLS is not a very good technology, except that it has proven
easy to deploy. When we finally figure out how to get IPsec deployed as a
general purpose solution for privacy and authentication, it would be quite
annoying to have to run TLS, too, because HTTP2 requires it.

I'm not entirely sure why the existing problematic solution with certs and
CAs was proposed here, as the intended goal, keep the NSA and friends out
of our business, would be hard to reach that way. Using some kind of
opportunistic encryption would serve that purpose much better, IMO.

(However, I do think there is value in making it possible to enable
encryption when needed/desired without requiring the use of the https URL
scheme.)

Deliberating exactly these kinds of issues is why the IAB gets paid the
big bucks. So I hope the IAB can take on this issue.

Iljitsch