On Fri, Nov 15, 2013 at 6:14 AM, Hannes Tschofenig <hannes.tschofenig@xxxxxxx> wrote:
Am 15.11.13 11:33, schrieb l.wood@xxxxxxxxxxxx:These are the types of claims we will never find out whether they are true or not (even though it is not even clear what you mean by 'open' here and whether it refers to the confidential communication between the browser and the Web site).
If we had widespread encryption earlier, we wouldn't have an open web.
Rob McCool and myself were implementing public key systems on the Web back in 1993-4. So were EKR and Alan Schifman at EIT.
Have you ever tried to do RSA1024 on a 20MHz machine? How many connections do you think can be supported on a 200MHz server?
The reason the Web does not have built in crypto is due to the problems of performance and key distribution. The SSL ecology is not designed to provide confidentiality except as a side effect. It was designed to make shopping online as safe as shopping in a store.
What the 'SSL everywhere' schemes are offering is more encryption but with a weak or non existent authentication model. That is ok if all SSL everywhere is intended to do is to replace traffic that is currently sent en-clair without authentication. But this is a weaker security control than current SSL.
Deferring the hard problems till later is the reason the Web works. Attempting to boil the ocean is the reason Xanadu could never be finished. We can make very definite statements about what was practical in 1993.
The strategy is very interesting, the claim is that HTTP 2.0 with SSL is just as fast as HTTP 1.0. Which sounds good until you remember that the purported benefit of HTTP 2.0 is performance rather than security.
That might turn out to be a problem.
Website: http://hallambaker.com/