Re: Fwd: New Version Notification for draft-nottingham-http2-encryption-02.txt

I'm pretty sure this argument (there are free certs so we should all use 
them for everything) has been floated and sunk about 3 times on this 

Maybe we need some place where we can collect these arguments and the 
results of them so we can post referrals to that place instead of doing 
that work over and over?


------ Original Message ------
From: "Brian Smith" <>
To: "Stephen Farrell" <>
Cc: "William Chan (陈智昌)" <>; "Paul Hoffman" 
<>; "HTTP Working Group" <>
Sent: 15/12/2013 09:40:58
Subject: Re: Fwd: New Version Notification for 
>On Sat, Dec 14, 2013 at 11:20 AM, Stephen Farrell 
><> wrote:
>>Possibly a different thread really but...
>>On 12/14/2013 05:20 AM, William Chan (陈智昌) wrote:
>> > Anyhow,
>> > we don't support any type of opportunistic encryption, especially
>> > unauthenticated. We want people to use https://, therefore we more 
>> > less only plan to support HTTP/2 for https:// URIs. Let me know if
>> > this still leaves anything unclear.
>>What that leaves unclear for me is how the current 30-40% of web
>>sites that are setup for some form of TLS will suddenly become
>>99%. Without some other action on helping sites get certs, it
>>just won't happen would be my prediction.
>We need to focus our effort on that problem.
>There are already at least three commercial CAs, that browsers trust, 
>that give away free certificates: StartCom (restricted to non-business 
>use), GlobalSign (restricted to open source projects), and GoDaddy 
>(restricted to open source projects). These CAs give away an inferior 
>good (presumably) in the hopes of you eventually upgrading to their 
>non-free goods. The main problem with these CAs' freemium models is 
>that their decision process for whether you qualify for the free 
>product isn't (and cannot be) automated. However, I believe there is an 
>opportunity for us (browser makers in particular, and the IETF 
>community in general) to create a new kind of inferior good in the 
>certificate space that CAs (possibly other than the ones I mentioned) 
>may be willing to give away for free in a way that allows CAs to be 
>comfortable with, without jeopardizing their businesses. Note: when I 
>say "inferior good," I use "inferior" in the economic sense only; I 
>think we'd insist that such certificates have security properties at 
>least as good as what we already accept as the minimum in browsers 
>Even if such efforts were to fail, we still wouldn't be at the point 
>where completely unauthenticated encryption is the only option left. 
>There are other ways of authenticating servers than punting to a 
>commercial CA. We should make sure we have thoroughly exhausted these 
>alternatives before giving up.
>>I think its all the more puzzling when contrasted with other cases
>>where people claim that we can't do X because that'd cause a problem
>>for 1% of the web, but yet here you seem to be saying its ok to
>>do this when it'd cause a problem for 60-70% of the web. (I don't
>>recall whether or not you've made such claim William.)
>When it comes to breaking interoperability or regressing performance, 
>small percentages like 1% matter. The fact that most connections web 
>browsers make are not encrypted+authenticated is a huge problem that 
>needs to be addressed with strong action, but it isn't acute in the way 
>that a compatibility or performance regression is.
>Difficulty with certificates doesn't explain why,, 
>,,, and other top sites aren't 
>HTTPS-only. Social issues (wikipedia has been very open about how 
>politics affects their HTTPS deployment) and performance issues are 
>much more serious issues, and those issues won't be properly addressed 
>by adding opportunistic encryption to HTTP/2.
>Do third-party advertising sites (the kind whose cookies are being used 
>to de-anonymize users) use HTTP instead of HTTPS because they can't 
>afford certificates? No. Performance, scalability, the pain or 
>migrating websites from http:// to https:// URLs, and lack of 
>motivation seem to be the problems. Web browsers can encourage them to 
>move to HTTPS by getting them on our HSTS preload lists (so the browser 
>"fixes" those http:// links to https:// links automatically) and by 
>doing other things. For example, at Mozilla we've long had a desire to 
>strip cookies from third-party requests that aren't HTTPS. It seems 
>like now is the time to figure out how to make that work. We've already 
>seen big advertisers make changes like this to accomodate our recent 
>mixed-content blocking changes. I'm confident that such advertisers 
>would be willing to accomodate further changes, if nudged a little bit.
>>Even if only as a backup in case that 30-40% -> 99% transition
>>fails, I'd hope folks do continue working on ways to provide
>>opportunistic encryption for HTTP/2.0.
>I agree that it is reasonable to continue to explore unauthenticated 
>encryption options. However, I encourage people to support the efforts 
>to go further--to try to hit a home run instead of trying to bunt.
>>On the current draft - its seems quite odd to ignore the existing
>>anon-DH ciphersuites when trying to do opportunistic encryption.
>The way cipher suites are currently negotiated in TLS, with the client 
>saying which cipher suites it supports and the server choosing one, 
>suffers from the same problem that ALPN causes for http2-tls-relaxed: 
>the client is telling potential MitMs whether or not they will get 
>caught. I appreciate, and agree with, the fundemental aims of the 
>perpass effort. However, I think way too much emphasis is being put on 
>the "passive" part. We need to remember that perfect is the enemy of 
>the good, but at the same time it would be unfortunate to spend a huge 
>amount of effort trying to prevent passive attacks while making active 
>attacks easier to carry out.
>Mozilla Networking/Crypto/Security (Necko/NSS/PSM)

Received on Sunday, 15 December 2013 22:14:23 UTC