Re: [TLS] Call for Consensus on removal of renegotiation

[Brian Hamon <[e-mail]>]

Yes, I do understand this is a discussion of renegotiation, and I'm in basic agreement with your proposal. However, it does leave unaddressed the issue I presented regarding the loss of the ability of the server to postpone demanding client authentication until the contents of one or more ApplicationData messages are examined.

I was trying to gather a list of objections that some have expressed of the use of renegotiation to deal with this specific situation, and as far as I can recall the only ones dealt with resumption. Am I forgetting any others?

My main concern is that if your proposal is accepted, what is the solution for a server that must examine some ApplicationData in order to determine if client authentication is required?  

-----Original Message-----
From: Martin Thomson [mailto:[e-mail]] 
Sent: Thursday, June 26, 2014 2:29 PM
To: Brian Hamon
Cc: [e-mail]; Joseph Salowey (jsalowey); <[e-mail]>
Subject: Re: [TLS] Call for Consensus on removal of renegotiation

On 26 June 2014 14:18, Brian Hamon <[e-mail]> wrote:
> I would argue that session resumption should revert the client's identity back to "unauthenticated". This avoids a lot of the complicated workarounds required for #2.

I hope that you realize that we're not talking about resumption right at this moment.  There are some interactions with resumption, but I think we're trying not to overcomplicate the issue by considering the implications of those just yet.

That said, if you did mean renegotiation in the above statement.  I don't think that making any recommendation to this effect will have any material impact on the sorts of bugs we've seen with renegotiation.  You can *say* that people should do this or that because it's good practice, but I think that that is just wishful thinking.