Re: [TLS] proposal to encrypt ContentType for TLS 1.3

[[e-mail] (Martin Rex)]

Daniel Kahn Gillmor wrote:

> On 07/14/2014 10:12 AM, Martin Rex wrote:
>> Daniel Kahn Gillmor wrote:
>>> The fact that network-facing code won't know with certainty when a
>>> handshake completes and when application data starts flowing seems like
>>> a feature, not a bug, if we want to protect the communication.
>> 
>> I definitely see it as a feature.
> 
> I'm glad to hear it :)

Ooops, my hands wrote the opposite of what my mind was thinking of.

I meant to say that it is a feature of TLS to be able to distinguish
handshake phase from application data phase on the outside.


> 
>> The bug here would be an unnecessary difference to existing TLS protocol versions.
>  [...]
> > Drop-in replacement of only the TLS stack will be impossible.
> 
> Yes, I'm suggesting that for people in the unusual situation of having
> moved their network i/o state machine entirely outside of their TLS
> stack, they'll want to update their network i/o state machines first (to
> understand the new traffic patterns), before upgrading their TLS stacks
> to support TLS 1.3.


That is not ususual at all.  It is actually the most reasonable approach
if you want to do network I/O with arbitrary mixtures of non-blocking
and application-specified timeouts.   Microsoft's SChannel SSP AFAIK has
always been transport-free.  In addition to non-blocking, I also
implemented server-side "TLS extension SNI" purely at the application
layer, the server-side of the SSL stack doesn't care about it.


> 
> It seems entirely possible that this will be the case anyway, given the
> new handshake flows proposed for TLS 1.3.
> 
> I don't think this is a good reason to avoid the improvement.


The more bogus and backwards-incompatible changes go into TLSv1.3,
the more unlikely it will be that we will ever support it.


-Martin