While I certainly support the idea of confidential DNS, I wonder whether it is 
a good idea to impose the overhead involved in TLS on the high volume name 
servers?
--
Glen Wiley
KK4SFV
Sr. Engineer
The Hive, Verisign, Inc.

From: Ted Hardie <[email protected]<mailto:[email protected]>>
Date: Wednesday, November 13, 2013 11:16 AM
To: Martin Thomson <[email protected]<mailto:[email protected]>>
Cc: "Wiley, Glen" <[email protected]<mailto:[email protected]>>, perpass 
<[email protected]<mailto:[email protected]>>, Stephane Bortzmeyer 
<[email protected]<mailto:[email protected]>>, Andy Wilson 
<[email protected]<mailto:[email protected]>>, Stephen Farrell 
<[email protected]<mailto:[email protected]>>
Subject: Re: [perpass] DNS confidentiality

On Tue, Nov 12, 2013 at 5:16 PM, Martin Thomson 
<[email protected]<mailto:[email protected]>> wrote:
On 12 November 2013 08:12, Ted Hardie 
<[email protected]<mailto:[email protected]>> wrote:
> The DNS query tells you which resource was the target even if the HTTP flow
> was protected by TLS.

In practice, since server name indication is sent in the clear, even
this doesn't help.  Unless you are running a browser from 2001, you
are sending SNI.

That said, SNI may be pushed into an encrypted payload in TLS 1.3.
The challenge there is that servers often use SNI to select what
credentials to offer.

True; I'd been thinking about the blogspot-style use cases where you get an 
initial negotiation at one name followed by a large set of alternate names, but 
that's not the common case.  The VPN case is still an issue, though.

Having read through the rest of the thread, pushing SNI into the encrypted 
portion of TLS in 1.3 seems like a good thing to do.

Ted
_______________________________________________
perpass mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/perpass

Reply via email to