On 7/29/2014 8:48 AM, Eric Rescorla wrote:

On Tue, Jul 29, 2014 at 8:41 AM, Joe Touch <[email protected]
<mailto:[email protected]>> wrote:

    On 7/29/2014 8:36 AM, Eric Rescorla wrote:

        On Tue, Jul 29, 2014 at 8:33 AM, Joe Touch <[email protected]

             That would be my view as well; TLS should never be involved
             with or triggered by TCP.

        Can you say more about why?

    It's a payload mechanism. IMO, involving TLS with TCP is layer
    crossing, and I don't see any good reason for it.

Hmm... I'm not sure I follow this. We already use the TCP port
number to indicate which protocol you are speaking. The problem
is that it's not a negotiation. What's wrong with having a protocol
offer and selection mechanism  at the TCP layer (which is what
my draft basically is).

TCP ports are like a "next header" tag. Yes, I would have preferred it to be a header after TCP that identifies the session protocol, and we're paying for that not being the case.

However, TCP options are for changing the behavior of TCP - not of the next layer. TLS doesn't change TCP at all. I don't like using a TCP option as a flag on a port number.

Besides, TLS already works just fine without it.

Joe

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

Reply via email to