* Deadlock freedom and buffer space control. TCP implementations
provide control over buffer space via socket options such as
SO_SNDBUF
and SO_RCVBUF. Applications should be able to send data
simultaneously in both directions without deadlocking if the data
fits
in the specified
On Tue, Nov 18, 2014 at 02:33:46AM +, Bob Briscoe wrote:
I think the point Ekr (and Nico) are making is that TLS is really a
control channel within the TCP Data, so if an app understands that
distinction, it can preempt the handshake. Therefore it /can/ read
and process such control
On 11/18/2014 9:59 AM, Nico Williams wrote:
On Tue, Nov 18, 2014 at 02:33:46AM +, Bob Briscoe wrote:
I think the point Ekr (and Nico) are making is that TLS is really a
control channel within the TCP Data, so if an app understands that
distinction, it can preempt the handshake.
On Tue, Nov 18, 2014 at 10:16:50AM -0800, Joe Touch wrote:
On 11/18/2014 9:59 AM, Nico Williams wrote:
On Tue, Nov 18, 2014 at 02:33:46AM +, Bob Briscoe wrote:
I think the point Ekr (and Nico) are making is that TLS is really a
control channel within the TCP Data, so if an app
On 11/18/2014 10:46 AM, Nico Williams wrote:
AFAICT that would require the specification of a new API over TCP;
otherwise, you have to translate TLS's expectation of a reliable,
byte-ordered stream into what TCP provides within TCP (which is neither).
Not in the design I gave, which could
On Tue, Nov 18, 2014 at 03:10:58PM -0800, Joe Touch wrote:
I'm speaking of the private interface. Yes, it could be
implementation-specific, but if you are intending to define a complete
spec for using SSL over raw TCP segments you first have to explain what
you expect from raw TCP segments.
On Tue, Nov 18, 2014 at 04:02:32PM -0800, Joe Touch wrote:
On 11/18/2014 3:51 PM, Nico Williams wrote:
On Tue, Nov 18, 2014 at 03:10:58PM -0800, Joe Touch wrote:
I'm speaking of the private interface. Yes, it could be
implementation-specific, but if you are intending to define a complete
Scharf, Michael (Michael) michael.sch...@alcatel-lucent.com writes:
* Deadlock freedom and buffer space control. TCP implementations
provide control over buffer space via socket options such as SO_SNDBUF
and SO_RCVBUF. Applications should be able to send data
simultaneously in both
On Tue, Nov 18, 2014 at 05:31:54PM -0800, Joe Touch wrote:
I think we're agreeing, but my concern is how much the early data
handling by the system needs to be spec'd out, not just left to the
implementation.
Joe,
No one is disagreeing with this.