Re: [tcpm] A possible simplification for AccECN servers

["Richard Scheffenegger" <[e-mail]>]

My reading of bob’s suggestion is

Syn - accecn
Syn-ack no ect

I don’t see any problem there ;

The handshake is only complete (and the server continues to use ip ect in the final ack), if the client responded by either non-ect synack(ae/esce =0, cwr=0, ece =0), a rfc3168 
Synack(ae/esce=0, cwr=1, ece=0), or one of the four synack accecn codepoints;

The only violation would be a server trying an ecn++ / accecn handshake with a ect0/ect1 marked syn...

Am i missing something?

Richard
Gesendet mit der GMX Mail App

Am 23.11.19 um 15:06 schrieb Jonathan Morton

> I believe that taking advantage of such a provision would result in a
> violation of RFC-3168, as seen by a server supporting that specification
> but not AccECN.
> 
> If I happen to be wrong about that, could you explain how the server gets
> to know that the connection has reverted to Not-ECT mode after sending a
> SYN-ACK confirming RFC-3168 support in response to a valid RFC-3168 SYN?
> 
> - Jonathan Morton