On 7/6/2017 6:02 AM, [email protected] wrote: > > Ultimately, putting any non-data info in the SYN payload violates the > requirement that TCP options can be ignored by receivers that don't > support them *without* impacting the ability of *that* connection > attempt to succeed. > > [Med] You are right… if we are talking about TCP options that are > inserted in the payload of the SYN to every remote peer. This is not > the case for the MPTCP proxy case: this is about proxy-supplied data > that is sent to the proxy, which is provisioned by the provider. > Proxy-supplied data is not received by the remote peer. > You have not considered what happens when the remote peer changes or disappears.
In that case, your E2E TCP connection is no longer falling back safely and correctly, based on the principle that all unknown options can be ignored - as is required by RFC793. Further, I have no idea what you think is happening above, but at *best* it's TCP between the proxies, perhaps triggered by packets between the endpoints. However, it's no longer a TCP "connection" in any sense of any IETF standard of which I am familiar. Joe
_______________________________________________ Int-area mailing list [email protected] https://www.ietf.org/mailman/listinfo/int-area
