30 likes | 160 Views
RDDP: TCP Alignment David L. Black (RDDP WG Chair). RDDP WG rough consensus (mailing list): The RDDP protocols must allow implementations over unmodified TCP/IP stacks. Implication: receivers MUST cope with sender loss of alignment (or meddling middlebox) Observations:
E N D
RDDP: TCP AlignmentDavid L. Black (RDDP WG Chair) • RDDP WG rough consensus (mailing list): • The RDDP protocols must allow implementations over unmodified TCP/IP stacks. • Implication: receivers MUST cope with sender loss of alignment (or meddling middlebox) • Observations: • Those in favor of sender alignment appear to be opposed to telling the receiver that the sender is aligning (beyond normal TCP behavior) • There does not appear to be a functional requirement for this sort of notification
RFC 2119 Section 6:Use of MUST and SHOULD • Imperatives of the type defined in this memo must be used with care and sparingly. In particular, they MUST only be used where it is actually required for interoperation or to limit behavior which has potential for causing harm (e.g., limiting retransmissions) For example, they must not be used to try to impose a particular method on implementors where the method is not required for interoperability. • Observation: Alignment is not required for interoperation. • Question: Potential for causing harm? • NB: receivers can’t rely on alignment
TCP Alignment Summary & Proposal • Alignment is • Not required for interoperability • Not something an RDDP receiver can rely on • A potentially useful optimization • Proposal: Treat as optimization only • No “MUST” or “SHOULD” • Describe circumstances under which sender use of alignment improves performance • Discussion?