Document Limit on Peers MTU

[mikkelfj]

The current transport draft only discusses minimum MTU. However, a large MTU drains resources on the receiving end and could lead DoS attacks on validation of invalid very large packets. IPv6 has a limit of 64K, but apparently there can also be gigabyte sized jumbograms. Futher, this adds an implicit dependency on IPv6 that might be overlooked when adapting QUIC to other transports.

A receiver can always drop large data grams and thus force a MTU limit, but perhaps some text is needed to make this explicit?

[marten-seemann]

We already have the max_packet_size parameter in the transport parameters.

[mikkelfj]

Of course I overlooked max_packet_size in transport parameters.

But I still think some mention is needed in section 9.2 https://quicwg.github.io/base-drafts/draft-ietf-quic-transport.html#rfc.section.9.2

[mikkelfj]

We already have the max_packet_size parameter in the transport parameters.

yes I just realized

[martinthomson]

PR welcome.

[martinthomson]

After looking at this, this is fine.