Re: [tsvwg] ECN encapsulation draft - proposed resolution

[Jonathan Morton <[e-mail]>]

> On 6 Jun, 2021, at 2:06 am, Bob Briscoe <[e-mail]> wrote:
> 
> RFC3168 reassembly does not even have any suggestion on what to do if framing boundaries overlap packet boundaries or where packets are smaller than frames - so it is not 'known-to-work' in any case, let alone in every case.

Actually, it does offer such a suggestion, if you read between the lines a little bit.

The basic premise that both the marking and reassembly rules are based on is: "if you would have dropped a Not-ECT packet, try marking an ECT packet with CE instead".  If any fragment of a packet is dropped, then the whole packet is effectively lost; this is mirrored in the rules that propagate a CE mark to the whole reassembled packet if any fragment of it was so marked.

Applying the same logic to the case where a single Layer 2 frame may span (parts of) more than one IP packet, dropping that frame would cause all the packets it (partially) spans to be lost.  Therefore, the behaviour implied by RFC-3168 would be to propagate a congestion mark found on the Layer 2 frame to all of the packets that would have been lost if that frame were dropped.

We can debate whether this is better or worse than converting a single Layer 2 mark into a single CE mark, but the logic is at least consistent this way.

 - Jonathan Morton