[Scone] Talking about the spin bit?

[Martin Thomson <[e-mail]>]

The editors are currently discussing whether to include a discussion in the draft about how SCONE interacts with the QUIC spin bit.

We're a little bit split on whether to say anything, so we have option A (do nothing) and option B (the text below).  Which do you prefer?  Why?

---
## Interactions With The Latency Spin Bit {#spin}

Endpoints that observe the recommendations for setting the latency spin bit,
as defined in {{Section 17.4 of QUIC}},
generate two competing signals for elements on the network path.

A QUIC packet that is coalesced with a SCONE packet
might make it harder for network elements to detect the latency spin bit,
because the SCONE packet moves the bit from the first byte of the datagram
to after the SCONE packet.

Network elements that ignore datagrams with a SCONE packet
could miss changes in the spin bit.
This could introduce measurement error
that inflates round-trip time estimates.

Endpoints that enable the spin bit can avoid this issue
by deferring the inclusion of SCONE packets
when a spin bit edge is being communicated in a QUIC packet.