Re: [core] 🔔 Confirmation call: CoAP over TCP/TLS #396: Revert L1 selection, select L3

[Carsten Bormann <[e-mail]>]

Hi Simon,

good to hear from you -- this is exactly why we validate in-room
decisions on the mailing list.

> I know one of the main concern of L3 in the pass was the possibility of
> head-of-line blocking, should we still consider bert draft that was
> proposed.

That is not a concern with L3, but a concern with large messages.
Yes, these do exhibit head-of-line blocking.
BERT is a good way to control the extent of this blocking, so I think
this is still a useful extension to pursue.

> The other concern was on how smaller device fit in this, if you send
> large payload that are out of scope for a constraint device, how should
> this be handle (if should be handle that the procol level)

As we said, there is no intention to revoke the SHOULDs in section 4.6
of RFC 7252.  However, there is also no intention to enforce these
SHOULDs (policy) by providing an artificial limitation to the message
size (mechanism) at the encapsulation level.

L3 does not prevent you from shooting yourself in the foot by ignoring
section 4.6 (note that L2 didn't really do that, either, 64 KiB is still
larger than 1152).  But L3 also doesn't stop you from transcending 4.6
significantly if you have a good reason to do so.

Grüße, Carsten