Re: [Lake] LoRaWAN use case; Re: WGLC for draft-ietf-lake-reqs-01

[Göran Selander <[e-mail]>]

Hi Dominique,

(trimmed down as instructed)

I’m trying to compare your challenge against the current benchmark: 1 message per packet at SF12.

Since there is no automatic mapping between uplink/downlink and flights of the AKE we need to consider both cases. (One potential way to address collisions between multiple devices accessing at the same time is that the device triggers the other endpoint start the AKE, which can act as initiator and schedule the AKE with each device.)

For the case of message 1 and message 3 being uplink, assuming key exchange protocol messages of size 51 bytes at link layer gives an air-time equal to 5.6 seconds (assuming the spreadsheet I’m using is correct). This is in the higher end of the interval 4-6 seconds of your challenge. (For EDHOC with raw public keys based ECDH the corresponding air-time is 4.4 seconds.)

For the setting where message 2 is uplink, there is a need for an additional message to trigger the AKE. Assuming a few bytes of trigger and 51 bytes at link layer in message 2 this gives an air-time for uplink slightly above 4 seconds (assuming the spreadsheet is correct; similar for EDHOC RPK).

So, the current benchmark and available data points are all compliant with your challenge.

Now message 2 is typically the largest message in the AKE, and as you acknowledged there is an uncertainty and potential cost with downlink (and also with simultaneous access). I think a benchmark should include all messages of the AKE and I would propose that we don’t formulate it in terms of uplink/downlink.

What do you think? Does the current benchmark make sense or would you prefer more restricted message sizes?

Göran