[core] draft-eggert-core-congestion-control

["Stok, Peter van der" <[e-mail]>]

Dear Lars Eggert,

I read your draft and had some additional thoughts, not solutions, on the congestion problem.
Congestion in wired networks is caused by routers not consuming and forwarding the packets.
Congestion control also reacts to the rate at which the destination can consume the packets.
The loss of packets has an additional non-intended effect on the congestion behavior.

For low-resource wireless networks, one of the dominating effects may actually be the low link throughput and the interference by many senders, more than the throughput of the routers. The saturation of space around the sender translates itself in long back-off waits.

Consequently, may it not be more natural to measure the back-off time that a packet suffered and use that as indication for the congestion control?

Also I should like to react to the 2-3 second time-out interval. In lighting control applications I expect that most messages will remain local inside the small subnet. I expect that response times of 300 ms are already too long. Consequently, messages may have a deadline of 300 ms attached to them. Once the backoff sequence exceeds 300 ms, a large part of the messages will be rejected and not sent. Accordingly, I expect that steering on deadline may already have a beneficial effect on the congestion problem.

Looking forward to your answer,

Peter van der stok

The information contained in this message may be confidential and legally protected under applicable law. The message is intended solely for the addressee(s). If you are not the intended recipient, you are hereby notified that any use, forwarding, dissemination, or reproduction of this message is strictly prohibited and may be unlawful. If you are not the intended recipient, please contact the sender by return e-mail and destroy all copies of the original message.