Closed ekinnear closed 6 months ago
This was Lars's contribution. I'm not sure there's anything to do here given what's already in the draft.
Agree, let's keep the issue open and re-read the draft after all edits - at most I expect this would need to be called out earlier.
@ekinnear In 4.2, we have: "In contexts where differing congestion control algorithms are used, it is important to understand whether the proposed congestion control algorithm could result in more harm than previously defined algorithms to flows sharing a common bottleneck."
Eric is just recording Lars's comment at the mic. I'm fairly sure Lars hasn't read the draft, so as you said, let's make sure this is covered when we re-read it.
IMHO "previously defined algorithms" is not exactly what we mean, and this could be made more precise: "previous standards-track algorithms (e.g. {{!RFC5681}}, {{!RFC9002}}, {{!RFC9438}})"
AFAICT the goal is to consider harm induced relative to the harm induced by Reno/CUBIC. There may be other experimental algorithms that become "defined", which are not intended as a yardstick by which to measure harm.
So I have proposed an edit for that: https://github.com/ietf-wg-ccwg/rfc5033bis/pull/108
We also have recommendations about bufferbloat and packet losses, which are also very much about "not causing harm to others". That plus #108 means we can probably close the issue.
Raised at IETF 119: We may want to consider mentioning that you shouldn't impair the path for others.