-
Notifications
You must be signed in to change notification settings - Fork 14
Conversation
See #132. |
Does this mean that we don't need to update 5681 anymore, because 7661 OK'ed CUBIC's behavior? Should 7661 have updated 5681? |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Please fix Markdown nits and add changelog entry.
The issue I think is: RFC 7661 describes a framework for application-limited flows to determine an appropriate cwnd. It's a relaxation with caveats also about how a flow needs to behave when an "unvalidated" use of cwnd was not safe (i.e. the "normal" case of using cwnd proved to be unsafe). RFC 7661 could have been a PS and could have updated 5681 - at the moment neither is the case. As far as I know, some Cubic implementations do part of what RFC 7661 says, and others might do all. The final sentence of the suggested PR acknowledges this, without trying to interpret this, seeking to avoid blocking publication of this specific PS. I'll re-echo your question to others: Should 7661 have updated 5681? Should it become PS? |
At the moment, this document is intending to update 5681 to allow CUBIC's behavior. Since that is on the "extreme" side of what 7661 allowed, maybe 7661 should? |
This still needs a changelog entry, but is otherwise good to go. |
@gorryfair please add a changelog line. |
@larseggert @goelvidhi Happy for someone to add a changeling entry. |
Take 2 at a PR is intended to amend the text to align with my understanding of the RFC-series. This normatively would require ssthresh to be set from flight_size; however, this is explicitly relaxed in RFC7661, so I suggest wording should be careful around this topic. I retained the original text observation that in fact some cubic implementations are known to set ssthresh from cwnd.
remove whitespace Co-authored-by: Vidhi Goel <goel.vidhi07@gmail.com>
That's OK Co-authored-by: Vidhi Goel <goel.vidhi07@gmail.com>
I think this may be obvious from the preceding section, but it seems technically more correct, so I'll commit this. Co-authored-by: Vidhi Goel <goel.vidhi07@gmail.com>
Now able to edit. md file fixed to remove trailing whitespace.
This PR is intended to amend the text to align with my understanding of the RFC-series. This normatively would require ssthresh to be set from flight_size; however, this is explicitly relaxed in RFC7661, so I suggest wording should be careful around this topic. I retained the original text observation that in fact some cubic implementations are known to set ssthresh from cwnd.
Note: I expected the REF to "RFC7661" to be informative (EXP).