Re: [quicwg/base-drafts] How many probed paths should a server remember? (#3489)

[MikkelFJ <[e-mail]>]

I tend to agree with Eric, but I think there are some aspects to consider:

You don't want to limit yourself to one extra path as Ian suggests because you might have 3 or 4 paths forward, e.g. sittig on Wlan 2.5GHz with optional 5GHz if the first becomes congested, along with a 4G and 5G modem path. You need to monitor these and choose the best path forward once the preferred path fails. Maybe this is a bit extreme, but just to illustrate path migration isn't linear.

At the same time, the peer need to issue enough IDs to not stall progress in addition to accommodating the number of paths forward a typical client might need. I'd say having double the number of CIDs in spare is probably OK, but maybe there isn't a need for a spare CID for every path. So it could be number of paths + 2 or something.

Thus, a recommendation could be that a client SHOULD not track paths beying having 2 spare CIDs and a server SHOULD be prepared to track the number of issued CIDs minus two, as a general rule. Exceeding that may require delayed probing which could lead to temporary delays.

-- 
You are receiving this because you are subscribed to this thread.
Reply to this email directly or view it on GitHub:
https://github.com/quicwg/base-drafts/issues/3489#issuecomment-598420935