moq-wg / moq-transport

draft-ietf-moq-transport
Other
83 stars 22 forks source link

Clarify meaning of Timeout subscribe error #531

Open englishm opened 1 month ago

englishm commented 1 month ago

As discussed on the call today when looking at https://github.com/moq-wg/moq-transport/pull/529

englishm commented 1 month ago

None of the other errors in that section have detailed descriptions yet either, but all the rest are more obvious.

@fluffy mentioned this likely being something to do with relays needing to conserve resources. Can probably check relevant meeting minutes/recording for additional details.

englishm commented 1 month ago

@ianswett (on 2024-02-14):

As discussed today, it'd be nice to have one indicating "Timeout". https://github.com/moq-wg/moq-transport/issues/347#issuecomment-1944334099

The Datatracker doesn't have anything recorded for that date.

I think this must have been discussed on an authors' call before we switched to the weekly interim format.

Any additional detail from folks who were there for that discussion would be helpful.

fluffy commented 1 month ago

I have on a vague recollection of priori discussion of this so take with a huge grain of salt but it seems like we might need something to deal with the case of S sends a subscribe to relay R, R sends subscribe publisher P, P never returns any response to the subscribe. Eventually R will want to give up and tell S that the subscribe did not work. In this case R needs some error to return to S.

ianswett commented 1 month ago

I think @fluffy is correct.

I'm curious what HTTP proxies do in this case today? Without looking at our code, I think a timeout would get converted into a 500, but I'm unsure?

LPardue commented 1 month ago

I think my HTTP proxy tends to return a 502 or 504 depending on the exact circumstances

afrind commented 1 month ago

I'm pretty sure this is always a 504 in our proxy.