Re: [webrtc-stats] Clarify qualityLimitationReason when limited by multiple reasons (#440)

> To expand on the above, my interpretation of the usecase for quality limitation was

I must admit that I don't know how the quality reason is determined, but what you write sounds very reasonable. I think there are problems in how these reasons are related though.

Any given point in time you might limit limit quality because of a reason. Assuming there are multiple reasons at this point in time, which reason is reported first would be implementation-specific, so straight away we might have a problem. But let's say we're only limited because of X.

OK so now you are limited because of X, and time passes. Things change, but you are still in a lower resolution. Are you also limited because of Y? Would you know without ramping up the resolution and seeing if you get problems? For bandwidth, maybe, can it estimate higher rates than current throughput? For cpu, I would assume we don't know unless we try and fail? But I don't actually know how these things are estimated.

Let's assume you are also limited by Y, you just have not experienced Y lowering the resolution because you're already lowered. OK. So which reason is the most limiting reason; X or Y? Could you know? Perhaps whether X or Y was the first reported reason earlier was a coin flip, and any attempt to test this would be flaky. Perhaps not, I guess it would depend on the implementation. The point is, X and Y are related, and I think we're getting into implementation-specific territory.

Options:
1. We define the reason to be the most limited factor, but because of the problems in determining this, we define this as an implementation-specific _best effort_.
2. We prioritize "bandwidth" over "cpu", which is Chrome's best effort.

Maybe even 2) is problematic. If you're "cpu" limited, would you be able to detect that you are also "bandwidth" limited? I'm not sure, but I'm hoping the answer is "yes" because of the bandwidth estimator.

-- 
GitHub Notification of comment by henbos
Please view or discuss this issue at https://github.com/w3c/webrtc-stats/issues/440#issuecomment-506694804 using your GitHub account

Received on Friday, 28 June 2019 11:01:54 UTC