Re: Data Channel Buffer Management Proposal

Yeah, I think that problem is independent of this.  If you watch
bufferedAmount using a fast polling loop, you have the same/worse problem.

FYI, the latest draft changes the names to "bufferedAmountLowThreshold" and
"onbufferedamountlow".

On Thu, May 28, 2015 at 12:29 PM, Martin Thomson <martin.thomson@gmail.com>
wrote:

> On 28 May 2015 at 07:38, Benjamin Schwartz <bemasc@google.com> wrote:
> > Pull request is up: https://github.com/w3c/webrtc-pc/pull/233
>
>
> Looks reasonable.
>
> I'll note that this feature, in any of its forms, potentially causes
> something like silly window syndrome, where clients watch for the
> buffer to go below x, at which point they feed in exactly enough data
> to lift the buffer to x, resulting in lots of busy work.  I guess that
> the response to that would be that the acknowledgements from the peer
> would have to be silly in order for that to be a problem, so we don't
> need to worry about it.
>

Received on Monday, 1 June 2015 16:01:25 UTC