On Tue, 22 Dec 2015, Ilpo Järvinen wrote:
> On Tue, 1 Dec 2015, Wesley Eddy wrote:
>
> > This is the start of a working group last call for the PIE specification:
> > https://tools.ietf.org/html/draft-ietf-aqm-pie-03
> >
> > Please send any comments on the documents by the end of December.
> Section 4.4:
> - If there are no enqueue events, burst_allowance is being reduced by
> T_UPDATE timer. Could it occur first enqueue after idle sees
> burst_allowance <= 0 but non-zero drop_prob_? I suppose this might not be
> a problem as idle period would very likely mean that drop_prob_ is also 0
> so the random drop won't drop it regardless of burst_allowance, is this
> correct or is there some corner-case where drop_prob_ > 0?
Another hazard with this unconditional subtracting is an underflow. It
seems that at least the example code in Section 11 and 12 limit the
substraction of T_UPDATE to positive burst_allowance which avoids both
issues but it would be much better to have that in 4.4 code fragment
too.
--
i.
_______________________________________________
aqm mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/aqm
