[altq 866] RED implementation

[Nguyen-Tuong Long Le <le@cs.unc.edu>]

Hi all,

I am wondering why the drop_early function has a different drop
probability than in the original RED paper ( prob/(2 - count*prob)
instead of prob/(1 - count*prob) ). Does it mean that we have to
provide redd with a value of inv_pmax/2 instead of inv_pmax?

Thanks,
-- long