On Thursday 24 January 2008 19:51, Michael Rogers wrote:
> Matthew Toseland wrote:
> >> This time {SUB_MAX_PING/MAX_PING} as the reasons. Examining the source
> >> I find that the reject is always computed as an average across all
> >> nodes.
> >>
> >> Usually that's fine, but one of my opennet peers was CRAZY SLOW: 92B/
> >> sec delay 11129ms (RTT 16347ms window 1.4688538312911987
> >>
> > We take the median of the average ping time for each node. Then we smooth
it
> > over a bit over time.
>
> Would it make sense for rejection to be based on the median of the
> peers' median ping times, rather than the mean of the peers' median ping
> times? That would exclude outlying peers as well as outlying ping times.
Umm...
It's the median of the peers' mean ping times. We further smooth it by a
time-based klein filter.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 189 bytes
Desc: not available
URL:
<https://emu.freenetproject.org/pipermail/devl/attachments/20080124/7baedf8a/attachment.pgp>
