Be careful with the terminology, since there are two types of head drop: 1) head drop based on sojourn time 2) head drop based on current queue depth
That paper used type 2. In type 1, the drop is based on the queue that existed at the time "this" packet was enqueued (i.e. the packets that arrived *before* this packet), and thus is equivalent* to tail drop. In type 2, the drop is based on the queue caused by the packets that arrive *after* this packet. * modulo the difference in predicted queue delay vs measured queue delay. Type 2 detects and signals congestion more quickly than tail drop. Type 1 is the same as tail drop. -Greg On 10/10/21, 12:32 PM, "Bloat on behalf of Dave Taht" <[email protected] on behalf of [email protected]> wrote: We've never really characterized the differences between head and tail drop. A good tidbit went by here, where in one experiment 15 v 22 drops were observed. https://journal.ub.tu-berlin.de/eceasst/article/view/1137/1092 Last paper I'm forwarding for the day! -- Fixing Starlink's Latencies: https://www.youtube.com/watch?v=c9gLo6Xrwgw Dave Täht CEO, TekLibre, LLC _______________________________________________ Bloat mailing list [email protected] https://lists.bufferbloat.net/listinfo/bloat _______________________________________________ Bloat mailing list [email protected] https://lists.bufferbloat.net/listinfo/bloat
