BILLW@SCORE.STANFORD.EDU (William "Chops" Westfield) (11/03/86)
how about a couple of TCP options: This is retransmission # n This contains an ACK for retransmission #n or Timestamp This is an ACK for packet containing TIMESTAMP. This would enable RTT's to be determined exactly. Of course, it's rather late to be adding things to the TCP spec. Sigh. billW -------
MILLS@A.ISI.EDU (11/07/86)
In response to the message sent 5 Nov 1986 10:23-EST from CLYNN@G.BBN.COM Charlie, And I thought I was the only one making mischief in underutilized header bandwidth. Fuzzballs happen to stamp the low-order 16 bits of the local millisecond clock in the IP sequence field (generating more than one datagram per millisecond is unlikely in these fossils). As you know, fuzzies usually keep rather good time, so this might be useful in determining one-way delays. Also, fuzzies stamp the size of the current retransmit queue in the urgent field (when that field is not being used for the purposes intended). This was intended for experiments in adaptive receiver strategies in much the same way your abuse of the urgent field was. Someday when we both are grey we oughta try some of those experiments. Dave -------