> And, wearing my official IETF hat, let me state unambiguously: there's no
> IETF document that TCP/IP is limited to paths with link latencies under
> 125 msec; speaking plainly, that's absurd; and at a recent research meeting
> I was at, we debated whether TCP can work with RTTs of several minutes.
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
On a lighter note:
In the past, we've emulated stuff up to (and including) Mars like delays
in the lab with TCP; TCP "works" (after some timer default diddling) but
you really don't want to be closing TCP's congestion control loop much
past a lunar distance. Everything works, but performance/utilization is
too low. An active open to Mars (and then slow-start) was quite an
experience. I won't even mention what happened when a segment was lost...
:o)
TCP works at arbitrary delays - it is quite robust. Effective bandwidth
utilization as the RTT increases is another story.
Regards,
Eric
This archive was generated by hypermail 2b29 : Mon Feb 14 2000 - 16:14:43 EST