>But, doesn't the issue of the TCP RTT calculation need to be taken into
>account? Are you assuming that the TCP stack is using an appropriate RTT
>calculation for the environment that it is in? If so, I agree with your
>statement but it seems to exclude the wireless link being between two
>intermediate nodes (where the TCP stacks may not even be aware that a wireless
>link is involved).
What's wrong with the standard TCP RTT estimation algorithm,
especially with the timestamp option?
TCP is supposed to be a generic transport protocol that's deliberately
unaware of the details of the subnets carrying its packets. That's a
feature, not a bug.
That's not to say that we can't improve the mechanisms already in TCP
to make them work better or over a wider range of generic path
characteristics, but I just don't think TCP is the place to be making
a lot of ad-hoc changes to deal with specific link types.
Phil
This archive was generated by hypermail 2b29 : Mon Jan 28 2002 - 09:12:20 EST