[e2e] TCP un-friendly congestion control

[Mark Allman]

> The original poster asserted that in an environment where the
> network went at 1 Gbps and had 50ms of delay, TCP was hopeless.
> 
> The point I was trying to drive home is that it is not hopeless.

But, it is "practically hopeless", I think.

I think it boils down to TCP performance being fundementally
governed by two different limits: 

    (1) a protocol limit (based on the RTT and the maximum
        advertised window size), and 

    (2) an algorithmic limit (based on the packet size, the RTT and
        the total loss rate)

(And, in practice, governed by the minimum of (1) and (2).)

I think what you (Craig) are trying to say is that the environment
described above can be handled if you look at limit (1) (and, in
fact, that is all we really have enough information to look at).
And, I think the impicit assumption of the original post was that
the loss rate (congestion, corruption, host error, router error,
whatever) will pretty much doom TCP to less than optimal performance
in nearly every practical situation.  I buy both of those points of
view.

allman


--
Mark Allman -- BBN/NASA GRC -- http://roland.grc.nasa.gov/~mallman/