MW> Yes, but you cannot outrule physics. If you drop the
MW> guarantee of delivery at link level you will have to keep
MW> the propability for a bit error extremely low.
Let me try to be more clear. I can see an argument for guarantee of delivery
at physical layer, but not at link layer. When we use telephone modems, for
example, we have V.42 or MNP4 error correction, but SLIP and PPP do not
implement any sort of retry mechanism. (PPP discards on a bad FCS.)
It is a mistake to bind error control into the link layer protocol, and nowhere
is that more clear to me than in radio. The link layer should not have to know
whether the link is on HF or VHF, narrowband or wideband, noisy or clear, and
so on. When we put error control into AX.25, we made the protocol close to
useless on HF. My point is that any error control mechanism has to be closely
tailored to physical layer issues, and that is where we have fallen down with
packet radio. Furthermore, it is my opinion that the half-baked error control
in AX.25 is largely responsible for arresting the development of intelligent
physical layer error control which would have been far more useful and
effective.
-- Mike, N1BEE