Re: Extended sequence numbers for AX.25

[rob@sys3.pe1chl.ampr.org (Rob Janssen)]

According to Phil Karn:
> >> Your network tests are very interesting. You should publish them.  I
> >> note that the usual rule of thumb for pure ARQ protocols (e.g., TCP)
> >> is that the packet loss rate should not exceed 1% for good
> >> performance.
>
> >With TCP (which has no feedback about lost packets and always assumes
> >lost packets indicate congestion, not loss by transmission error) this
> >is very true.  That is a reason why most of our network uses VC mode for
> >IP traffic.  (actually, the current software automatically switches between
>
> The 1% rule is not just because of TCP's backoff algorithms. Any ARQ
> protocol (TCP or LAPB) wastes valuable RF energy and channel time
> whenever it discards a frame. A channel that loses half the packets
> sent over it essentially wastes 3dB of transmitter power.

But how do you get at the 1% figure that way?  That would be a 0.04 dB
loss only, very insignificant.  Based on wasted energy alone I would not
think a 10% loss (which I routinely see) is all that bad.
What makes it troublesome is the timeout waiting for retransmission, IMHO.
That is usually more of a problem for TCP than for AX.25, especially as
TCP does the retransmission end-to-end, and AX.25 is normally dooing it on
a hop-by-hop basis.  (assuming some reasonable network, not a chain of
digipeaters)

This is why I switch the individual hops dynamically between VC and UI
mode, depending on link quality.  For near-perfect links UI is best, but
when there is some packet loss VC is normally better as it puts the
retransmissions right where the loss is, not at the endpoints of the
connection.

Rob

--
+------------------------------------+--------------------------------------+
| Rob Janssen         rob@knoware.nl | AMPRnet:   rob@pe1chl.ampr.org       |
| e-mail: pe1chl@wab-tis.rabobank.nl | AX.25 BBS: PE1CHL@PI8UTR.#UTR.NLD.EU |
+------------------------------------+--------------------------------------+