Matthias Welwarsky wrote in a message to Mike Bilow:
> This is true, but AX.25 doesn't really have a channel access
> algorithm. Most of the textbook models do not take into account
> the hidden transmitter problem, and we have never really
> developed an effective method of dealing with that. What we run
> in practice tends to reduce to pure Aloha. Seen in that light,
> the importance of T2 is more sensible.
MW> Don't mix things up please. T2 has _nothing_ to do with
MW> channel access. It's solely for delayed ACKs and gives the
MW> chance to merge multiple ACKs into one or even into an
MW> implicit ACK via an I-Frame. You won't be able to cope with
MW> hidden stations anyway. You can achieve the same effect with
MW> a proper slottime/persistence setting, without rapeing an
MW> AX.25 protocol element for MAC purposes ;-) Aloha generally
MW> means: No throughput when many stations.
No, I think you misunderstand me. I realize that this was not the purpose of
T2 as designed, and I thought I pretty much said that in my comments which you
quoted above, but my point is that the practical result of T2 is to improve
channel yield. I am not asserting that T2 could improve channel capacity,
which would be obviously absurd. By "yield," I mean here the proportion of
attempted frames which succeed, primarily due to the absence of collision. I
also fully appreciate that channel yield, in this sense, is not something that
one necessarily desires to maximize, since it can be traded off against channel
capacity. In a way, maximizing channel yield is analogous to bragging about
how effective your dummy load is because it has a great SWR.
> This is, of course, completely true of a point-to-point link,
> or at least of any link generally where there are no hidden
> transmitters. However, as soon as hidden transmitters are
> introduced, the channel utilization starts to follow a
> classical Poisson distribution, so the introduction of
> strategic delays is actually the only thing that prevents
> degeneration into total chaos.
MW> Correct. Thats what channel access algorithms deal with. But
MW> the mathematical facts wipe all attempts to be clever, as long
MW> as you don't really solve the problem of the hidden stations.
MW> In a contention environment with competing stations and pure
MW> ALOHA you can at best achieve something like 21% of the maximum
MW> possible throughput of a channel. No matter how clever you try
MW> to be with persistence or T2 settings.
I would concede that this is theoretically true, and the 21% number comes from
precisely the Poisson model that I mention. However, it is important to
recognize that this number only applies in the context of measuring channel
capacity. In other words, pure ALOHA falls off to 21% throughput
asymptotically under the specific condition of approaching maximum channel
loading to capacity. This is exactly why one might choose to trade capacity
for yield, as I discuss above, and why maximizing yield might be reasonable if
not carried to extremes. That is, SWR really is important, but it is not a
measure of absolute effectiveness or we would trade antennas for dummy loads.
> I'm not really sure what sort of pattern results from ARQ, but
> I would assume that it tends to look more or less binormal. As
> long as it is something predictable, we should have no trouble
> adapting our rtt measurements.
MW> Hm, it's not fully predictable. No ARQ scheme is. CSMA/CD
MW> isn't either, you can have inifinite delay of delivery. No
MW> prediction scheme can deal with this.
My remark was fairly off-hand. I'm not sure what you get for statistics when
you try to measure something like this empirically, but it would probably be
interesting and I assume people have done it. My comment about a binormal
distribution a was purely intuitive guess based on no hard evidence. If we had
any real data, I suppose curve-fitting would be a reasonable approach. If you
are asserting that no such curve could be found, then I think such an assertion
in the absence of real data is at least as unsupportable as mine. :)
-- Mike
