[aprssig] RE: aprssig Digest, Vol 3, Issue 28

[Robert Bruninga]

>>> n9yrx at hotmail.com 9/27/04 2:42:48 PM >>>
>...I am having to use an ever increasing WideX-X to get 
>a message to check-ins... [becasuse]... there are more 
>digis that "burn off" a Wide than... a year ago.  I am now 
>up to  Wide7-7 to get a message reliabily 125 miles away in 
>5 re-tries.

Wow, thanks for the example. ALL KINDS of things come
to mind in this example:  Let me preach to the choir if I may:

1)  APRS is for local tactical real-time operations, not
statewide nets.

2) The 1200 baud APRS channel can not handle reliably
any more stations than about 50 to 100 in any given 
area.  Attempts to do so violate the laws of space and time.
The "size" has little to do with geography, only the size
of the area that contains the 50 to 100 users in your area.  
In LA it might be only 15 miles.  In Wyoming it might be 
150 miles.

3) No net on APRS or any packet system using dumb digis
should expect to operate reliably over more than one or 
at most 2 hops or so.

4) Yes too many digis make too many QRM packets
but they dont "burn off" a hop that would otherwise
cover the distance between digi A and C even if
B uses a hop in the middle. 

5) Adding MORE hops shoots oneself in the foot.
You are making more QRM rather than reducing it.

6) If the probability of your msg packet getting through
each digi, is X, then the chance of getting it out N
hops is X to the Nth power.  This means that the
probability of getting back an ACK is X to the 2*Nth
power.  Or terrible!  

7) Do NOT expect ACKS beyond 1 or 2 hops!
If X is your chance of success through one digi is say .5, 
then your chance of getting out 3 hops is  about 12%.  
But your chance of getting an ACK is about 1% !!!!!!!
Do not expect ACKS beyond 1 or 2 hops!  And by all 
means then dont flood the network trying to get them...

8) Only APRSdos, Xastir, and APRS+SA fully implemented
all of the original APRSdos ACK-IMPROVEMENT 
features to help in this (well proven unbalanced) situation.  
Most other programs never implemented them.  Thus most 
of these programs are very poor at sending ACKS.  Here 
is what APRSdos (and APRS+ and Xastir) do that most do 
not:

a) APRSdos sends new messages on a decaying algorithm
   so that new message lines are delivered quickly and
   repeatedly but then decay out of the system.
b) APRSdos schedules a 30 second delayed duplicate ACK
   that doubles the probability of getting an ACK without
   quadrupeling the number of message copies having to 
   be sent!
c) APRSdos then further increases the number of redundant
    acks by the number of HOPS on the incoming message.
    Again, doing far less damage to the network then 
   continuing to send the original LINE over and over
   to too many hops!
d) APRSdos embeds ACKS in outgoing messages.  Thus
    in a dialog (like your net) the chance of an APRSdos
   ACK is only X to the Nth instead of X to the 2*Nth.  This
   difference is the diffeence between the 12% and the 1%
   in the original example.

So what you are seeing is over expectations in using a
1200 baud channel beyond your ALOHA range and
in some extent, the general degradation in the
APRS network as more and more people use software
that has less and less of the original algorithms in them.

My recommendation to you is to simply plot the PHG 
circles of the digis in your area so you can see how 
many hops it takes to cover the intended area.  Then
use only WIDEn-N to cover those N hops.  If your
number of N is greater than 2, then you expect too much
from APRS.

Also consider using an APRS program that 
1) Easily shows you the RANGE circles of
    digipeaters so you know how the network
    works in your area and
2) has fully implemented the delayed ack,
    redundant ack, and REPLY-ACK algorithms.

de WB4APR, Bob