[aprssig] Why RELAY,WIDE... is so bad....

[Robert Bruninga]

>>> mwrobertson at comcast.net 3/31/05 12:24:21 PM >>>
>Please explain how RELAY causes DUPES?  

Thanks for asking, because it is important that eveyone 
understand this and it is why RELAY in our current overloaded
nets is now part of the problem and not the solution...
Lets assume a mobile is using the path of RELAY,WIDE:

If all he hits is a rare RELAY-only digi in the boonies, then
there is one copy digipeated and then that copy is then
digipeated by lets say 3 other WIDES.  Those are all 4 copies
of the original packet, one for each digipeater.  We dont call
those dupes.  Those are what we want.  4 copies for 4 digis.

MOBILE>APRS,RELAY*,WIDE
MOBILE>APRS,RELAY,DIGI1*
MOBILE>APRS,RELAY,DIGI2*
MOBILE>APRS,RELAY,DIGI3*

But all the other 95% of the time while the mobile is driving
around town his RELAY,WIDE packet is being heard by
lets say 3 other digis.  Now then here is what we see:

MOBILE>APRS,DIGI1*,WIDE
MOBILE>APRS,DIGI2*,WIDE
MOBILE>APRS,DIGI3*,WIDE
MOBILE>APRS,DIGI1,DIGI2*
MOBILE>APRS,DIGI1,DIGI3*
MOBILE>APRS,DIGI2,DIGI1*
MOBILE>APRS,DIGI2,DIGI3*
MOBILE>APRS,DIGI3,DIGI1*
MOBILE>APRS,DIGI3,DIGI2*

for a total of 9 packets.  Three are the intended
packets from the 3 surrounding digis and the other
SIX are "DUPES".  Each digipeater is digipeating
the same packet TWICE.  Now imagine each time
that mobile passes by a HOME RELAY, that adds
another original for each of the digis to also digi
a 3rd time. 

Not only is this so obvious to see happening and
so obviously a waste of bandwidth by causing
a factor of 3 dupes on every packet trasnmitted,
but the stations doing it the most are MOBILES
which already generate 10 to 20 times as many
positions per hour as any other stations.

It is this totally unneeded multiplication of DUPES
from R,W paths that is the biggest load on the
APRS chnnel and also the *easiest* to get rid of.

Some people say that they have to hve a RELAY in
a certain area because it is a dead zone.  True, but
the "cost" of having mobiles use the R,W path just
for that one packet in the morning and that one packet
in the evening in that "hole" is causing the other
40 of his commute packets to be generating over
120 "DUPES" all the rest of the time when he is
closer to town on his 20 minute commute.

To me, we are shooting ourselves in the foot with all
these mobiles driving around using a RELAY,WIDE
path.  Yes, I advocated that for years for backwards
compatiblity with all the RELAY-only and WIDE only
TNC's we had back here on the east coast.  But
it is time to move beyond that.  In fact, we should
have done it years ago.  

But I have to admit that since I have been spending
12 hours a day for the last 4 years buildign satellites,
I had not been watching what was going on on the
local 144.39.  A few months ago, I had a weekend
lull and was wathcing RAW packets and was 
*astonished* at what I saw from all the dupes of all
the RELAY, and WIDE paths.

By the way, WIDE,WIDE does exactly the same thing.
if there are 3 digis that hear that path, then there
are the potential for 9 copies just from the first 3 digis.
We MUST phase out WIDE completely and discourage
any routine use of RELAY.

On the other hand the path of WIDE2-2 goes just as
far, but because the WIDEn-N algorithm has perfect
dupe elimination, then if there are 3 digis, then there
are only 3 copies (no dupes) heard in the original 
area of packet origination.  Copies only radiate outward
and never fold back.

Where as with RELAY,WIDE and WIDE,WIDE paths
there will always be dupe multiplicatiion.

But you ask, what about callsign-substitution? Isnt
that supposed to eliminate dupes and keep a digi
from digipeating something it has digipeated before?
Yes, but look back at the original 6 dupes in the 
original example above.  Notice that none of the
6 dupes was ever transmitted by the same digi
twice.  Those would look like MOBILE>APRS,DIGI1,DIGI1*
but this only begins to take effect at 3 hops ane
beyond.  It only begins to work to prevent this:

MOBILE>APRS,DIGI1,DIGI2,DIGI1*

This packet would NOT occur because of callsign
substitution.   But in the meant time there could be 
as many as 15 copies heard from these 3 digis when
only 3 would be heard if WIDEn-N were used.

MOBILE>APRS,DIGI1*,WIDE,WIDE
MOBILE>APRS,DIGI2*,WIDE,WIDE
MOBILE>APRS,DIGI3*,WIDE,WIDE
MOBILE>APRS,DIGI1,DIGI2*,WIDE
MOBILE>APRS,DIGI1,DIGI3*,WIDE
MOBILE>APRS,DIGI2,DIGI1*,WIDE
MOBILE>APRS,DIGI2,DIGI3*,WIDE
MOBILE>APRS,DIGI3,DIGI1*,WIDE
MOBILE>APRS,DIGI3,DIGI2*,WIDE
MOBILE>APRS,DIGI1,DIGI2,DIGI3*
MOBILE>APRS,DIGI1,DIGI3,DIGI2*
MOBILE>APRS,DIGI2,DIGI1,DIGI3*
MOBILE>APRS,DIGI2,DIGI3,DIGI1*
MOBILE>APRS,DIGI3,DIGI1,DIGI2*
MOBILE>APRS,DIGI3,DIGI2,DIGI1*

That is why I say that any use of RELAY,WIDE... paths
is causing 2 to 5 times the duplication of every packet
transmitted.  These are unnnecessary DUPES and
not just COPIES I am talking about.  They are just
wasted QRM...

>So, if home RELAY fill ins are spaced properly, what is 
>the problem?   They are doing what they were put there 
>to do...

So true.   But is the other 95% of the time that that mobile
is transmitting RELAY,WIDE and causing horrendous
duplication all the rest of the time when he *is* in range
of digis that is the problem...

>I fail to understand your statement that home RELAYs 
>cause so much duplication.. 

I'm sorry.  If I said that, what I was meaning was that the
"concept of home RELAYS which is the basis for mobiles
using the RELAY,WIDE... path is causing horrendous
duplication most of the time." is the better way to say it.

>So in short, a DUPE to me is a packet sent by a RELAY 
>that was received from another RELAY. It is not a DUPE 
>if two RELAYs attempt to digipeat the same received 
>direct from the tracker's packet.

Yes.  A DUPE is the same digi digipeating the same
packet more than once...

>Simple solution is to limit unnecessary home RELAYs.

Best solution is to eliminate the use of the RELAY,WIDE 
and WIDE,WIDE paths completely.  This will cause a
huge 2 to 5 TIMES reduction in dupes.  This means
the network load could go down to 50% or even 20%
of what it is now.  With that kind of a quiet channel
there probaby is no need for the fill-in digi in the
first place, because now the regular digis might hear
enough dead silence to hear the mobile in what used
to be called the dead zone in the first place.

And if the dead zone is really dead, then put a WIDE2-2
digi there to replace the RELAY.  You still get the digipeat
but not the QRM all the rest of the time when the mobile
is not there....

Hope that clears it up.

Bob, Wb4APR