[aprssig] time for APRS second generation network?

[Jason Winningham]

> there is no such thing as "routing" in APRS:

Putting "RELAY,WIDE" in a packet and aliasing digipeaters to respond to 
the names RELAY and WIDE it meets the standard definition of source 
routing.  I agree that source routing isn't "real" routing.

My RF LAN consists of all stations that can hear me and all stations I 
can hear (my definition).  If a digipeater forwards a packet so that a 
station that cannot hear me directly receives that packet then the 
digipeater has routed my packet between two LANs - mine and the other 
station's.

The problem is that our routers (digipeaters) are only capable of 
source routing.  That's a poor way to route a packet, especially in a 
dynamic (mobile, especially cross country) or unknown (SAR, come as you 
are) environment.


If APRS uses a network, we should be able to apply some of the basic 
networking terminology to it.  I tend to think in terms of the 7 layer 
ISO reference model (even if we don't use all layers).

> 1) APRS is defined as one-to-all information sharing

This statement is either application layer information, or network 
layer addressing information, depending on how you look at it.  It has 
nothing directly to do with the underlying network.  It does influence 
design decisions, but it does not define lower layers of the network 
(OK, it may define the _interface_ between this layer and the next, but 
it does not define the internals of the layers below).

> 2) It is physically impossible to have more than 60 or so
>     users in a single RF domain.

Wouldn't it be nice if the nodes on our network responsible for the 
physical area we see were aware of that fact, and could adapt to it 
dynamically?

> Thus, by definition, there simply is no such thing as "routing"
> on the APRS RF domain

There is no such thing as "real" routing in APRS RF networks now.  
That's my point, and I think that is its biggest weakness.

This is _not_ to say that there _cannot_ be routing in APRS RF 
networks.  If it really is a network, it can be routed.

Using a standard networking protocol layer breakdown, RF is simply the 
physical layer.  There's a lot more network above that.  Routing is two 
layers up, and doesn't necessarily care what's happening at layer 1.

You _can_ route on an RF network, and the routing model I'm talking 
about is roughly analogous to logical subnetting of multiple IP 
networks on a single LAN.  Right now you can't _practically_ route on 
APRS-RF, simply because there are no routers to perform this function.

>  except in non-routine, unusual circumstances.  Otherwise "routing" is 
> simply the APRS-IS.

IGates do the only real routing in the global APRS network, but only 
between RF and IS.  Unfortunately, that's not where we need it the 
most.

A real APRS router would:
	- simplify client configuration
	- stop client abuse of the network with excessive hops
	- statically determine how many hops are allowed in its physical area
	- avoid loops (the ping-pong effect)
	- reduce packet size (no "RELAY,WIDE2-2" to double the size of a Mic-E 
packet)

in addition, it could:
	- dynamically resize it's definition of "local" to match traffic to 
capability
	- only transmit packets that the LAN can handle

-Jason
kg4wsv