[aprssig] time for APRS second generation network?

[Jason Winningham]

Reading (and learning) about ways to fix the problems with the APRS 
network (which needs various amounts of fixing in different areas) 
reminds me a great deal of growing pains seen in some LANs that I've 
witnessed.  One particular ethernet network that had way too many nodes 
to be a LAN had all sorts of tricks applied to keep it going:  bridges, 
switches, twiddling ARP cache timeouts on individual nodes, and so on.  
All these things did bring _incremental_ improvement, but they did not 
solve the basic problem: the network was too complicated for a single 
simple LAN.  The solution was to _route_ packets, not repeat them.  
This took the network to the next level.  Dumb infrastructure 
components were replaced with smart ones, less was asked of the client 
in the way of configuration, and network performance and reliability 
improved by an order of magnitude.

It appears the same sort of thing is in order for APRS.  Digipeaters, 
be they dumb or smart, are not up to the task of routing packets.  
Asking end users (clients) to source-route their own packets has 
brought us the WIDE7-7 problem.

In order to take the APRS network to the next level of performance and 
reliability (and forget about 9600 baud) digipeaters will have to be 
replaced with routers.  Let the routers examine packets, network load, 
and other factors to determine how many times to re-transmit a packet.  
Let the user supply a requested hop-count, or nothing at all! (how's 
that for client configuration simplicity), but the network-aware router 
will have the final decision.  This can be accomplished hardware-wise 
with a microcontroller based module attached to current digipeaters.  
Such a module coupled with a simple KISS TNC should cost less than a 
TNC capable of "smart digi" operation, but would be capable of so much 
more.  For that matter, such hardware could be easily configured to be 
a router, a mobile tracker, a weather station, a flight computer for a 
balloon experiment, etc more easily and more capably than the 
ubiquitous KPC-3+ is today.

I'm not familiar with the innermost details of AX.25 and how APRS uses 
it, but I am confident that not only can a modern router be designed 
and manage the APRS network more efficiently, but it can be made 
backward-compatible with all existing clients the way they operate 
today.

We can easily expect such a device to be aware not only of hop counts 
but of other important factors like previous packet payloads (for dupe 
suppression), the router's own aloha circle, physical distance 
associated with the packet, and other things.

We're asking too much of users and of layer 2 (data link) digipeaters.  
It's time for a layer 3 router in the APRS network.

-Jason
kg4wsv