[aprssig] time for APRS second generation network?

[Jason Winningham]

Thanks for the explanation of the 9600 baud system.  Turns out I did 
understand the details, but not the intent.

> If you are talking about Nodes,

Not sure what a Node (capital N is) but when I said node I meant it in 
the common networking sense - a device attached to the network (client, 
tracker, digi, whatever).


> But your example is not a one-to-all.  It is a point-to-point

A single unbridged ethernet LAN (or an 802.11 wLAN) _is_ one to all, at 
layer 1 and the bottom part of layer 2.  I accept there are very few of 
these old ethernets around, but we probably still have some here (:

Ignoring the suitability of my example, one-to-all/broadcast is an 
issue handled different ways at different layers in the network.  At 
the application level this is simply assumed to be broadcast and we 
don't have a real destination address.  At the data link layer (with 
the current system) it's how aliases, callsign substitutions, and so 
forth are handled.  At  the physical layer, it's antenna design and 
placement (and TX power).

>  and has 10 megabyit bandwidth.

or 100.  or 1000.  It's still a CSMA/CD network with layer 2 switching.

> That is on another planet compared to
> APRS one-to-all communications with 1200 baud radios.

No it isn't - if you take a simple IEEE802.3 network (coax or hub, no 
switch) and compare it to a single ARPS digipeater's network, there is 
a very close correspondence.  If you examine a bridged 802.3 network 
there is a great deal of correlation to an area-wide APRS network with 
multiple digis.  There simply is not a great deal of difference between 
our CSMA/ALOHA hybrid and CSMA/CD.  RF is a signal radiating 
spherically is space, ethernet is a signal propagating linearly along 
the coax.  It is basically a difference between 2 dimensions and 3.


On the theory side of the house, the current APRS network consists of 
an application layer, a (heavy) data link layer, and a physical layer.  
This abstract network I'm proposing has the same application layer, a 
new network layer, the same data link layer (used lightly), and the 
same physical layer.

This IN NO WAY ignores "one-to-all".  On the contrary, I'm suggesting 
we _design_for_ one-to-all in the entire network, instead of trying to 
retrofit to an archaic source-routed data link layer and trick it into 
doing layer 3 tasks.

Source routing is fine for packet.  APRS has outgrown this and is ready 
for something better, hence the "second generation" in the subject 
line.  We can attain incremental improvements in performance, and we 
can attain drastic improvements in ease of client configuration and 
abuse suppression, and we can do it without Joe Ham doing anything to 
his d700/d7/opentracker/tinytrack/gps-attached kpc3+/whatever (if we're 
careful).

APRS is not a creature unique in the universe - it is just another 
network.  We can use standard network theory to describe it, we can use 
standard network analysis techniques to examine it, and we can use 
standard network design techniques to build it (or rebuild it).

-Jason
kg4wsv