[aprssig] APRS routing strategies (was: New n-N success in North Carolina)

[Jason Winningham]

On Feb 12, 2005, at 4:20 PM, Wes Johnston wrote:

> I love a good discussion....

yep. (:

> Well, yes, we are kinda saying the same thing.... what you are doing 
> is putting
> a cap on the number of hops a packet can take beyond you.... what I'm
> suggesting is that we pass it along as-is it hasn't hopped too many 
> times
> already....

OK, I see the difference now, and I see some merits for both 
approaches.  I need to draw myself some pictures.  There may be a way 
to use a combined approach, or have it user-configurable for the 
digi/router operator.

I haven't brought it up yet to keep examples simple, but when the 
router is calculating the requested number of hops (or the number of 
hops already taken) it should examine the _entire_ path, not just the 
first unflagged call.  some example pseudo-code:

total_hops_requested=0
for each call in path
	total_hops_requested = total_hops_requested + hops requested by this 
call

... and the hop count limit code from before:

hop_count=total_hops_requested

if total_hops_requested > max_allowed_hops
	hop_count = max_allowed_hops
hop_count--
if hop_count > 0
	transmit packet

This would allow only a sane number of hops for a path like 
WIDE2-2,WIDE2-2,WIDE2-2,WIDE2-2,WIDE2-2,WIDE2-2.  We'd need to decide 
on a mechanism for rewriting such a path; probably a total rewrite is 
in order.



While we're talking about hypothetical new hardware, we can designate a 
new path specification: Hn, the letter H followed by a single digit 
indicating the number of hops requested.  H3 would be equivalent to 
R,W,W, or R,W2-2, R,R,R, etc.  Or, we could stick with WIDEn-n, provide 
for an abbreviation of Wn-n and drop RELAY.

I'm not sure of the consensus on the value of a path trace, but it 
would be simple to make adding the call of the router to the path, 
e.g., H3 is path received by router KG4WSV-7, "KG4WSV-7*,H2" is the 
path as transmitted.  I haven't examined the possibility of adding 
flags, but maybe something like inserting a call of "FLxyz*" (flagged 
as digipeated) as the first call in the path would work.  We could use 
flags for things like

- "add a trace to this packet"
- "digipeat only until I get to an Igate"
- "this packet should stay within the area/state/region/etc"
- "this is a serious situation and I _need_ for this packet to go 7 
hops" (to override hopcount reduction)


There should probably be a "first hop only" router, analogous to the 
current RELAY only digi.  This device would transmit a packet only if 
it had not been transmitted by another digi/router.


There is also the issue of the APRS router acting as a standard AX.25 
digipeater.  We are making assumptions on the path and modifying it to 
suite an ARPS network, and the operations may not be compatible with 
standard AX.25 operation.  Once choice is to designate this device as 
ARPS-only - no other AX.25 operations would be allowed.  Another method 
would be to try and figure out if we're handling an APRS packet or not 
and act accordingly.  This may be possible by examining the destination 
address, and see if it matches one of the generic APRS addresses or an 
ALTNET designated by the digi/router operator.

It seems to me that the APRS network is busy enough without non-APRS 
traffic on the frequency, so it wouldn't necessarily be a bad thing to 
"break" non-APRS packet operation.

-Jason
kg4wsv