[aprssig] Re: Tracker Smart Pathing: user types, alternatives

[apratt at bestbits.org]

On Thu, 23 Mar 2006 08:42:05 -0500, "Robert Bruninga" wrote:
> a) But is your wife really looking at FINDU 
> every single minute ...

Of course not. The reason to want a fast rate in the boonies is not a
desire for fast updates, but because so few of the packets get through
at all. In a mountainous remote area, two factors come together:
unreliable access to any digi, and lots of hops to an IGate. The point
is not that I'm far from HOME, but that I'm far from the local DIGI:
between distance and terrain, my packets do not have a good chance of
being heard at all. If I am 3 hops from an IGate, Proportional Pathing
gives me an effective beacon rate of eight minutes. If my digi hit
success rate poor - say, 25% - I'm down to 32 minutes. Remember: any
hit with less than three hops is the same as no hit at all.

Once again, we talk about different users. In my application, when I'm
in a high-success one-hop region, I absolutely do not need one-minute
beaconing, or even three. I would accept ten minutes or more. Also, by
the way, any DIRECT packets in Proportional Pathing are almost
certainly wasted for me - how often is an IGate reachable simplex?
With the proposed Proportional Pathing system, I'm kind of stuck with
one-minute overkill. If I dial the base rate that back to two minutes,
I'd probably end up with an effective rate of 64 minutes in the wilds.

In another message, Bob wrote, "You've clearly never tried to
rendezvous with somebody who's doing 5 minute beaconing." I say I
never plan to try to rendezvous at all! That's what I keep saying
about different users.

> b) If your goal -is- FINDU, most packets get into FINDU
> via one hop or maybe 2.

Not in the areas where I want to be tracked: rural California, in the
mountains and deserts. That is, the big empty areas on that
ten-year-old map on the New n-N Paradigm page. Your own advice on that
same page is to run WIDE1-1,WIDE3-3 or more in areas like that. 

> So this means you prefer to run 1 minute updates from  
> your mobile with 3 hops all the time in California?

Yes to three hops, but probably three-minute updates. You're making my
argument for me: this is a bad path in well-connected areas. The whole
point of this auto-hop-count idea is to get a three-hop route when I
need it, and a one-hop route when that's enough. Today, I have to pick
a single route that will work acceptably well in the WORST areas I
pass through; that path is rude and abusive in the BEST areas. Q.E.D.
(I think I've already described why proportional pathing is not
acceptable for me in those "worst" areas.)

I asked, "What hop count should such a digi write in?" Bob answered,
"None! Only the sender knows what he needs that day." We're talking
past each other: I've already said that a digi will not rewrite a path
unless it's given permission (with "AUTO" in the path), so the user
who "knows" can dictate the path as always. But what I "know" as a
sender (for my user type) is that the network can make a better choice
than I can, as I move from one digi footprint area to another.

Bob also wrote (paraphrasing): "A mobile operator can fix his path
instantly if he finds the web-based advice isn't working for him." Not
me - different users again. While I'm on a trip I can't tell whether
my packets are getting out or not, whether they're failing to hit
digis or don't have enough hops to reach an IGate. Any news I get to
that effect (e.g. at the end of the day) will be so old as to be
useless - by then I'll be out of the relevant area.

A digi that writes "one hop" into the AUTO path when the local IGate
is down will indeed fail me. But I'd have failed in that area anyway
if I manually ran WIDE1-1, thinking that one-hop IGate was still
alive, so I'm no worse off. A local user can manually recover by
setting WIDE2-2 and removing AUTO until things get worked out. As a
mobile traveler trusting digis for my hop counts, I know there are no
guarantees. The digi operator should program the hop count to one or
more "reliable" IGates, but we all know it's a crap shoot. Is that
really a deal-breaker for a system that can improve capacity in urban
areas (fewer unnecessary 2- or 3-hop packets) while improving access
in remote areas?

I don't want to get this mixed up with ideas that involve intelligent,
adaptive digi configuration based on conditions. That would be great
but it's asking a lot, and I see this proposal as an achievable
compromise.

This idea doesn't require "replacing every TNC digi in the country." I
haven't heard yet that it's impossible with current digi hardware, so
it's a matter of configuration. I am not proposing "intelligent digis"
that auto-configure based on conditions - they're just not that smart
today. A mobile user with a path like WIDE2-2,AUTO will deliver the
capacity advantage of one hop in an urban area that's been upgraded,
and the IGate-reaching advantage of three hops in a rural area that's
been upgraded, while getting the usual two hops in un-upgraded areas.

-- Allan Pratt, apratt at bestbits.org