[aprssig] More digis are changing to the new paradigm
Jason Winningham
jdw at eng.uah.edu
Tue Jul 5 13:48:14 EDT 2005
On Jul 5, 2005, at 8:27 AM, Robert Bruninga wrote:
> This will guarantee several things:
>
> 1) the sender will have no idea how far his packets will go
Why? How do we know how far they'll go _now_? (answer: we don't!)
> 2) the network will always be 100% busy with routine
> lights-are-on-noone-home packets doing nothing
Again, why? If we design it to saturate the channel, sure it will be
that way. If we are just a little bit smart when we design it, there
will be no more traffic than the channel can support.
> 3) The user that needs to get packets from A to B cannot
Why? If we design a network that is smarter and more efficient, it can
be easier and more reliable to get from A to B. For that matter, we
can easily design in backwards (and I do mean backwards) compatibility
for those who insist on using source routing.
> 4) The priority user with a LIVE immediate need cannot
> communicate any better than a home station with non
> moving beacons.
Why?
> And he wont have any clue if the system
> will deliver his packets too where they need to go right-now.
APRS uses AX.25 datagrams (UI frames). Datagrams are unverified,
unacknowledged, best effort unguaranteed delivery _by_ _definition_.
If you want guaranteed delivery, use connected mode AX.25, or TCP, or
some protocol that supports it.
By intelligently managing the network we can increase the successful
delivery rate of these datagrams. We can also allow the user to more
intelligently specify where he wants the packets to go (to an igate, X
number of hops, Y number of miles, state/region/country, etc) for a
better success rate and decreased network load.
> 5) The network becomes completely indeterminant.
The network is _already_ completely indeterminant. If you want
determinism, you need connected mode, etc.
> 6) The network is extremely dependent on sysop-hands
> on-tweaking to do anyting other than "routine". And
> experience shows that getting a SYSOP to make real
> time changes in some areas takes not just hours, but
> months.
Thanks for making my point. This is _exactly_ why the infrastructure
needs to have smart well designed components that can react
appropriately to conditions on a minute-to-minute basis, not a
year-to-year basis.
-Jason
kg4wsv
More information about the aprssig
mailing list