[aprssig] Throttleing EchoLink Objects

[Lynn W. Deffenbaugh (Mr)]

Jason KG4WSV wrote:
> On Thu, Jul 23, 2009 at 10:45 AM, Robert Bruninga<bruninga at usna.edu> wrote:
>   
>> can you help me understand why we cannot have
>> regional filtered feeds.
>>     
> The regional filtered feeds provide a subset of the full APRS-IS data
> stream.  In order for data to appear on the regional filtered feed, it
> must be on the global APRS-IS.
>   
I believe what Bob was suggesting is individual, localized feeds going 
directly into a set of regional servers, augmenting what they'd already 
be getting from the global APRS-IS.

The biggest issue I see is that you want SOME of what's injected into 
the regional feed to NOT be fed back to the global feed.  Right now, all 
of the APRS-IS servers automatically send to their upstream host 
EVERYTHING received from their downstream clients.  And the three core 
servers automatically send everything between themselves (unless, of 
course, it's a duplicate packet).

To pull off some sort of regional server, you'd have to have a filter on 
the outgoing end of the server to prevent these locally injected objects 
from going global, but yet have all of the currently acceptable traffic 
continue going global.

And all it would take is ONE mis-configured client to take a feed from 
one region and feed it into another region or central server and BAM, 
everything is global once more.

Regional servers MIGHT be possible if they were somehow centralized 
under one operation, but then we'd lose the self-healing type of 
redundancy that we enjoy with the current ad-hoc network that makes up 
the APRS-IS.  We tend to refer to it as if it was one thing, but it's 
actually more like a coral reef, a bunch of individual organisms 
(servers) each built upon and extending the others to which it is connected.

Lynn (D) - KJ4ERJ - Providing my understanding of one of the potential 
issues

PS.  Pete's "No" was much easier to type!  But was admittedly only 
correct given the current APRS-IS architecture and implementation.