[aprssig] Someone's Server Mangling Packets

[Gregg Wonderly]

On 1/25/2012 8:16 AM, Andre wrote:
> Op 25-1-2012 14:44, Bob Bruninga schreef:
>> Concatenated packets:
>>
>> I have seen these rarely since the beginning of APRS.  I assumed they were
>> my own DOS processor taking a few milliseconds break and as a result
>> concatenating packets.  Since the TNC would not make such a mistake due to
>> the CRC check, I assumed it was in the  APRS client.
>>
>> However, if we realize that almost every packet heard these days is a
>> concatenation of one weaker packet overlayed with the most recent strongest
>> packet, and since the ending flag is the same as a beginning flag, there is
>> a finite probability that the sum of a few bit errors could result in a
>> concatenated packet with a valid checksum at the end...
>>
>> I think I theorized this once before, and maybe someone shot it down, but
>> I'd like to see the probabiltites again...
>>
>> Bob, Wb4APR
>>
> CRC is 16 bit so the probabilety if a mangled packet having the correct CRC is 
> one in 2^16 or 65536.
> I think it is far more likely that a TNC/PC link in converse mode drops a 
> lf/cr or in simple kiss mode drops 2 FEND bytes in a row.
> Also telnet links can drop characters like a TNC in converse mode.
It is not possible for "telnet" to drop characters, because it is a TCP 
service.  What is possible however, is that igate software, can be doing 
automatic connection "management", and when the link disconnects, it can either 
"use" a partial packet as a complete packet because it didn't check to make sure 
that the CR/LF was on the end, or it might leave those characters in a buffer 
because there was no CR/LF at the end to trigger processing, and then it 
continues receiving more characters when the connection is set back up, and thus 
appends the next complete packet onto the end of the previously received characters.

It's that kind of behavior, no mater whether it's a network connection, or an 
RF/serial interface into the software, that creates these kinds of problems.

Gregg Wonderly