APNIC IPv6 transit exchange

Bernhard Schmidt berni at birkenwald.de
Tue Dec 4 16:44:19 CET 2007

bmanning at vacation.karoshi.com wrote:

>> Germany-Italy-US-APNIC-IX (Japan?)-Australia-APNIC(again?)-Japan-US-Germany
> 	I can't help but think that you (and others that complain)
> 	have an odd understanding about "broken"...  For me (and
> 	I suspect many others), broken means the packets don't get
> 	there.  Connectivity, even if not what I would consider 
> 	optimal paths, is better than no connectivity at all.

No, not if you are trying to get ordinary users to "just use" IPv6. If 
IPv6 is so much worse regarding latency and throughput (let's just call 
it "user experience") users will complain and eventually just disable 
IPv6. Seen for so many cases, for example with broken recursers in SOHO 
routers that timeout on AAAA records, instead of fixing them the 
solution in almost every Linux distribution wiki/forum is to disable IPv6.
> 	and now the two points of interest:
> 	) mapping an ASN to a specific economy is often a false reading.
> 	  for a higher confidence metric, it would be useful to see the
> 	  latency btwn each of the these hops in the AS path.

Can't really help you with that. I did (before Terry fixed the filters) 
a trace from uschi01.sixxs.net (within OCCAID network) to the prefix in 
question and saw my mapping confirmed. 450ms RTT and 20% loss to the 
destination (~100ms would be expected). All on the forward path of 
course, return path was as one would have expected. Now add the 120ms 
25484 41692 30071 is going to take and we are in the 600ms region.

@Terry: Thanks for fixing this, but by saying you fixed the export 
filter for 1221->38610 you basically tell me that v6TE is not operating 
in the state you want it to be and the only cause why it is not leaking 
such routes at the moment is because it doesn't receive a fulltable. 


More information about the ipv6-ops mailing list