Bad IPv6 connectivity or why not to announce more specifics (Was: IPv6 news)

Jeroen Massar jeroen at unfix.org
Thu Oct 13 19:06:30 UTC 2005


On Thu, 2005-10-13 at 14:42 -0400, Valdis.Kletnieks at vt.edu wrote:
> On Thu, 13 Oct 2005 11:05:58 PDT, Peter Lothberg said:
> 
> > Is there anyone who can talk to it using IPv6 on the Nanog list? 
> > 
> > (Time20.Stupi.SE, 2001:0440:1880:1000::0020)
> 
> % ntpdate -q 2001:0440:1880:1000::0020
> server 2001:440:1880:1000::20, stratum 1, offset 0.012038, delay 0.45547
> 13 Oct 14:37:22 ntpdate[30374]: adjust time server 2001:440:1880:1000::20 offset 0.012038 sec
> 
> % traceroute6 2001:0440:1880:1000::0020
> traceroute to 2001:0440:1880:1000::0020 (2001:440:1880:1000::20) from 2001:468:c80:2103:206:5bff:feea:8e4e, 30 hops max, 16 byte packets

Well Valdis, that bad route also has to do with your side of the
equation, you might want to check who you are actually using as transits
and if the routes they are providing to you are sane enough.

Text version:
2001:468::/32 is in the routing table, getting accepted by most ISP's.
This one has a reasonable route, going over GBLX (3549) in most places.
Though some get this over BT (1752), who have 'nice' (ahem) tunneled
connectivity and transit with everybody on the planet. ESNET (293) seem
to be the third 'transit', with OpenTransit (5011) being the fourth one
and ISC being the fifth. Many routes seem to go over VIAGENIE (10566)
who seem to have some connectivity problems too most of the time.
Path wise most of it looks pretty sane.

Then there is a chunk of /40's, which are visible inside Abilene, GRH
does see them, but other ISP's don't. Most people will thus take the /32
towards your IP, which might go over some laggy tunneled networks.
Fortunately not criss cross world yet, but...

then the fun part:
2001:468:e00::/40 though seem to be visible globally, getting announced
by University of California, directly going to: Korea! :)
And then coming back to the rest of the world over Viagenie (10566)

One of those nice paths:
2001:468:e00::/40 16150 (SE) 6667 (FI) 3549 (US) 6939 (US) 6939 (US)
10566 (CA) 3786 (KR) 17832 (KR) 1237 (KR) 17579 (KR) 2153 (US)

Neatly around the world, you might want to hint this University to not
do 'transit' uplinks themselves with Korean networks :)

Then there is also a 2001:468:e9c::/48 which also goes over Korea.

The colored version:
http://www.sixxs.net/tools/grh/lg/?find=2001:468::/32

The above simply happens because most ISP's sanely filter on /32
boundaries, as per:
http://www.space.net/~gert/RIPE/ipv6-filters.html

For further reading see Gert Doering's excellent presentations at:
http://www.space.net/~gert/RIPE/

Greets,
 Jeroen

-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 240 bytes
Desc: This is a digitally signed message part
URL: <http://mailman.nanog.org/pipermail/nanog/attachments/20051013/e1e5a3a8/attachment.sig>


More information about the NANOG mailing list