Fixing MTU issues - would somebody like a 'tracepath6 from 47 different ISPs'? (Was: Test your connectivity for World IPv6 Day)

Bill Owens owens at
Tue Jun 7 14:33:40 CEST 2011

On Tue, Jun 07, 2011 at 11:46:53AM +0200, Jeroen Massar wrote:
> First an insult: Folks who think that blocking ICMP is 'smart' and
> 'secure' please upgrade your brain and fix your local issue and at least
> don't block ICMP Packet Too Big, it really is needed for a proper
> functioning of the Internet.

+1 (+1000 actually)

Small but important correction:
> Scamper is afaik only BSD.

Scamper builds just fine on Linux, BSD, MacOS X, Solaris and something called DragonFly (never heard of it, myself ;)

The TBIT PMTUD-probing code only works on FreeBSD and OSX because it needs IPFW:

> Both can tell you when a node does not respond to Packet Too Big.
> When that happens, you pretty much know where the blackhole is.

It turns out that there are quite a few ways for PMTUD to fail. Matthew has been researching them for several years and has written a couple of papers on the subject, which make great reading if you're really interested in geeking out about this stuff ;)

> This shows that a traceroute works, if enough people want though, then
> yell here on the list with a "+1" and I'll work on it tonight that I get
> a tracepath6 exposed from the PoPs that can offer it, that or option B
> would be that I set up a tunnel from each of the PoPs to one host, and
> then set an MTU of 1280 on those tunnels and allow people to get a
> result of a tracepath6 from that, that will show you if you are
> reachable from behind 47 locations in the world..... would anybody see a
> need for that?

That sounds like it's worth a +1. I'll make another much less scalable offer - if someone has an IPv6 destination (or IPv4 for that matter) that they'd like to check the path to, I have a 9000-byte capable host that has scamper installed, and upon email request I will test the path and shoot the results back to you. Sadly I have no web programming skillz so I won't be able to provide a nice interface to place requests ;)


More information about the ipv6-ops mailing list