op 06-11-13 03:01, Philip Homburg schreef:

On 2013/11/06 8:27 , Tim Chown wrote:

Just to rewind a bit, the concern over support on-path for IPv6 Extension Headers, esp. the Fragmenttation Header, has always been there, it’s just that with the recent measurements by Fernando and an MSc student here, the numbers are rather worse than many people expected, moreso the longer the EH used. Fernando did the test extending his own IPv6 toolset, our student used scapy (which supports the four main EHs) to craft HTTP requests.  We both targeted the top Alexa sites.  The Fragmentation Header success rate across the sites tested was at best under 70%.

A couple of remarks:

NLNet Labs used RIPE Atlas to measure fragmentation issues in the
context of DNS. Though the error rate is high (around 10%) it is nowhere
near 30%.

Things seems to get better over time.
Our most recent measurements (Oct 2013) showed that 6.9% of all 1059 IPv6 probes were not able to receive fragments.
See slide 21 of https://ripe67.ripe.net/presentations/230-pmtud4dns.pdf

We measured it by querying for a larger DNS record that will be answered fragmented at 1280 MTU. This is a more realistic setting than fragmented TCP which shouldn't occur in the wild, right?

We measured the ability of probes (client side of DNS) to receive fragments over UDP.
With Emile (10% IPv6 frag filters, https://labs.ripe.net/Members/emileaben/ripe-atlas-packet-size-matters ), the ability of the probe to sent out and receive fragmented ICMP messages was measured .
The direction of the traffic and protocols used seems to matter...

Currently we have a student working on a framework that can track these network properties over time. It is our intention to make the framework open source so others can review and redo our measurements exactly as we did them ourselves, and can build upon it (and contribute back :).