RIPE Meeting: 60
Working Group: Test Traffic
Status: Final
Revision Number: 1

Test Traffic Working Group
RIPE 60, Prague
Thursday, 6 May 2010, 09:00-10:30
Scribe: Rene Wilhelm, RIPE NCC
Jabber: Robert Kisteleki, RIPE NCC

A. Administrative Matters

Co-chair Ian Meikle welcomed attendees and opened the last session of the Test Traffic Working Group.

B. Re-chartering of Working Group - Ian Meikle, Working Group co-Chair

Following the discussion at RIPE59 in Lisbon, Ian sent the proposed new charter for the working group to tt-wg mailing list. The revised charter widens the scope of the working group such that it can accommodate all discussions on Internet measurements and analysis, not only those related to the RIPE NCC TTM project. Only positive comments were received, no objections raised. Ian asked the working group to formally accept the new charter and re-launch as the Measurement, Analysis and Tools Working Group (mat-wg).

As part of the rejuvenation, Henk Uijterwaal decided to step down as co-chair and make room for new blood in running the group. Henk served as co-chair for the Test Traffic working group for many years and was thanked for this by the audience.

Two persons already had expressed their interest in a co-chair position: Richard Barnes from BBN and Christian Kaufmann from Akamai. In the absence of any objections, Ian proposed to appoint Richard and Christian by acclamation.

The working group adopted the new charter and welcomed the new co-chairs.

C. Measuring IPv6 at Web Clients and Caching Resolvers - Emile Aben, RIPE NCC

Emile presented results of his work on measuring IPv6 capabilities of clients visiting and the caching DNS name servers they use.


Geoff Huston (APNIC) reported he has been doing similar work. He was surprised by the diversity of results. Theory had it that when given the option, dual stack clients would connect via IPv6. But this is not the case, most prefer IPv4. The small amount of Teredo is surprising too. The real surprising outcome is the relative amount of IPv6 capable clients. 5% is much higher than we expected.

An audience speaker remarked the data set is biased. Geoff acknowledged visitors of RIPE NCC and APNIC websites may not form a representative sample of the general public; however, the sites do attract thousands of unique visitors, the outcome is still a surprise.

Emile stated we could have better, more representative numbers if the code is deployed on websites, which are targeting the average Internet user. He asked for people who can help in that area to approach him.

D. Measurement Network Update - Erik Romijn, RIPE NCC

Erik gave an update on developments in the TTM measurement network and other projects run by RIPE NCC Information Services department.


Ian Meikle asked why the graph which shows the number of active TTM boxes over time has such a drop between 2004 and 2006

Mark Dranse clarified the TTM network is a community run thing; people subscribe to the service and host a box. However, sometimes the hosting party looses interest; that makes boxes go away.

E. Discussion of TTM Improvements - Rubens Kühl, NIC.BR

Rubens talked about some improvement opportunities see for TTM. Specifically, they are interested in IPv4 path mtu measurements, survival of power failures and more affordable, more flexible GPS antenna solution. For the latter they've built their own GPS based on Garmin 18x LVC unit. Rubens also enquired if, with the death of tt23 in Ann Arbor, more old TTM boxes are reaching End-of-Life.

Mark Dranse thanked Ruben for the input, and stated the RIPE NCC Information Services group would take the suggestions on board. The IS team are about to embark on a fairly large-scale exercise to revisit the software and look at what improvements can be made.

As for older boxes going off-line, Mark reminds the audience RIPE NCC do not manage, own or run the vast majority of the boxes. From time to time, people do lose interest, go away. RIPE NCC does try and encourage fresh deployments to keep the network alive.

F. Counting MAC in IPv6 - George Michaelson, APNIC (15 min)

George extracted statistics on Vendors from DNS client IPv6 addresses and the PTR queries they made at servers for ranges. Intel is most prominent when looking at source addresses. The addresses returned by PTR queries, however, are dominated by Apple. Are they really popular with folk who run IPv6?

George warns not to draw any hard conclusions from this. It's not who you buy your MAC address from; it's how easy it is to use in IPv6 which counts.

Daniel Karrenberg remarked that the Intel numbers are likely overestimated because they include virtualisation. Parallels on the Mac, for example, will emulate an Intel network interface.

G. DNS Visualization Tool - Casey Deccio, Sandia National Labs

Casey presented DNSviz, a tool that visualizes DNSSEC and can help with troubleshooting.

Richard Barnes asked if Casey is collecting any statistics on where he sees DNSSEC and where not. Casey confirmed he has statistics, but the data are far from complete. The tool has only been up for a month. Most users who submitted data have been looking at unsigned zones.

Daniel Karrenberg inquired if there were any plans to package the software once it is stable, such that people could run it on other websites. Casey acknowledged that would be a desirable and viable option for the future, but some work still needs to be done in terms of packaging and licensing; the latter can be a lengthy process.


There was no other business. Ian thanked the speakers for their contributions and closed the session.