RIPE 75 MAT Working Group Minutes

24 October 2017, 9:00-10:30
WG co-Chairs: Christian Kauffman, Nina Bargisen, Brian Trammell
Scribe: Alun Davies

A. Introduction

The presentation is available at: 
https://ripe75.ripe.net/archives/video/114/

Christian Kaufmann (MAT WG Chair) opened the WG session and welcomed participants and introduced Brian Trammell, incoming WG Chair. Christian asked if there was anyone against approving the minutes from the previous RIPE Meeting. With no objections, the minutes were declared approved.

B. Introduction of Brian Trammell [5 min]

The presentation is available at:
https://ripe75.ripe.net/archives/video/115/

Brian Trammell introduced himself as new MAT WG Chair.

C. Comparing a Set of Latency Measurements

Agustin Formoso (LACNIC)

The presentation is available at:
https://ripe75.ripe.net/archives/video/117/

Kaveh Ranjibar (RIPE NCC) raised some points regarding collision avoidance in relation to some of the issues the speaker had raised. He went to ask whether, in the browser, it’s possible to know if you’re on wireless or not? Agustin responded that, in the wild (as it were) it wouldn’t be possible for him to tell.

Geoff Houston (APNIC) commented that he was confused about where the presenter was getting RTT measurements from. Browser noise may make it difficult to get a proper signal. He pointed out that you need to look as kernel operation. When speaker sees that signals are erratic, it’s because browsers are erratic. After some back and forth between him and the speaker on the topic, Geoff made the suggestion that looking at the signal on the other side of the browser would enable the presenter and his team to get a much cleaner set of measurements.

D. A New Almost Reliable UDP Protocol

Nuno M. Garcia (Universidade da Beira Interior)

The presentation is available at:
https://ripe75.ripe.net/archives/video/118/

A speaker from the audience stated his support for the presenter’s idea. He went on to ask whether they’d managed to produce any measurements on performance? He also asked to which kind of application this this can be applied.

Nuno responded that he and his colleagues don’t know exactly which applications this can be applied in yet. It is not yet known exactly where the protocol can be applied. He added that it may be the case that it can be used wherever there’s a UDP connection. On the first question, he suggested we shouldn’t worry about performance. He clarified that the role of his team is to provide the algorithms, and that they would generally leave others to worry about speed.

Brian Trammell suggested that the team might have looked into building a decoding layer. He added that what they are providing is not a transfer protocol, and that they’d need to build a protocol on top of this. He raised an ‘open issue’ suggesting that they will need to reinvent TCP on top of this to make it work. Suggested that they should look at what the transport community has already attempted and failed at. Also suggested looking at failures to due NAT TP. The idea is that by looking at those encodings, we see some of the issues here go away.

In response, Nuno pointed out they have looked at other proposals. The argument is that they don’t need to focus on the protocol ‘on top’ of this – we won’t necessarily need real time reporting. Brian also pointed out some security concerns.

Jeff Osborn (ISC) pointed to the loss of a certain result in the packet table provided by Nuno. Nuno explained that packets can replace others in the process – in some cases packets are not elected at all; in others, packets that were elected were not picked up.

Ruediger Volk (Deutsche Telekom) returned the discussion to Brian’s remark about security. He suggested it would be useful to do an analysis of the attack vectors that become available on this approach. He pointed to a few potential security vulnerabilities for the process.

E. [Questions were held until the three updates had been given]

E1. RIPE NCC R&D Update

Chris Amin

The presentation is available at:
https://ripe75.ripe.net/archives/video/119/

E2. Multi Approach Infrastructure Geolocation

Massimo Candela

The presentation is available at:
https://ripe75.ripe.net/archives/video/121/

E3. Crowdsource your Infrastructure Location with OpenIPmap

Massimo Candela

The presentation is available at:
https://ripe75.ripe.net/archives/video/203/

Questions were raised about OpenIPMap. One audience speaker pointed out he’d had some issues with using it. Massimo responded that it’s because they were still releasing during use.

The audience speaker also asked about TraceMON issues – i.e. specific probes not showing up. Massimo explained that they should look into this.

Brian Trammell pointed out his interest in the on demand stuff associated with this, and asked why ten milliseconds had been chosen as the threshold? Massimo thanked Brian for previous work on this, and responded that anything over this threshold and things aren’t useful any more.

Brian asked whether Massimo and the team have been getting useful results with that threshold? Massimo responded that they had, and that it’s possible to add a score to the results and so on.

Wolfgang asked whether there’s been any though of using DNS records. Massimo responded that there had, and that this would be added in the near future, as well as reverse DNS.

F. Reflections and Saying Bye

Christian Kaufman

Z. AOB.

Daniel thanked Christian for all his contributions to the WG