Tony Finch – blog

This weekend I was in Rotterdam for the RIPE DNS Hackathon.

About 50 people gathered with several ideas for potential projects: things like easier DNSSEC provisioning, monitoring DNS activity in the network, what is the environmental cost of the DNS, …

At the start of the weekend we were asked to introduce ourselves and say what our goals were. My goal was to do something different from my day job working on BIND. I was successful, tho I did help some others out with advice on a few of BIND’s obscurities.

The team I joined was very successful at producing a working prototype and a cool demo.

runner up

The project that was the second most interesting to me was “DNS OOPS”, out-of-protocol signalling. The idea there was to find out things like when a zone has been loaded and is ready to serve, so that it can be added to a BGP route advertisement.

I talked about it with Willem Toorop, and I said OOPS sounded close to what I had done with nsnotifyd, but OOPS would need a little more information added to the NOTIFY messages, for instance if you want to monitor key rollovers.

The OOPS team ended up using nsnotifyd as part of their demo, and Lars-Johann Liman reported a usability bug to me, which I fixed.

our project

I joined a team to work on Emile Aben’s idea for a scripting language that could run on RIPE Atlas measurement probes. Instead of collecting voluminous data that later needs to be reduced in a complicated manner, a script on the probe could strip out unwanted data before collection.

I suggested that Starlark might be a good language to use. It’s a simplified dialect of Python designed for sandboxed scripts embedded in applications. It was originally created for the Bazel build system, but it’s a general purpose alternative to Tcl, Lua, Guile, etc.

Annika Hennig suggested WASM as an alternative, which sounded cool to me. But it was too difficult to get a WASM runtime up and running with functions plugged in for DNS resolution. So we switched to the Golang implementation of Starlark.

results

We got Starlark to invoke dig or the RIPE Atlas tool evdig I suggested using dig +yaml but we ended up using jc --dig which is new to me.

Annika Hennig got us started with Starlark-go, and made an awesome front end based on Codepen that could dynamically deploy scripts to probes. Raffaele Sommese wrote a bunch of Starlark scripts, and got it working with evdig. Jonas Andersson set up a RIPE Atlas probe onto which we could install our prototype software, and sketched out some security considerations. Emile gathered use-cases and wrote the presentation. I hooked more functions up to Starlark and examined security of the Starlark implementation and our prototype code.

The demo was deploying a Starlark script from the Codepen front-end to a couple of probes; the script monitored changes to the .com SOA serial number across gtld-servers.net.

conclusion

Starlark turned out to be nice to work with, and Golang is a good hackathon language. I have not used Golang much before so I didn’t write much code; I was much more successful at coming up with ideas and suggestions.

For the context of this project, it looks like Starlark will be more accessible for data scientists than the possible alternatives, as I hoped.