#182
Mar 13, 2022
Work
I started my new job at GoCardless on Tuesday.
I’ve never worked in finance before, and their infrastructure is built on Kubernetes, which I only have a basic knowledge of, but most of the code is Rails, which I’ve been using for years now at GDS. Of course, every project using a particular framework does so in its own idiosyncratic way, but there tend to be more similarities than differences.
So some things are new, other things are not so new.
My team seems nice, and owns some of the core components which the rest of the business depend on. It’s satisfying to work on something big and important.
I basically spent this week reading documentation and running through the technical onboarding. I’ve got some more business-focussed onboarding sessions next week, but I’m hoping to start contributing code as well. I much prefer to dive straight in and learn by doing over just reading and shadowing.
I’ll still be mostly working from home, which is great. Everyone on the team tries to come into the office on Wednesdays, but other than that it’s up to whatever you feel like.
Since this all is now closed-source private sector work, I’m not entirely sure what the future of this section of my weeknotes holds. It’ll certainly be more general, and maybe it’ll be more focussed on big-picture career things than the day-to-day.
Books
This week I read:
Volumes 7, 8, and 9 of So I’m a Spider, So What? by Okina Baba
I’ve got to the big plot twist which I saw people online vaguely hinting at, and which I’d looked up in advance because I just do that sort of thing (if a book is ruined by spoilers, it’s not a good book). It definitely feels like a good place to have an anime series cliffhanger, so I wonder if series two will do that. It would only be four books (since the first series covered volumes 1 through 5), which is a little short, but I’m sure it could work out: volume 7 has a lot of action, which could easily take up a few episodes.
I’m now half-way through volume 10, so I expect to finish all the translated volumes next week.
resolved
Not that much progress on my DNS server this week. I did make one big decision though: I decided to bring authoritative zones into scope. Previously I’d decided that static records were enough, but I think that properly supporting zones will be easier in the long run.
I’m not yet running resolved
for my LAN DNS. But I’m close now. Before I switch that on I want to:
- Implement the zone file parser
- Add the necessary configuration to support RFC 6761
- Package it for NixOS
Nearly there now.
I also wrote a memo on how the cache works, which got featured in This Week in Rust 433.
Code wise, I merged two PRs:
Implement zones & refactor nonrecursive resolution
Which is a significant change to the nonrecursive resolution: it makes the code look much more like the algorithm in RFC 1034; unifies my implementation of static records and root hints; and lets me return authoritative NXDOMAIN responses for local zones.
With this in place, I can now start on implementing a parser for zone files, which will let me remove the existing (much more verbose) configuration mechanism, and means I’ll be able to implement RFC 6761 with, hopefully, no code changes whatsoever—just configuration.
Drop support for non-IN record classes
Since I don’t have access to CSNET, the Chaos network, or the Hesiod network, supporting records for them just makes the code more verbose, since I have to check that classes are consistent all over the place. By just dropping support for those classes entirely, I simplified a bunch of logic.
And I opened the following new issues:
Review & implement RFCs 2181, 4343, and 4592
These three RFCs clarify a bunch of implementation concerns. Not all the content is relevant to what I’ve built, but I definitely want to go through them all to see if I’ve made any common mistakes.
Implement RFC 6761: Special-Use Domain Names
This RFC defines some special domain names which should be answered based on local information, rather than forwarded to the root nameservers. This issue was a major driving force in me deciding to support authoritative zones: if I support authoritative zones, this RFC is just a configuration change; but it’s a code change if I don’t.
Wildcard queries should consult authoritative nameservers
I noticed a discrepancy when running
dig A barrucadu.co.uk && dig ANY barrucadu.co.uk
.resolved
only returns theA
record for theANY
query, whereas my Pi-hole forwards the query upstream and returns all the records.Basic overview of how DNS works
This is a documentation issue, to just write a little something about how the protocols and types work.
Document how configuration works
This is another documentation issue. When
resolved
supports zone files, it’ll be very helpful to just have a quick summary of the format. And also what all the other configuration options are.-
Really, this should have been the first thing I built. Instead, I built a whole nameserver before thinking that a little tool to query an upstream nameserver and print out the result would be a handy debugging aid. Sure, there’s
dig
itself, but debugging my DNS resolution by starting the server and querying it so that it’ll in turn query an upstream nameserver, is more awkward than just querying that upstream nameserver directly. Add executable to convert a hosts file into a zone file
This probably won’t be very useful. But it should also be very simple to implement. So why not?
-
This is, I think, the only must have before I can switch over my LAN DNS. I need to actually be able to build and run
resolved
from my NixOS configuration, so that I can manage it in exactly the same way as I manage all the other services on my systems.