May 9, 2021


Planning my week worked well last week, so I gave it a go again this week. On Monday morning I noted down a few bullet points for each day. I didn’t get everything done by the day I’d planned, but I got everything done overall. Just about: there’s one PR still being reviewed.

I’ll try it out next week too.

Levels of Authentication

The team finished working through this. Everything is feature-flagged off but, when we’re ready to switch it on, it should just work.

I prefer proof to hope, so I think next week I’ll switch it on in our integration or staging environment to make sure.

Service Level Objectives

GOV.UK doesn’t really do SLOs. We have an alert if the error rate is too high, but that’s just a number picked by developers, who revise it up or down if it seems too sensitive or not sensitive enough. There’s no product commitment behind it. We also have lots of alerts for things which don’t say anything really about our level of service, like memory being “too low” (but unused RAM is wasted RAM!).

Introducing SLOs for GOV.UK as a whole would be a big task. But we can at least adopt them on our team for the account-api, since everything account-related is going to go through that app. We need it to be fast and reliable.

So I gave a presentation on what SLIs and SLOs are and why they’re useful, drafted an SLO document & error budget policy, and now we have SLOs on latency and error rate! They’re very conservative at the moment, which is because we’re not used to working in this way and are planning to launch some experiments in the near future; but over time I plan to tighten the bounds, so we get more useful indications of quality.


This week I read:


After moving to Thursdays last week, so a new player could join, they dropped out! So we’re back to Sundays.

I think there are things we could have done better. The new player joined while both campaigns were in the middle of plot arcs, so they were thrown into the deep end; and also I got the impression they were looking for a different tone of game (more wacky, whereas our games are more serious). Well, those are lessons to learn for next time.

The D&D GM said that in the future they might be up for running a West Marches campaign using Whitehack, so I’ve been looking into that. It seems a pretty nice system: the core mechanic is d20 roll-under-stat-but-above-difficulty (very nice), there’s a very flexible class + group (species, profession, and other such affiliations) system, and there’s spell creation. There’s also a very cool “auction” mechanic for handling contests between characters where a single roll won’t cut it, where players effectively make bets on how well they’ll roll.

I’m looking forward to giving it a try.