Every go dev goes through a rustspringa

I’m also hyped for errors.AsType!

Seems up for me?

charcuterie drift

I'm a fan of using

#![cfg_attr(debug_assertions, deny(clippy::panic))]
#![cfg_attr(debug_assertions, deny(clippy::todo))]
#![cfg_attr(debug_assertions, deny(clippy::expect_used))]
#![cfg_attr(debug_assertions, deny(clippy::unwrap_used))]

wow those are some great hashtags lmao

The real production appview itself uses Scylla, but there are other services using Postgres that are part of serving the whole Bsky experience (discover is the biggest, but also chat, the relay, ozone, etc…)

I run the on-prem Postgres instances we use for the appview and I can assure you it does not spark joy lmao

RDS is alright for smallish workloads, just expensive

I mean this in the most respectful way possible, but no, this is not correct

Totally agreed. We do a lot of SQLite within the bsky appview itself and it all has to go because it’s now load bearing in the worst possible way

The problem IMO is that the other obvious options have a huge amount of accreted cruft. We need a much better Postgres!

Everyone hates when I say this but sqlite on the server is a blunder

Perhaps the only exception is for software that’s written to be deployed widely by a variety of people (ie the PDS)

I’ve been deeply impressed with Umass Lowell’s low level, hands on approach

Not sure if that’s what you’re looking for, but for instance he’s taking an OS class that requires him to fork the Linux kernel and add new features

Emission of a record over the PDS websocket does not mean the record has been indexed in the relevant appview db yet. It has to traverse the relay (or several) then eventually land in the db

I am dumb as rocks and totally blanked that this was last night!! I'm sorry! I'll catch you at the next one!!

Bunny rocks! They’re great!

we've had one outage, yes, but what about second outage?

do you think it feels good for the carrots when they're harvested?