Stream: ideas

Topic: Drag Racing Roc Packages


view this post on Zulip Luke Boswell (Oct 28 2026 at 04:49):

@Zak Kohler made a drag race for his web framework Roux... and talked about it in the last meetup. I just had an idea while looking at roc-awesome and thinking about the growing list of packages in the ecosystem ... I wonder if we could extend the drag race idea, and generalize it somehow to run races for basically every package.

Getting the environment could be a challenge, but I would pick a linux machine running Nix and roc-blueprint, then we can pin the Go/Rust/Python/Javascript tooling.

I don't know if this is a crazy idea... but basically pick the smallest DO droplet or something, and have it continuously updating and running the drag races and reporting that in a website?

I've done a bit of work on all my parsers for example, and things like roc-time or roc-pdf etc... basically I'd love to know how everything compares, and track that over time too.

I've mostly been benchmarking the implementation to help me find obvious bugs and ensure it's efficient. I haven't been aiming to be the fasted or anything. Just doing this alone has caught an enormous amount of correctness bugs.

I imagine it would be beneficial for the community to understand how Roc packages are performing across measures like speed, memory etc. I've had a lot of fun challenging myself with performance goals on various things.

Interested to know what others think about this.

view this post on Zulip Luke Boswell (Oct 28 2026 at 05:13):

This is a can of worms ... :sweat_smile: so many cool ideas keep popping up

view this post on Zulip Luke Boswell (Oct 28 2026 at 06:43):

Definitely getting carried away with this... here's an idea I've been iterating on with Claude

https://gist.github.com/lukewilliamboswell/16bb4ac647b0e3094842dd5bb377a731

I had Gemini imagine what it could look like https://share.gemini.google/05kJR74QK6nX

Screencast from 2026-10-28 17-40-57.webm

view this post on Zulip Karakatiza (Oct 28 2026 at 07:11):

Definitely fun, but I imagine there will be few groups of packages that want to race against each other - once a package for a thing is out there there is little reason to create another one, unless there is a lot to improve on API UX or performance. Web and mackend frameworks, parsers are a notable example

view this post on Zulip Karakatiza (Oct 28 2026 at 07:13):

Benching platforms will likely be a thing to try, but some of those need special environment/runtime

view this post on Zulip Krzysztof Skowronek (Oct 28 2026 at 07:26):

I've been watching this yt channel recently: https://youtu.be/-BFt49ula08?is=byZ_Zf0_YUXwGlfl

I got Roc on roux to be comparable with the winners, at least in my laptop. I applied the most common optimizations, so batching writes to sqlite, parking waiting connections and so on. It can now serve 130k users with 0.1% fail rate on sqlite write :)

They are taking submissions for the next round with multiple CPUs

view this post on Zulip notramo (Oct 28 2026 at 11:06):

Karakatiza said:

I imagine there will be few groups of packages that want to race against each other - once a package for a thing is out there there is little reason to create another one

Race these agains other languages, just like the version in the video. Rust, Go, Zig, etc.

view this post on Zulip removewingman (Oct 28 2026 at 18:40):

Krzysztof Skowronek said:

They are taking submissions for the next round with multiple CPUs

would probably be good advertisement for roc, so close before 0.1.0

view this post on Zulip Zak Kohler (Oct 28 2026 at 19:33):

Roux is sqlite only, or else i would have done it.
I was #2 with fourneau for a bit:
image.png

view this post on Zulip Karakatiza (Oct 28 2026 at 19:38):

would probably be good advertisement for roc

Roc gets only one first impression, like anything else. If it gets spread around too fast too early it will leave the impression "this is interesting, but is too early in development. Will check it out in a few years", where I imagine Roc will get much more mature (relatively) and ready to go public in 6-9 months time

view this post on Zulip Zak Kohler (Oct 28 2026 at 19:43):

remember python 2->3 but that was after a decade. and the claim after the fact was "we had no idea how much production code there really was"

view this post on Zulip Luke Boswell (Oct 28 2026 at 21:39):

Re the above drag race idea for packages ... I dont have much capacity to build this right now. I'm trying to clean up all my packages and get everything ticking along with more maintainable workflows etc.


Last updated: Nov 04 2026 at 18:54 UTC