The technical blog of William Manley
I recently came across this article by Matteo Vaccari of Thoughtworks via HTMX discord. As part of developing Stb-tester’s web interface we’ve increasingly been moving to a traditional hypermedia approach1. This has meant a different (and I think superior) approach to unit testing than with SPAs. Traditional Hypermedia to me this means:
I recently discovered this 2010 JSConf presentation via HTMX discord:
See my colleague David Röthlisberger’s website for an article we authored together: Simple declarative schema migration for SQLite.
I often see Rust’s cargo package manager dismissed in online discussions by analogy to pip and npm. Cargo/rust doesn’t suffer from many of the problems that pip does. Some of these reasons are not just about cargo, they’re also about how Rust is different to Python and about how Cargo and Rust integrate:
From an LWN comment by Neil Brown, kernal hacker and author of the Ghosts of Unix Past LWN article series (among others):
Here’s a comment from Ericson2314 that I agree with wholeheartedly:
An article written by my colleague David Röthlisberger in part describing how the build system that we built at stb-tester works and the philosophy behind it: Merkle trees and build systems.
An idea that I’ve had burrowing into my mind for quite a few years now is that free software distros are held back by the crap build tools that are used to build software - but the lack of availability of better tools isn’t the cause of the problem.
TL;DR version: Improve integration test sandboxing with systemd socket passing. You can allocate a random port for your daemon and don’t need to wait for the daemon to start up to run your test. It’s fast, robust, race-free and doesn’t depend on systemd.