miniblog.

Type systems can be nominal or structural, they might force an option type, and they might even encode side effects (IO or exceptions). I think the biggest single improvement (where compiles≈correct) is exhaustiveness checking. This produces thoroughness and often robustness.

I don't always have anything to add, but I always appreciate folks responding to my toots <3

An incredibly ambitious project to archive all software source code, *and* all commits, to preserve it for future historians: https://m-cacm.acm.org/magazines/2018/10/231366-building-the-universal-archive-of-source-code/fulltext Impressively, they've already mirrored GitHub and Debian!

It was amazingly easy to add Tramp support to deadgrep.el. A single function change and suddenly you can run searches on remote machines too!

Whilst IPv6 addresses are often longer, I do like that ::1 is less typing than 127.0.0.1.

This is a neat idea (alias works in zsh, bash needs a script). Even better, perhaps our websites need to exclude $ from being copied the way line numbers are usually excluded. https://twitter.com/brandon_rhodes/status/1050570678032850944

Websites designed for desktop often don't work well on mobile. I've started noticing sites where the opposite happens: they've gone mobile-first, and the desktop UI suffers. It probably makes sense to prioritise mobile, but it's hard to do a good job on such diverse platforms.

Travis CI is adding Windows support! https://blog.travis-ci.com/2018-10-11-windows-early-release (Article also includes some interesting statistics on Windows popularity and Travis popularity in the npm community.)

OpenBSD adds unveil(), a new syscall for restricting filesystem access. Like pledge(), it aims to maximise adoption through having a limited (but very comprehensible) API.

I'm really impressed with the Grasshopper app for teaching coding to complete beginners. In this typo, it tells me the code is incorrect even though it would be a valid JS program!

Rust Remacs 2018:

On making parsers more human-friendly by adding grammar prodctions to make the parser total: https://duriansoftware.com/joe/Constructing-human-grade-parsers.html (allow everything in the AST, but mark errors)

Code quality metrics are very subjective and depend on the domain and the programming language. They can add value, but I rarely see tools that work well without initial configuration.

GitLab has an excellent, candid discussion of how they're doing 'open core' development, and resolving the tensions that occur:

A remarkable short story and thought experiment on charitable giving in a smart contract system: