I run my Mastodon instance in a Docker container because Mastodon’s code is such a rat’s nest it won’t install on my bare server because it’s not ubuntu. This morning its database container decided to shit the bed for no reason and corrupted the database. Thankfully I had backups, but it’s not exactly what I wanted to do this morning.
Shit like this is happening almost daily. Npm should require all releases and commits be signed.
I’ve been working on a php DOM-related project lately trying to wrap the built-in DOM, so been pouring over the spec a bit. I’m sure others have noticed this before. You can’t create CDATA nodes in an html document, but you can import them from an xml document and append. Why? If CDATA nodes aren’t in the HTML spec then why are they allowed to be imported? Seems to me if the document is an html document they should be imported as regular text nodes.
It's nearly impossible to tell which tab is the active one at any given time because there's no continuity with the rest of the UI, and I can envision sometime in the not so distant future where some clever nasty people start phishing off of people using Safari because of how the bottom bar just disappears into the content. Even when colorized Vivaldi still has a crystal clear separation between UI and content.
I decided to install Safari's preview version on my Mac today, and I know people have flogged this dead carcass over the past week but damn is it fucking terrible. They rip off Vivaldi wholesale with tab groups and colorization based upon content but fail in really weird amateurish ways to make anything usable.
"Why CAPTCHA Pictures Are So Unbearably Depressing" by Clive Thompson https://onezero.medium.com/why-captcha-pictures-are-so-unbearably-depressing-20679b8cf84a
A hilarious read. And yes, I too hate doing free labor for Google, especially when it's a bit creepy and unsettling.
Babylonians Used Applied Geometry 1,000 Years Before Pythagoras
Mastodon requires 3 separate processes, and one, sidekiq, leaks memory like a cheap diaper leaks piss. It slowly consumes storage space because there's no mechanism built in to cull cached avatars and header images of every single person who sees you on Mastodon. To solve these I wrote two cron jobs: one to restart the sidekiq process every 12 hours, and another a hackish attempt to delete unseen month old images. This shouldn't be needed.
Developer, washed-up designer, and occasional Illustrator. Ook!
Single user instance for Dustin Wilson