they’re different files generally, the only client that will automatically request them is a debugger.

you turn them off because you don’t want to expose your full source code. if you would be ok making your webpage git repo public then making sourcemaps available is fine.

most things should have an alternate implementation, just in the unit tests. imo that’s the main justification for most of SOLID.

but also I’ve noticed that being explicit about your interfaces does produce better thought out code. if you program to an interface and limit your assumptions about implementation, you’ll end up with easier to reason about code.

the other chunk is consistency is the most important thing in a large codebase. some of these rules are followed too closely in areas, but if I’m working my way through an unfamiliar area of the code, I can assume that it is structured based on the corporate conventions.

I’m not really an oop guy, but in an oop language I write pretty standard SOLID style code. in rust a lot of idiomatic code does follow SOLID, but the patterns are different. writing traits for everything instead of interfaces isn’t any different but is pretty common

all of their examples are pre ai anyway. it’s almost impossible to change a core data model thing without ai too.

all the legacy codebases I’ve worked in have very much been shaped by the original abstractions they decided on. as much as I wanted to change them, there wasn’t really a way to do so because of the scale and backwards compatibility requirements

some things are easier to change than others. it’s easy to slop on new features, rewrite them etc. changing core models after you’ve built a ton on them isn’t easy, even with ai. the odds it comes up with the perfect data model aren’t great, but for isolated features that doesn’t matter since it’s easy to throw them away and rewrite

steam is one of the few commonly used 32 bit apps left on linux.

I imagine most of it is bc most other things are oss and have been updated/rebuilt already. having to run a 10 year old binary happens way less on linux than it does windows.

a handful of distros have tried to remove 32 but support they’ve gotten backlash bc they’d lose steam support. linux the kernel won’t drop it any time soon, but there’s a good chance that if steam drops 32 bit, so will fedora etc

yeah, ofc it should only be a curated set of errors where the consumer can do something about it. unknown errors should just be opaque 500s

I love clj. in general it takes more of a pragmatic approach to functional programming than others. you get most of the purity for way less effort. same with specs over proper static types. it just ends up being a very enjoyable and productive language

sql as the language executed by the db hasn’t changed notably, but I do think there’s been significant developments in ORMs. for a lot of developmers sql is now just an intermediate target

more directly, sqlite was originally for tcl which is why they share the semantics.

also I’d argue that sqlite is a bigger contribution than tk, but I suppose in a more roundabout way

early js/html liked to do something in all cases instead of throwing or whatever. I think it’s mostly just a collection of them trying to do something smart on nonsense input and not being consistent about it.

side note, I’m so excited for Temporal, some browsers already support it and you can polyfill for the rest.

I have a GL.iNet GL-MT6000, works great. not sure about newer models but people generally seem to like their higher end ones

it ships with almost vanilla openwrt, and it’s easy to install an upstream build if you’d like. the big advantage of GL.iNet is that it’s officially supported, so I’ve been able to send emails to support about openwrt stuff and they’ve been helpful

AdGuard works great and is built in to their version, so you wouldn’t even need the separate device for pihole

I also like supporting a company that supports oss software

NixOS, plasma rn but sometimes jump to sway. I’d say distro is more relevant. for the most part I just have an editor and a browser open, DE doesn’t change much about my workflow. NixOS definitely does though

chosen by my team, company at large doesn’t care but it’s nice for everyone to be on something consistent. company devices

NixOS is a nice balance of the two

I generally just copy my personal setup, which I’ve spent a decent amount of time on, but because I enjoy it

not particularly, but nix supports all of the big ones

language and stack a little bit, it’s all stuff that has good integration with nix. we deploy nix containers and then have consistent environment everywhere without having to work in a container. my team is a pretty standard team maintaining some full stack web stuff

idk if clojure has really faded though. some dialects have done well (jvm, js) and some haven’t gotten much use (go, clr), but it feels like a reasonable path. there’s a good chance you can tap into a decent chunk of the existing clj ecosystem too

I feel like it’s probably not a high priority, but the company I worked at that selfhosted gitlab was also paranoid about dependencies disappearing and so mirrored every repo they had a dep on.

I imagine that’s not that rare of a situation and it would have been a nice qol kinda thing if we could have federated with the upstream and gotten a backup of issues and such and could do everything on the one platform. definitely not important and requires upstream to also federate, which will never happen for github so not important

I’m saying we weren’t taught when react was the way people wrote sites. if I was writing a site with pure html, css is great, especially modern css.

but if I’m already using react and their abstractions, opinions on that part aside, I’d personally rather lean on the react component as the unit of reuse. tailwind removes the abstraction that you don’t need, since many people in react tend towards one scoped css file per component with classes for each element anyway

at this point I’d be more inclined to say for many sites the api and data fetching things are the content and html+css is presentation. csszengarden is cool but I haven’t seen the html/css split help an end user, or really even me as a developer.

instead of using classes you just use whatever your ui library provides for reuse. stick a classname string in a variable and you have a class. use a component and it just contains all its styles.

unless you mean that if you look in the inspector you see a mess of classnames. I don’t have a solution there

shadcn is the primary one for react at least. they’ve done a great job filling the space where you’re trying to build up a design system but don’t want to start from scratch, but they’re great if you just want prebuilt components too

all the components build on something else like radix, and are pretty simple themselves. normally just the radix component with styles. Installing a component just copypastes the source into your project at configured locations.

if you’ve ever fought against something like mui to get it to fit design changes or change specific behavior, shadcn is great. at some point the extension points of a library aren’t enough, but if you own all the code that’ll never be a problem.

except we generally use higher level abstractions now, like component based frameworks. If you’re writing raw html with tailwind and no library you’re doing it wrong and css is a better fit.

well written straight css ends up building it’s own tree of components. if you’re using react too you’re either only selecting a single component (inline styles but have to open two files) or writing good css (duplicating the component hierarchy in css).

tailwind is just the former but better since it encourages using a projectwide set of specific sizes/colors/breakpoints and small scope, the two actual problems with inline styles after organization and resuse, which react etc solves.

I’m sure all of them are just cherrypicked hotfixes from main tho

repr is generally assumed to be side effect free and cheap to run, so things like debuggers tend to show repr of things in scope, including possibly exit

also then it behaves differently between repl and script, since repr never gets run. to do it properly it has to be a new repl keyword I imagine, but I still don’t know if I’m sold on the idea