One counterpoint I can think of: most forms of electronic payment require the Internet now. Credit card transactions, Venmo et al. You could transition back to cash but there would be enormous switching costs and short-term chaos, and I could imagine paper-based transactions are also way less efficient in terms of transaction fees and literal loss of the cash.
One of the bizarre ironies of the world is that mobile phone based payment systems achieved widespread adoption in third world countries before the USA! (That's largely because the USA had less need, there were plenty of established alternatives.)
That's one of the reasons why poor villages found mobile phones to be so essential.
it's interesting that the drops tend to collect in straight lines. I wonder what's happening in the sim code to keep them from collecting into round droplets?
It is strange to me that omarchy took off and not Regolith Desktop [1], which is a very similar project, and has been around a lot longer. I suppose the DHH effect is real. There is definitely a critical mass accumulating around the hyprland ecosystem. They seem to be forming their own culture separate from the wider FOSS community that I find concerning.
Oh man, I forgot about Regolith! I ran it for a few weeks on an old ThinkPad a few years ago when it had a new release and it was pretty nice compared to configuring i3 and all that myself.
It sounds like a natural expansion of AWS GovCloud offerings to me. Servicing the US government and it's contractors has been very lucrative for AWS. Taking that successful model into new markets makes sense.
I don't think you could 100% nail the Win7 glass or Win11 mica material without having transparency effects that have access to the compositor pipeline. But that doesn't mean we shouldn't try to get close enough!
HomeAssistant's dependency requirements are so complex as to make a deployment of it essentially read-only. Trying to mod it with HACS is so hard that I gave up even though I write python for a living. I can't recommend it to anyone that doesn't know how to use linux at an expert level. This detracts from it's mission IMO; I would like to tell my family to use it instead of Google Home or Amazon whatever but it is so fragile that I can't do that in good conscience.
My install has been essentially maintenance free for years. I don't imagine many casual users that just want an app/gui for their smart outlets will be reaching for HACS. HA even warns you when you install HACS that you're voiding all the warranties.
HACS is very easy _if_ you’re using the HA OS installation method, in which Home Assistant manages its own containers (requires bare metal or a dedicated VM). I’d have no hesitation recommending it to someone who is just tech savvy enough to operate Home Assistant itself (e.g. setting up automations).
I’ll also echo the sibling commenter re: running it for years with no maintenance besides occasionally installing the updates.
Another selfish reason: web pages just work a lot better when you use the actual HTML elements, especially when you compose them together. React projects often mix several component libraries together to make a comprehensive UI. All of these libraries behave differently in subtle ways. When you compose them together, the differences compound: focus is not restored to the button that opened a dialog when it is dismissed, there are 4 different blues used on the page, the date input doesn't use your country's date format.
When you use HTML primitives like inputs with associated labels, the new popover API, dialogs, details + summary elements, their behaviors are all made by the browser vendor and are designed to compose with each other. It really is a difference of night and day, and for free. We don't take advantage of the amazingly powerful tools we have been given.
Good intentions and some good ideas. But ultimately lead to a fractured, overly complicated and bloated web.
I say yes to lower level, more general capabilities. But why does that have to mean we get bad APIs (web components, indexeddb, service workers...) and lack fundamental and well established UI components (combobox, toggles, tabs, tooltips, data grids...) and incomplete implementations (lacking HTTP verbs for forms)?
(Also why we're at it, I'm going to sneak in a complaint about half baked and broken language features like `const` and `import`. Also we shouldn't have abandoned Do-Not-Track...)
"Const" is half baked as it implies that you declare an immutable value, but it only prevents redeclaration and direct reassignment.
"Import" breaks the browser REPL and the dynamic nature of JS. You can't redeclare, look or move into a module. Now JS is already broken in that you can't redeclare variables declared with let/const. But not being able to move into modules an extra crime.
`const` works that way in most of the C family that uses it as a keyword or is restricted only to "primitive" types. To be fair, I agree that JS isn't really a C-style language under the hood and as a functional programming enthusiast I wish `const` was called `let` and `let` called something like `mut` or `let mut`, but JS likes its pretend-to-be-C-family surface level so I'm not upset about it.
`import` works great in the REPL, the various REPLs I use have caught up. Most of them in REPL mode let you redeclare imports and reimport modules as needed. They also all support top-level await so it's just as easy to use `let { thing } = await import('some-module.js')` if you want more of a safety net to redeclare as you wish.
Just some blatantly obvious examples: Select Multiple was some Windows 3 garbage, and HTML didn't have a 'combobox', much less a 'multi-combobox'. (MDN says datalist is not supported on Firefox, so maybe it still doesn't.) So write it yourself, or use a library. The form validation stuff is still bad, and idk if modern desktop browsers have a good date control.
I can agree that it is bad. And there might be some examples where <select multiple> is the "best" way, but I cannot ever remember a real-world use case where it was not better solved using a list of <input type=checkbox> anyway.
I must be confused because the use case I'm thinking of is a list of 5,000 somethings that are filtered down through user-input text in the combobox, that they can then select, and selected items are then displayed as bubbles within the input field or elsewhere. Certainly can't just display 5,000 checkboxes instead :P
Yeah sure, the 'right' way was checkboxes in an iframe. Just that the weblords care alot more about adding syntax to javascript than whatever multi-select-searchable-accessible-box you might want. The form stuff still mostly sucks.
To you and OP, as the movie goes: "Yeah, well, that's just like, your opinion, man."
My opinion is that we could use fewer "web applications" and more web pages that simply provide structured semantic data and then get out of the way and let the User Agent (and user himself) control the presentation of that data. Sure, some things have to be a web application. I'll grit my teeth and admit that. But many, many sites we visit could easily be raw HTML and maybe some CSS (which the User Agent is free to cherry-pick through or ignore). They would be fast, functional, accessible, (probably) more secure, (probably) more private, free of telemetry, and would better serve the user's interests over the web company's interests.
Instead we get JavaScript "instructions" from the web site, which require the User Agent to faithfully execute, often removing the user's direct access to the actual content and enforcing the web developer's opinion on how it should be presented.
If we had more use of standard controls and light markup, and less JavaSludge, maybe browsers would devote more time towards fixing and improving their standard controls.
Just another random person's opinion, but this has been a terrible direction for the web.
The whole reason we need web "applications" when a simple web site will do all boils down to the demands of marketing and advertising. If you were to go to a major news organization's web site, the bulk of the javascript isn't for frontend UI frameworks, but advertising. Go to a video streaming platform, and they do Weird Things not to prevent scraping, but to inject marketing analytics for advertisers even when you're on an ad-free plan.
Marketing and advertising makes the world a worse place. When you look at the real numbers, it rarely achieves whatever business goals it's supposed to achieve, though for some reason they keep throwing money at the crap because maybe this time it will be different.
Last product I worked on I made this a thing. Every modern browser now renders a very competent date picker for input type="date".
I find a lot of the arguments for building inaccessible UIs seem to conflate historical web issues with modern times. Virtually everything that sucked in the core web toolkit circa 2010 has essentially been solved in 2025, even on ultra-restrictive platforms like iOS Safari.
Native date inputs have their problems, but I default to them usually because they typically have the best behavior on mobile and are localized by default.
Very cool that you can use the XCode interface builder to build the UI.
I wonder why this type of style hasn't caught on with React and friends? It would be really nice to be able to have an AppKit-quality UI programmable in React or Svelte.
I know I know mobile blah blah. But lots of web apps are complicated enough to only be useful on a large screen, like Figma.
reply