n4te

Use pi. Can't trust OAI or Anthropic to make good tools.

Swing layout was bad then (gridbaglayout!) and is bad now. The framework's approach is bad. Components provide min, pref, max sizes that are very often just ignored. Components need a layout mystery object in order to participate. The mechanisms to measure and constrain components sizes and positions is bad. None of the layouts are sane and it's difficult to add your own. 95% of layouts are nonoverlapping components arranged in row and columns, but provided layouts do other things.

Swing was themeable but good luck if you ever try to do it. Customizing an existing theme is too hard. All components have opaque theme objects that are part of the complexity problem I mentioned and often hide functionality and parts you want to customize.

It's easy to shit on everything, but I really think UI could be done much better and that most or all toolkits make big mistakes that hurt productivity and cause a lot of suffering.

True those platforms have security controls, better than what we've ever seen on desktop. Desktop also needs controls like that: to use the system tray, filesystem, hardware, etc. There's at least a work multiplier of 3 for Windows, Mac, Linux. It's just one of many things needed.

Building the web toolkit took enormous effort and it sucks in many ways (see article). Theoretically a new one could be created and could also work for desktop, but there are lots of reasons why that won't happen.

Hard disagree. Lots exist but none are great. Usually layout isn't sane. Another blunder is components are too complex with too many layers. Customization and creating your own components becomes difficult. Often the event system is terrible.

Eg Swing has all those problems, in addition to ugly themes. The uncanny valley resulting from trying to mimic native UI was bad. At least nowadays users don't necessarily expect native looking UI, even browsers don't do it.

It's not a question of can you make a nice UI with it. You can drive a nail with a rock. The high pain needed to be productive with bad tools leads to Stockholm syndrome. People would rather stick with what they know than go through such pain again to learn something new, and rightly so when the new one is likely just as bad in new ways.

Ah. One difference is a normal desktop can access a lot more or even all of your system. Otherwise cross platform UI toolkits exist: GTK, Qt, Swing, SWT, etc. They could be great but they aren't, not because it can't be done well just no one has managed to do it well. They could technically be sandboxed but they aren't, or not well. Java applets tried long ago and a lot of systems got owned.

You'd never design from the start the bonkers HTML/CSS, box model, and other junk we are stuck with, but browsers have a number of good aspects that you'd probably end up with in a complete redesign.

The problem is multifaceted. UI toolkits are very hard to do well (I have a lot of opinions on this, and my own crossplatform toolkit), even on a single platform. Just that alone is huge. There are basically no overall great UI toolkits, even if some have good parts. Even a theoretical great UI toolkit will have quite a task to achieve parity with what browser UIs can do.

If we could get past that (spoiler: we can't, no way in hell) then we'd also have to be willing to ditch all the legacy browser stuff to switch billions of users and devices to something new AND we'd still have many other big problems to deal with: getting the committee/everyone to agree without ruining it, security, privacy, etc.

We've accepted we can't really fix it, so the only thing we can do is keep extending it. That's how we got here.

No, tables are the right solution: nonoverlapping elements arranged in a sane way that is easy to reason about. The arguments about semantics so everything can be a div are hilarious -- it really doesn't matter.

Yes, a terrible one. That's what the article is about.

Windows Firewall Control is good, binisoft iirc.

The Bay Area start up money bought nice marketing that makes it look popular, but it's not really.

Close! It was 1 dev and 1 artist for a long time. Eventually the runtime and support load became too high and it's slowly grown to 9 people. Running lean was satisfying but exhausting.

The only thing that makes sense to me is that the body gets more efficient.

Good point, that also works. For some reason I never remember to use it.

Ha, it works great! I like it, thank you!

The blog header popping up constantly makes the page unreadable.