Comment by cosmodisk
4 years ago
It was probably 20 or so years ago when I played for some time with Delphi. It was extraordinary easy to make program interfaces. Fast forward to today and we are in this clusterfuck, where everyone keeps reinventing the wheel and complexities just keep growing.
Yeah I feel that in a way, the best time to be a developer was back in the 1980s/90s. Your tools were limited but those constraints took away a lot of the "overhead" thinking about what frameworks to use and you could just focus on functionality. You didn't have Google or StackOverflow, but had a few books on your desk that covered pretty much everything you needed to know. Or if you were working on Unix, you had man pages, K&R, and Kernigan and Pike's The UNIX Programming Environment
The reason was you weren't competing with the world class. Programming was great because you could be a local hero.
Now everyone strives to make as beautiful websites as <insert big blue with their genius web framework>
We also didn't have security issues, which helped.
In the same way that Visual Basic offered simple, drag & drop interface builders.
What happened is also changing hardware, UIs that need to adapt to changing screen sizes, different needs, theming, and many more.
Making resolution-agnostic applications in C# Winforms wasn't hard, it was a simple flag to tell the OS how to scale the GUI. And if you used the native widgets and set tab-indexes you'd be all set for changing sizes, blind users, OS re-theming, etc. A good UI framework should handle that stuff internally... even a bad one should do that (bad, like how winforms set the wrong default font).
Imho, the real reason we don't see stuff like this for the Web is that the web isn't designed for modularity. CSS, Javascript, and HTML IDs are all global.
Programming 101 lesson 1 is "don't use globals" and the Web is the perfect object-lesson in why not.
CSS - Yes. JS - Not since 2015, I'll admit that is somewhat recent. HTML IDs - IDs are only good if they're unique. Since HTML had no notion of scope, they became global. Shadow DOM is the web platform answer for modularization, however any JS framework will allow you to slice your CSS and HTML in components.
> if you used the native widgets and set tab-indexes you'd be all set for changing sizes, blind users, OS re-theming, etc
> Imho, the real reason we don't see stuff like this for the Web
Where are you looking on the web? tab-indexes and extending native web components gives you responsiveness and accessibility. The browsers provide theming capabilities for light and dark mode, and OS level color preferences (I use "red" for selected on Mac) easily show themselves on CSS `outline` etc
> Globals
No one uses globals on the web. This isn't 2000, or even 2013.
Everyone seemingly unaware that this still exists and you can still do it in Hejlsberg's C#? Drag and drop WPF components and program in code-behind? With themes and responsive design?
(Big asterisk: mostly Windows-only and has recently lost product direction coherency)
My biggest problem with the drag and drop builders was how it played with version control. The last I used was nearly 20 years ago - Windows Visual Studio C++ with MFC. Even if the builder produced C++ code, it was hard to know exactly where and how the changes were made. CVS (and maybe svn?) didn’t exactly like tracking those changes.
Is that any better now?
1 reply →
Also Android Studio, XCode, all the React builder websites, etc. Basically every GUI platform + IDE has some form of a drag & drop interface builder.
The problem is that outside of the iOS ecosystem, there are too many subtle differences in behavior to really trust the results. And because a lot of software gets worldwide distribution these days, it's economically worth it to squeeze that last bit of performance & user friendliness out of the framework. So most professional programmers learn how to do things programmatically and only use the interface builder if they're doing a quick internal tool.
WinForms still exists and is fine for the odd quick utility. Doing WPF without ever having to drop in the XAML is... unlikely to say the least. You're going to end up in there are one point, because styles are fucking you over or for some other reason (databinding? lol). But yes, C# is still keeping the drag & drop alive. So are plenty of tools (Lazarus, hell even Android Studio has a drag & drop designer).
Windows forms are great, whilst WPF is a bit odd,to say the least.. Here's an example:
https://docs.microsoft.com/en-us/visualstudio/get-started/cs...
1 reply →
Delphi still exists and you can still do it in Delphi. Responsive. Cross-platform on Android, iOS, macOS, Windows, and Linux. Single codebase single UI.
1 reply →
> UIs that need to adapt to changing screen sizes, different needs, theming, and many more.
People keep saying that as if that is this new thing that wasn't ever heard of before the web or smartphones.
Open any desktop app and resize it. Boom, you've got "changing screen sizes".
And yeah, "needs are different, and we need theming is this new thing that never existed before the iPhone ".
I think forced aspect ratio was very common in the era of these oldschool UI toolkits. Additionally, handling multiple classes of pixel density and input (touch vs mouse) was unheard of.
> Fast forward to today and we are in this clusterfuck, where everyone keeps reinventing the wheel and complexities just keep growing
Yes, it is quite funny to see how no toolkit exists to simply produce Web UIs in a meaningful way.
However, the complexity has grown largely from externalities that didn't exist during the time of effortless interface builders, which is screen sizes and aspect ratios and pixel densities of all sorts. To handle this, you need to have some lower level primitives, and of course any time you have to go to a lower level you surface more complexity.
Final point - as a person who develops web UIs professionally for 7 years now, I think that the "reinventing the wheel" has been actually quite beneficial to tame this complexity. Previously, untyped JS had to be bent into surfacing type-style error messages, and good luck with boundary crossing data. Now, TypeScript lets you describe every key in your application and have incredible confidence that a fully-typed piece of UI or logic (which of course must avoid `any`) will deliver exactly what you intended. GraphQL & codegen has given us the ability to type our boundary crossing data straight from our DB or resolvers without any runtime reflection. Runtime reflection tools like io-ts also bridge that gap admirably to program defensively in the situations it's needed. It's obviously been accompanied by a lot of churn, but with strictly typed component libraries, a bit of reusable layout logic, and Hasura, I can make sexy fully-themable UIs strictly typed all the way to and from the data source without significant effort. The complexity in my new paradigm is entirely in application-level tricks like UIs visually informing users of all the async actions, animations / transitions, avoiding dynamic content causing bad layout blips, and ensuring user input is never lost. I think this kind of thing wouldn't have been easy in any oldschool toolkit because it inherently requires some wiring that isn't easy to surface
> where everyone keeps reinventing the wheel and complexities just keep growing
Cue XKCD "Standards" comic. People look at an existing framework and declare "this is total shit, I can build a better, easier to use version!" They then start building the better-easier and realize why the old version is so hard to use--because it's a difficult fucking problem begetting awful complexity + shitty code.
This repeats itself every 18-24 months, giving us the current clusterfuck of JavaScript libraries. Lather, rinse, repeat for the past 30 years (n.b. XWindows Athena -> Xt -> Motif/Lesstif -> ...<aeons pass>... -> Qt -> Electron)