Subnostr

当你完全控制 relays 的时候，虽然会麻烦很多，但是也透明很多。现在有很多 nostrgram 这样的客户端，用户体验可以做得很好，但是它的“优化”可能也是有代价的，它对你可能不是很透明、可以完全掌握的。现在社区的趋势是一直在强调更智能的软件，降低用户门槛，但我们可能也要注意做所谓“优化“的时候不要牺牲到控制权和透明度

Replying to

Mike Dilger ☑️

I don't know, all I can tell you is what I have tried and how that went. #[0] is working out design ideas but I don't think he's interested in the coding part. Are you?

I tried Tauri and found the undebuggable magic that hooks Rust and JavaScript together to be a massive minus. I have performance problems and I could not figure out why. Using javascript rendering with VueJS or similar was nice (it is what I do for my day job), but it just didn't work. I hear "good things" from other places about Tauri, but it didn't work for me.

I tried gtk4 (since it's newer than gtk3 and new projects maybe should use the newer one), but the rust bindings are not documented and are not quite like the C++ bindings, so it was very painful trying to figure out how to do in rust what is well documented in C++. The people who wrote the rust bindings were no help at all. I found that slowed me down so much that I abandoned gtk4.

I did not try QT.

I did not try other rust-native UIs such as druid, iced, slint, or azul.

egui has two problems, but they aren't as bad as originally imagined.

First, it is not easy to do beautiful interfaces. You don't get beautiful fonts with color emojis or subpixel anti-aliasing or even hinting for that matter. The default widgets are ugly 1990s looking things, so you may have to build many of your own widgets. BUT given that it is OpenGL you can draw anything that OpenGL draws so strictly speaking there are no limits to what you can do. It's just that some things are very complicated and most people don't want to build the top-end of a GUI from scratch.

Second, it is immediate mode, which means it re-renders every frame. This turns out to be no big deal at all. I run it at 10 FPS and it looks fine. Many frames don't require a redraw as nothing has happened, so it might only draw 1 frame per second. And the rust code that specifies what to draw is much faster and simpler than the pixel shaders (for example) that determine what color each pixel needs to be, and we aren't doing anything like that in rust per frame. So I find my CPU appears pretty much idle, even though it is respecifying what to draw for up to 10 times per second. I find it "fun" to make sure the UI code doesn't block and doesn't do anything compute intensive, and optimizing it really is my cup of tea.

So in summary I'm not against sticking with egui, but to make it look beautiful and to fix the UX, it will probably require some uphill UI development which might have been easier in a different GUI.

/wall-of-text

ElectronicMonkey 2y ago

I have tried dioxus https://dioxuslabs.com/, it use webveiw(I think?) and offers a react-like style to write interface. I have this little relay GUI that can be run on desktop for non-coder people https://github.com/digi-monkey/fly-castle , but dioxus still not feels like a very good one since it lacks some tools compare to react.js

Replying to