Hacker News Reader

Comment by raspo | Hacker News Reader

I don't know much about WebGL (that's why I would be interested in reading something like this), but I'm turned off at the very beginning by the choice of this extra regl framework. To me, it sounds like having a tutorial titled "How to create an HTML5 form for your website", and as the very first step installing React or Vue... maybe that's just me. I'll probably still check it out at it seems like a very good write-up.

View on HN

OpenGL (which WebGL is a sister API of) is a relatively obnoxious API to work with, requiring lots of calls to set an maintain state, as well as pick which state you're working with for a give call. (conceptually you can think of there as being a bunch of static variables that point to the current state for a call). This is IMO one of the biggest barriers to OpenGL, although I believe they've made some of this more explicit than it used to be.

It looks like this API aims to keep the same abstraction level as OpenGLs API, but without the state management required.

So this won't allow you to go straight to OpenGL, but it will teach you the concepts and objects you use with OpenGL.

One of my favorite tutorials and references for learning graphics concepts is http://learnwebgl.brown37.net/.

Picogl, threejs or Babylonjs would have been better choices, given their usage across WebGL community.

Picogl looks like you still have to manage state, whereas ThreeJs and BabylonJs are both fairly high level iirc?

Seems like the framework used in the tutorial is somewhat in between, where you're still using opengl concepts but without having to do state management.

This is the opposite problem of why I enjoyed the article! I understand why they used Regl (as they say, it's because WebGL is ludicrously verbose)... so the article was great for me as someone comfortable with the the very basic basics!

If you're just getting started and want to do it from completely from scratch, then https://webglfundamentals.org (or https://webgl2fundamentals.org - though I'm starting to fear WebGL2 is not getting as much traction as it should... is it dead?!) is the place to go. This site is really fantastic at explaining what all the boilerplate is actually doing.

It's not supported on Mac/iOS. Apple's solution to this was going to be WebGPU, but it's been in committee hell for way too long, and Apple is seemingly intent on torpedoing any goodwill they have left.

Google has since picked up the slack and is working to add WebGL 2 to WebKit on Mac/iOS platforms.

https://bugs.webkit.org/show_bug.cgi?id=126404

Just to be clear, it is not supported on macOS Safari, but WebGL2 works with other browser engines. On iOS, there are no choices however because all browsers have to use the provided Apple one behind the scenes.

I'm not sure how to interpret your theory of what is happening behind the scenes - Switching WebGL to use ANGLE on Safari/WebKit is a Google-led project and Apple is "meh" on it?

Your link is just talking about the feature being blocked on the ANGLE switch.

Together with Apple employees, as explained in some Khronos talk available on their YouTube channel.

Google is also working alongside Apple on WebGPU (Chrome initial support is macOS only), with Mozilla finally having some initial work ongoing.

Really, it's more a tutorial on rendering algorithms than it is about WebGL per se. There are lots of tutorials out there where you can look up the minutiae of how to set up a canvas, clear the framebuffer and set up a rendering loop. This is about how to draw stuff using fragment shaders in a non-trivial (but still pleasingly simple!) way.

There is comparatively little content for that -- most of it gets sucked down into the mathematics of bump mapping or discussions about stencil buffer precision. I rather liked it.

Part of it is that it requires an art side of things. There's no tutorial that can tell you "how to shader", more than there is a tutorial that tells you how to draw. Well, there are plenty of "how to draw" tutorials but they don't get you that far, it's a process of self-discovery and a skill that you need to practice.

Similarly, there are plenty of tech art / VFX shader tutorials out there if you go looking in the right spot, but they can't tell you how to construct a scene like this. That's something that the art side of you needs to solve.

GL was written in the early 90s with a late 80s / early 90s programming context in mind. It's essentially a giant state machine with some limited programmability bolted on in later additions. It feels really clunky compared to modern frameworks.

I think anyone interested in it should find a basic WebGL tutorial and render some triangles and cubes with vertex coloring and some texture mapping to see how it's done in WebGL and to see how tedious it can be. Having to change a lot of different pieces of state every render call to do what seems like very basic things (is there anything more basic than rendering a triangle?) gets old quick.

Except almost all major APIs, even from those days (e.g. Glide, QuickDraw 3D, Warp3D) used structs to contain state, instead of a global state machine.

Had not been for miniGL and Quake, and GL would have had other relevance in history.

Reply on news.ycombinator.com