GDC 2017: "Insomniac's Web Tools: A Postmortem" by Andreas Fredriksson (@deplinenoise) of Insomniac Games https://gdcvault.com/play/1023961/Insomniac-s-Web-Tools-A
I found this completely fascinating. From 2010 to 2017, Insomniac made their game development tools (level editors, sequencers, animation editors, node-based visual script editors, etc.) with web technology, so they ran in Chrome. This presentation is about what they discovered by doing so, both technologically and culturally.
1/5
They chose web tech for a few reasons:
- Lots of existing libraries out there to use
- Hiring JS devs is easier than hiring C++ devs
- The web is where UI innovation is happening
- Zero-click installation
After 7 years, what they ended up with is actually a spaghetti of different systems: they had lots of JavaScript but also had a local server that communicated with it via WebSockets, but also a central server that had its own API entry points, a native browser plugin, a Flash plugin etc.
2/5
A high level summary of their observations are:
- The UI of their tools didn't feel native. User tests indicated it _felt weird_
- Having a strong barrier between tool data and game-ready data is very valuable
- Being able to use any JS framework for anything resulted in nothing being reusable between different tools
- Dynamic languages make refactoring impossible
- Off-the-shelf web frameworks are not fast enough for tens of thousands of objects
3/5
- Developer tools support of new JS features often lagged behind main browser support
- Browser auto-updates break everything all the time
- People have Chrome plugins which will mess with the guts of the tool
- Off the shelf components, even if fast enough, don't have the professional-user feature sets they need
- It's impossible to stop browsers from caching random JS files, leading to random parts of your app being out of date all the time
- Node.js is dependency hell
4/5
Overall, they made the point that, if you're building on a browser, you're not in charge of a core piece of your product. And, it ended up creating an "us vs them" mentality between the (JavaScript) tools team and the (C++) engine team. No one was happy. So, going forward, they switched their stack to Qt, and started rewriting tools.
Review: 8/10 It was very good, but I can't rate it higher, because I can't imagine anyone seriously considering writing AAA gamedev tools this way nowadays.
I don't actually think that's better than JS, for a number of reasons:
- The debugging / profiling story is worse, not better
- There isn't a proliferation of WASM web frameworks, like there is for JS
- The rendering still has to be done via the DOM, which is harder (impossible?) to use from WASM than from JS
If you're going to write C++ anyway, what's the point of running the result in a browser?
GDC Presentation Reviews
Josh Simmons
aeva
