Hacker News Reader

Comment by whstl | Hacker News Reader

I know I'm going against the grain and this is a super-unpopular opinion, but I really feel like there's room for more specialized engines and more independent engine middleware/tooling in 2023. The "game engines are impossibly hard, so we won't get new ones" discussion only ever talks about Unreal and Unity, where editor, engine and technology are all vertically integrated and the flexibility requires a lot of compromises. Plus you can't really replicate those two, even if you have a great team and 10 years of work on your hands (Godot). This oligopoly is super unhealthy for the industry, IMO.

View on HN

I don't think it's unpopular. We need more and better game software, what exists today is unsatisfactory for a plethora of reasons. I'm personally interested in this because I ship Planimeter Game Engine 2D, which comes with client-side prediction standard, and it's the largest pure-Lua game engine on GitHub.[1]

[1]: https://github.com/Planimeter/game-engine-2d

I think in most game development related forums, tutorials, etc, you will be correct about the unpopularity of building a new game engine. However, it’s definitely not a lost cause in general. I also agree that a vast majority of the ‘engines are impossible’ are predicated on Unreal/Unity as a baseline. I look at Arch Systemworks and see them using Unreal for their heavily stylized games (Guilty Gear, Drgaonball FighterZ, etc) and watch the various talks and articles about them having to bend over backward and work to fit Unreal’s expectations to make the engine render and animate in the manner they require and always wonder if a custom engine would not make workflow and maintenance simpler. I wonder the same for more specific genres, like RTS and HD2D-style JRPGs, where a custom engine would conceivably be a better long term option for developers. I always go back to Mike Acton talking at the first HandmadeCon, where he is talking about small dev teams and their benefits in specializing tools and wonder if that logic should be applied to engines as a whole.

Exactly! Good point about forums/tutorials/courses pushing unification.

And interesting, Mike Acton's point goes all the way back to the late Fred Brooks. Not the popular main topics of MMM, but the "Surgical Team" chapter, where he mentions that there should be someone taking care of the tooling. Now, in 1974 this could have been a person. In the 90s game studios it was common to have specialized teams building reusable tooling. This seems to be disappearing.

If I had to guess it becomes a question of how easily can you replace/train up new people to be proficient with the engine/maintain it. Fighting game engines are probably the simplest game engines there are to make but new people still need to be onboarded on it, the knowledge will be internal mostly, you can't find people with knowledge in $propietary_engine to hire right off the bat let alone maintain it. Capcom will use I believe their RE engine for sf6, already used it in dmc5 replacing MT Framework so it's not like it isn't happening in some places, but the AA developers where most of fighting games sit probably can't justify the costs to train and maintain a proprietary engine versus the cost of dealing with roadblocks in UE.

I definitely agree.

My original point, that was not well articulated at all, was more that there should be a middle ground. The choice shouldn't be only between "homegrown engine" and "Unreal/Unity oligopoly".

I'm saying there should be much more competition when it comes to third-party engines, but also in third-party tooling and third-party middleware.

And this competition should be more specialized in order to demand less of the "train and maintain a proprietary engine" but also less of the "dealing with roadblocks in UE".

HD2D games are made in UE4 and they seem okay with it.

(A lot of people think it's a custom engine though.)

> this is a super-unpopular opinion

The adverse reactions I occasionally receive when I talk about building my own engine are a primary source of my motivation these days.

Let me give you some even-more-unpopular opinions. I think the GPU is a gigantic distraction from fun. Low-level languages are largely a mistake for new engine/game development in 2023+. You can deliver production-quality, interactive experiences on top of GC'd languages that perform better than ASM/C/C++ implementations (both for technical reasons & human reasons).

For me, the best "game engine" on earth right now is .NET7. Nothing gives me as much control over such a broad scope of fundamental primitives. You might be balking at this point, but go look at their System.Numerics implementation. First-party, SIMD-enabled code for projection matrix calculations based upon FPS-style semantics is just the beginning. No royalty bullshit either - I get to keep 100% of the profits I extract.

I should tell you as the publisher of a large game engine that performance absolutely does matter and it will eat you alive after it is too late and you've invested too many years into a particular language ecosystem.

Performance matters but to what extent?

I am not looking to support building WoW2 or COD. This is for a very specific creative vision.

I’ve wondered how much of this comes from the huge amount of game programmers that dive right in with zero CS background. I’m not knocking just jumping in, but a game engine is definitely not going to be feasible with 0 experience/background. Even setting up a sophisticated enough build environment to glue together libraries and handle assets is not going to be realistic for someone who’s only experience is installing mods and maybe an OOP course.

Oh, I'm not talking about regular game developers making engines from scratch, I'm talking about the need for competition to Unity and Unreal from other companies, either in the form of new engines or new tooling. I believe new, more specialized, engines would actually make things easier for newbies.

But there's a point there, lots of recent game developers are strongly against the idea of "new" things since each new engine would be one new software and platform to learn.

Reply on news.ycombinator.com