[adamrezich]

If you use someone else's high-level game engine to make a game (metaphor: use CAD software to design mass-produced wooden products), but you could write a simple game yourself “from scratch (with libraries)” (you have carpentry experience), then you're a game developer (carpenter). If not, then you're not, you're a proficient tool-user, with no domain knowledge outside of how to use said tool.

> Also it’s a little funny you bring up the example of an XNA/FNA game, because that’s already a relatively high level framework, the kind of code those devs would use is really quite similar to using an engine, especially another C# engine like Unity.

This is untrue and you know it. XNA/FNA does not have a GUI editor, a scene graph, a physics engine, an entity/component system, and many other things that Unity et al. have. When you use XNA/FNA(/MonoGame), you have to figure out how to structure your game logic yourself. The structure you think in terms of is by and large something that you've written yourself, as opposed to with something like Unity, where a complete structure is provided for you, and you learn to think in terms of the primitives it exposes to you (GameObjects, Components, Prefabs, Scenes).

[MattRix]

Ok I think we’ve come to the root of the problem. You don’t realize how much software development people who use engines still have to do! Just because there are things like GameObjects and Components does not mean you don’t have to think about the game logic or structure. There are tons of ways to approach things, those are still just a relatively low level framework.

And just so you don’t think I’m making stuff up, I wrote a framework for Unity that acts much like XNA or some other 2D scene graph. It’s 100% code based and essentially doesn’t use Unity at all except as a publishing tool. It’s called Futile and it was used in popular games like Rain World and Mini Metro, as well as many of my own games.

However, I also have made games “the Unity way”, with GameObjects and Components etc. These two different approaches to making games are not nearly as different as you claim. I still have to think a lot about game architecture and system design even when using all of Unity’s built-in foundational stuff.

[adamrezich]

Yes, I do realize how much work people need to do to make games within arbitrary frameworks provided by game engines, and alluded to such in previous posts.

XNA never had a scene graph though, I don't know where you got that idea from. XNA and its descendants give you a basic game loop, and then it's entirely up to you how you structure your game logic from there.

There's nothing wrong with using someone else's structural backbone, like the library you made that you mentioned -- but everyone who wants to self-describe as a “game developer” should try to do this themselves, without such a thing, at least once.

Video games are computer applications, and computer applications consist of data, and functions that operate on said data. When you learn to think in these terms, stuff that seemed complicated before becomes very simple to reason about, and also very easy to write code for. Sometimes, a more generalized scene graph-type system is the best fit for what you're trying to do. Oftentimes though, it's not, and you're unwittingly contorting your thinking about the problem space to fit the shape of the predefined structure you've chosen to use.

One example of this is how Game Maker (in the old versions—I don't know much about the newer GameMaker Studio) is structured. You define “Objects”, and each Object has, in addition to many other properties, “Events”, each of which can have one or more “Actions”. You also define “Rooms” (scenes), with a 2D GUI editor that lets you place instances of Objects within the Room. The first Room in the Room list is where your gameplay begins.

When you first start using Game Maker, you make a Room for each level in your game, you define Objects, and you place instances of Objects in your Rooms. You define logic with Events within Objects -- init stuff goes in the “Create” event, and update stuff goes in the “Step” event. You drag and drop Actions into Events in your Objects to define gameplay logic.

This is all well and good, but eventually you grow out of using any Action except “Execute code”, which lets you type GML code in an integrated editor. Eventually, you learn that it's a good idea to keep your “controller” logic in a single instance of a single Object—separate from your player character Object—called “objController” by convention. You put a single instance of objController at (0,0) in your first Room, and mark the “Persistent” checkbox in objController so it persists across Room transitions.

At this point, you're “fighting the engine”. The engine does not provide a first-class means of defining “controller logic”, so your only means of doing something like this is as I described above: defining an entire Object type—which contains tons of properties and functionality that you'll never use for this Object—just to allow you to define some code that gets run every frame. As an advanced Game Maker user, you get used to it, but you wish that this common pattern was given first-class support, so you didn't feel like you were “fighting the engine” to do something so obviously common.

As you become an even more advanced Game Maker user, depending on the type of game you're making, you might end up eschewing most if not all of the structural features of the engine, just so it “gets out of your way” and lets you write the code you want to write to make the game you want to make. You might end up as I did—having only a single Room, containing only a single objController instance, and then handling everything else yourself from there.

When I reached this point as a teen, had Raylib or XNA existed and had somebody shown it to me, I would've said, holy crap, this is exactly what I'm looking for. (XNA did come about shortly thereafter, and it's where I ended up going after outgrowing Game Maker.) I just want to draw sprites and text and simple geometric shapes to the screen, and I don't really care about learning low-level graphics stuff to learn how to do that myself, so if you can handle that for me, that's great—but I can handle managing all of the game's data in structs and arrays myself just fine though.

But apparently this is not the case for many modern game engine users. They learn to “think in terms of” whatever primitives the engine provides them with, and then decide that that's good enough, they don't need to learn anything else. They're already thinking in Scenes/Prefabs/GameObjects/Components—why stop now?

I firmly believe this is huge mistake, because it's such a self-limiting way of thinking about game development, and it causes you to become entrenched in your limited way of thinking about gameplay logic. But before last year's Unity debacle, my views were very unpopular, because, what's wrong with just using Unity, man? Well, now we know: it's not a great idea to invest one's entire way of thinking about game development and one's entire business prospects so heavily in someone else's platform, when they can decide to fuck you at any point in time, causing you to suddenly realize how dependent you were on said platform.

Some of the upvotes I've been getting in this thread are somewhat encouraging—maybe people are starting to come around to my way of thinking? (Not that I have a monopoly on such thinking, by any means, of course.) But many of the comments are incredibly discouraging—people really like to tie their entire identity to the specific-engine-specific way of thinking about game design and development, apparently, even when pitfalls of doing so are clearly outlined, and it's blatantly demonstrated how easy it is to begin to think outside of that box.

[MattRix]

I think you’re getting consumed with HOW the game is made rather than WHAT is actually getting made. There are tons of super creative and unique games being made with engines like Unity all the time (or even with Game Maker, for example check out my cousin’s game Fight Knight).

The idea that the constraints of an engine are some big problem is absurd. Constraints are a good thing, they breed creativity. There’s a reason every game jam has a theme and a tight time limit, yet result in some of the most unique game concepts.

[adamrezich]

I completely agree that constraints can breed creativity—but so can trying to do things a different way than you're used to. If you've only ever made games with very generalized data structures organizing your game logic for you in some predefined way, then PICO-8 can give you the best of both worlds!

There's plenty of games made with Unity and other engines that I enjoy—I'm currently playing a Godot game and it's great.

But if nobody pushes back on the current status quo rhetoric, then many people will a.) continue to have extremely incorrect misconceptions about what necessarily constitutes a “game engine”, and how difficult it is to make one, b.) not be encouraged to try to learn something new that has more general applicability outside of the domain of their preferred tool, and c.) continue to find themselves in situations like Unity developers did last year, or like what some Godot developers are experiencing right now—feeling trapped, with no immediately-viable path forward.

In this past year, I've built my own “engine” that runs a networked open-world game that loads world chunks in and out of memory as players move around the world—but unlike other games like this that I've seen, certain types of NPCs can also freely move around the world, even when there's no players nearby. This was not trivial, but it would've been at least an order of magnitude more difficult, and much less performant (even though I haven't done any optimization work on it yet!) if I was using someone else's engine, shoehorning my approach to solving the problems that this design I wanted to make presents, into said engine's way of doing things—instead of just defining structs and functions that do what I want them to do and writing the whole system myself. It wasn't trivial, but it was MUCH easier than I would've imagined before trying it myself, given that I'm using a library for rendering, and another for network sockets.

So constraints can certainly breed creativity, but if these constraints are left unchallenged for long enough, then you're just limiting your potential for what you can do.