Hacker News Reader

Comment by Cthulhu_ | Hacker News Reader

Given the sheer amount of work that went into Factorio, it feels like a waste to just park it and move onto something else; it has a solid base to work off of. Its performance and the amount of Stuff it simulates is amazing.

I'm thinking they may get inspired by some of the mods out there; there's a mod to go into space (build a factory inside of a spaceship) and land on other planets, for example. They could reach out to those and buy it, iirc that's what happened with some of the KSP features and addons as well.

I wouldn't mind if they'd do something big with the enemies / monsters. Maybe introduce a new antagonist entirely.

Or a whole new class of sea and/or undersea base buildings, with unique resources that have to be launched up to the surface for your main base to process.

Or satellite bases too far away for even trains and the like, where you need long-term transports (container ships, ICBM's) to get stuff.

View on HN

>> where you need long-term transports

I've always thought that factorio should have bulk transports, things that arrive after multi-hour trips. There are a handful of mods that tie the character into an external economy. Rather than the survival scenario, I think Factorio 2.0 could be based on taskings surrounding the arrival/departure of transports. Rather than "build the rocket" the task could be "have 200k of blue circuits ready for the arrival of a transport in X days." Changing up the tasking would allow for truly different factories to be developed. I'd love it if the game suddenly told me that I needed to switch from one end product to another, any excuse to delete an entire factory and build a new one optimized for a different goal.

This is currently actually available as a scenario! External delivery using a very small area. Check it out if you haven't!

Regarding buying the mod that goes into space, they did one better and hired the developer of that mod.

From the blog post:

> One of the steps we have already taken is hiring a concept artist, he might be known to you under the nick of Earendel

Earendel's list of mods for those curious: https://mods.factorio.com/user/Earendel

Yes please to all your ideas!

> Or satellite bases too far away for even trains and the like, where you need long-term transports (container ships, ICBM's) to get stuff.

There's one thing that could work with this idea that I haven't seen in any game so far. You've built this complex factory and have it running like clockwork. The game should be able to observe it - to notice that the flow of inputs, intermediate products and outputs has been stable (even if cyclic, because you built a solar-powered factory or have some intermittent backpressure on the conveyors) - and then abstract away your factory, simulate its matter balance via closed-form mathematical functions closely approximating the steady-state behavior. This way, you could have many complex factories running nearly for free, while your attention is focused elsewhere.

Factorio itself, in its current form, may be a tad too complex for this idea to work (at the very least, there's extra complexity in managing finite resource deposits), but I think it's doable in principle. But for some reason, as far as I'm aware, nobody ever tried it.

(Even in Kerbal Space Program, the closed-form math only applies to stable orbits; internal state is not processed. There are mods that try to simulate e.g. life support or power flow for such craft traveling "on rails", but as far as I know, they all do that by running numerical simulations and have problems with error at high time accelerations - whereas my idea would be to approximate a continuous function of time for relevant values, and use it instead of numerical integration.)

I think the problem with this idea is that it has to work extremely reliably, or it basically doesn't work at all. And that's inherently very difficult to achieve in a game that's based on simulating a complex system.

I don't think you can retrofit an efficient closed-form model onto an existing simulation; it has to be designed in from the beginning. I suspect that's a big part of why KSP only simulates two-body conic section orbits. They knew they wanted to include time acceleration as a gameplay feature, and so they limited the simulation to closed-form orbits for which that could be implemented stably and reliably.

In Factorio, a well-designed factory can be approximated by its steady-state behavior. But part of the reason it's an interesting and challenging game is that it's not always obvious how well a factory satisfies that criterion. You can have bottlenecks that only show up in unexpected conditions, or systems whose throughput depends on the details of inserter positioning, or feedback loops that switch from controlled negative feedback to runaway positive feedback when some parameter crosses a threshold.

Given how easy it is for these kinds of simulations to accidentally end up being NP-complete or even Turing-complete, I'm very skeptical of the idea that an algorithm could make reliable decisions about when it's safe to "optimize" an entire factory without drastically changing its behavior. I think any game simple enough to make that tractable would effectively be an entirely different genre. (I would be pleased to be proven wrong, though!)

That said, you could certainly optimize the simulation of individual components that are themselves simple enough to not have chaotic behavior. In fact, Factorio does some of this already. For example, long transport belts are simulated by tracking the position of items relative to the belt, rather than their actual coordinates. So an item's "movement" isn't simulated on every frame; only when it reaches the end of the belt, or gets picked up by something, or if you're actually looking at it. (https://www.factorio.com/blog/post/fff-176)

I share your skepticism on the closed-form model. I do think there are other opportunities to optimize a sufficiently remote base though. right now (as I understand it), factorio needs to keep anything that can affect the global train state and systems connected to the same pipes, belts, or electricity grid consistent on a per-tick basis. in practice, this is basically everything. but if a remote base had its own train state and all the belt/pipe/wire connections were internal, the updates could be deferred to a more convenient time and/or offloaded to a background thread. that base would only need to be consistent when the player is looking at it or a bulk transport vehicle arrives to pickup/dropoff goods.

> Or satellite bases too far away for even trains and the like, where you need long-term transports (container ships, ICBM's) to get stuff.

I like this idea. imo the core of factorio is conquering scale. the base game is already very polished. at this point I just want more stuff to do post launch. taking the scale up another order of magnitude (huge ships, point to point air transport for latency sensitive goods, etc.). would also provide the opportunity to expand combat a bit to provide automated defenses for far away shipments. I want my robot navy, now shut up and take my money!

Reply on news.ycombinator.com