The Jupiter dream

Yes, there are NPCs that stay at their position, but thats basically all they do.

They do not actively do anything while you are far away. The game only has to remember the position of the NPCs (that doesn't really affect performance) and spawn them once you are next to them or in the same chunk. The NPCs do not interact with the world while you are far away.

In a classic RTS or simulation however NPCs are constantly interacting with each other and the world. (for example shooting at close enemies, collecting resources, path-finding, transporting or building something, capturing something, getting hungry, patrolling ...) which also requires path-finding and scanning the nearby area which means that the computer/server has to be fully aware of it (keep it in memory).



Yes, there are a few exceptions. Complex redstone circuits would be another example but as you said yourself they and TNT can easily kill performance because of being such exceptions.

You would never want such exceptions in a MMO. It would be terrible if player structures or actions could easily kill the server and there being only official servers (they would surely never allow player hosted servers as they want to make money via ingame-items).



Well, endless voxel words that can be completely manipulated and have NPCs are performance heavy.

I don't know many games that fulfill the criteria while at the same do much better in terms of performance than the current minecraft and more importantly aren't affected by the same compromises.

Actually I can't think of any. But even if there are games with a 3D (voxel) world that offer endless or at least planet-sized worlds that are procedurally generated and can be completely manipulated by the player while at the same time being more performant than the current minecraft, they in all likelihood have the same compromises as minecraft.

Or let's take a different approach.

What were the biggest maps, player limit and unit limit in a realtime strategy/simulation game that wasn't like a browser game (NPCs could interact with each other or the world like collecting resources)? I think the most ambitious would be the currently developed planetary annihilation which offers planets however each planet is more like a map in a classic RTS and not even barely comparable to minecraft proportions / real planets.

Yet alone massive multiplayer with thousands of players (I think PA only supports 8 or maybe even 16) or terraforming which affects wind and weather on top of that. And even more crazy erosion and dynamic water. And as if that wasn't ridiculous enough cross-plattform development with a mobile tablet.

The claims peter made were absolutely lunatic, no matter which game you compare it with.

Exactly. That is the point I try to make, that is exactly the reason why having such a world and having realistic erosion over time or complex wind and weather simulation or "simulating every grain of sand and every drop of water" was absolute nonsense by peter.

Having a "browser game" with a jupiter sized word - no problem.

Having "from dust" and "banished" with a jupiter sized world - technically impossible.


Having a lumberjack that produces X logs per hour without affecting the world - possible.

Having a lumberjack that actually goes out and chops trees down (requires pathfinding) - impossible with that scale.


Persistent multiplayer means that Godus plays forcedly like a browser game as every calculation has to be simple (for example doesn't require path-finding). Either that or it freezes like Sim City did (which wouldn't be true persistent multiplayer). And apparently even the singleplayer is meant to be persistent.

22cans imposed heavy limitations upon themselves with their design decisions.

Without relying on resource nodes or infinite trees - the need to continually expand and have each individual lumberjack needing to pathfind to find a tree every time it goes out to work.
Multiply that times a dozen or more lumberjacks per player. Add in a forester or other method to regrow trees (or infinitely expand outwards).
Then multiply that times hundreds/thousands/millions of players.

The main reason most browserbased games work off of buildings and nodes producing their resources is because then its simple calculations. Every hour you get X wood.
There's no further processing needed. If every single woodcutter would also need to have its pathfinding calculated, the trees have their full animation done, perhaps add in some physics to make it look fancy as it falls down. Every additional step gets exponentially larger consequences.


Then figure out which parts have to be done server-sided, which can be done clientsided and what degree of communication and cross-checking is needed. The infrastructure required would be insane. While it is *technically* possible. It isn't very likely to happen and would be ridiculously expensive to achieve. If they can even scale that up to the supposed 50 million+ players they claim to be able to support. (I suspect part of the reason they're holding off on so many things isn't just because they don't want to scare of new players, but also because they just wouldn't be able to run it on large scale.)

I don't see why. You can apply the same trick to the locations and types of trees and have it stored locally and then every once in a while perform synchronisation.



Why would a server ever have to handle pathfinding?



I don't think Banished uses a server.

Your original question was why it would be impossible to have a lumberjack that actually goes out to chop trees, like in banished where a lumberjack actually cuts trees and they require some time to regrow etc. and the further away the tree is from your lumberjack and his storage the more time it would require to get to that tree and carry it back. So actual resource management like in black and white, command and conquer and banished.

So instead of just placing a building and having X resources generated every hour you have to make strategic decisions about roads, storages, terrain, exhaustion of resources and efficiency.

Danjal already answered that question.

Also if you distribute important calculations to the client you have to somehow validate them afterwards, what if the client manipulates memory to make his lumberjack go through buildings or other cheats?

Generally nothing that is calculated or sent by the client can be trusted as it could be easily manipulated by said client and it should either be avoided or properly validated.




Yes, if it would use a server the performance would most likely be worse.

Online gaming requires more "trickery" (estimations) and compromises than you might expect.

If they don't workout you experience things like rubberbanding (when you're running in an online game and after a minute you're warped back to an earlier position - what happens is that server and client are out of synchronization and after a minute your client gets updated to his real position) or rollbacks (your progress wasn't saved and is lost).

Its unlikely that 22cans will set up a server infrastructure that exceeds battle.net or EA and as far as I know the latest Sim City already had multiple problems (you couldn't log in at times, progress wasn't proberly saved etc.) with very small and limited maps and it was actually more a singleplayer experience than multiplayer (everyone played on a server but everyone played for themselves, your city wasn't active if you weren't active etc.)

If EA had difficulties the company that required 5 months to implement dragging with your mouse and more than 1 year to get a working exit button will not perform better than that.

You are totally right, see here the new topic. Feel free to make suggestions for a name for it coz I don't have the foggiest.



By the way, I realize I haven't been around much lately. I sort of have more important stuff to attend to at the moment. This topic could have been split sooner. Sorry for not being around. I blame Qetesh for not keeping an eye out for me while I was off doing important stuff she wanted me to do.

No harm. I totally understand.

No idea about the thread title. Something that eludes to the discussion being about the reality of the Jupiter sized world.

There must be some amazing path finding algorithm that nobody has yet thought of.

Okay, here's an idea. (perhaps)

Say you consider a mapsize of 256 * 256.

Now you write a function that discards all the areas of the map that can not possibly be reached and stores the remaining coordinates in an array of some sort.

Then you write another function that calculates every possible route there is from every coordinate in the array to every other coordinate. A route consists of a list of all the coordinates on that particular route. You store all these routes in a database.

Then you write a function that goes through all the routes in the database and removes all those that don't make any sense. (or you find a way to remove them before they go into the database)

For each route you calculate the total distance or perhaps rather the total time it would take to travel along the entire route. (perhaps taking slopes into consideration or something like that)

Then you go through all the remaining routes and iterate over each coordinate in each route. While doing that you create a new database (or table) that has an entry for each valid coordinate and as values it will have a list of all the routes that have that coordinate in them.

This basically gives you a large database that has all the information you might want to have already pre-calculated.

Now when you need some path finding from coordinate A to coordinate Z all you need to do is query the database and retrieve the shortest route (or one of the shortest ones if you would want some type of randomization) and it will give all the coordinates the character needs to be able to travel along that route. No calculation required on either the client or server side.

Now let's add in some landscaping. Whenever the map is changed by the player the affected coordinates are sent to a server. (I think it's probably best to calculate server-side in this case) The server runs a calculation from every affected coordinate to any other reachable coordinate. (I skipped the data cleaning step, I know) These new routes are written into the database while the previous routes pertaining to those coordinates are removed from it. Thus the route database is completely up to date again and any character requiring path finding can simply query it to find a suitable route.

Now I don't know a lot about optimizing databases for speedy data retrieval, but as pointed out by splitterwind Google can retrieve data from the internet in microseconds. (of course Google also has loads of gigantic warehouses full of servers and data storage)

I don't know if applying this structure will be fast enough but there is probably plenty of room to cut some corners to make it faster.

This leaves of course the issue of multiple 256 * 256 maps that are stitched together. I haven't really thought about this yet, but let's say you want to move from point A on map 1 to point Z on map 2. You could try "drawing" a direct line between the two points and finding the coordinates of both maps where that line intersects with the map edges. Then you look up the route from point A on map 1 to that coordinate on the edge of the map. Then you look up the route from the coordinate at the edge of map 2 to point Z. Presto, you have found a path. I'm sure it's not ideal but I can picture it working fairly okay.

As for everything written above there are likely many kinks to work out along the way but I have a feeling this could work. What do you think?