I don't leave Minetest, but...

[Sokomine]

In my young days I was optimistic and all positive about it :)

Perhaps you just got spoiled and expect a lot more today than before :-) We all suffer from that more or less. The modding options are there - and quite a lot of things are possible. But it will never be a complete simulation of RL either, and there'll always be more to wish for.

I created a world with some mods like plantlife and some others. With this mods you still can play, but if you have mobs and if you punch them and nothing happens, or if you place a torch on a wall and it takes 5 seconds until the torch is rotated from horizontal to diagonal, or the mapgen is so slow that you can literally jump of the (ungenerated) cliff) than you can really feel that Lua is overcharged.

It all...depends. Plantlife is a rather expensive mod. All those plants have very specific demands regarding the places where and how they want to grow. I did take a look at it some time ago but didn't really see ways for speeding things up. At least not without simplifying matters (maybe plants can survive even if there isn't water very close by...). Worlds with full plantlife are extremly beautiful - and extremly expensive regarding mapgen. Newer versions of plantlife split generation up and do it in steps, but it's just distributed over a longer time with less lag and still takes the same (high) amount of time to calculate.

If you want fast games, take a look at servers. In particular, lag, maikerumine and rnd created worlds with extensive and aggressive mob life that worked extremly well and got as close to a shooter as might be possible. Mobs up until then where considered to be rather expensive regarding cpu time. Most established servers work very smoothly.

And don't forget one important point: You can screw up in any language. If something's too fast, there's still the option to create an infinite loop and burn away cpu time until some watchdog catches up. Even if your new game is far more efficient than MT and eventually gains modders who want to extend it - they will come up with ideas you'd never thought of and make the game slooow again. That's just life :-)

And now the devs want to add more functions to Lua? Bad idea

Having extensive api functions is in general a very good idea :-) It's bad if your particular idea cannot be done due to the engine not exposing the necessary function. Not all things are cpu-intensive.

[Linuxdirk]

The modding options are there - and quite a lot of things are possible.

When coming from Minecraft and never were able to write own mods one might think the Minetest modding API is the holy grail. But as soon as you dig deeper into all of this you realize that it’s more like a disposable cup from a no-name coffee to go company. It serves the purpose but it’s not a coffee house coffee.

[Byakuren]

The answer is simple. The client side scripts come from the server. The server/subgame decides, which scripts are executed on the client, not the player. Period.

Yes, that would be the best solution. Make the client as dumb as possible and the server as authoritative as possible. Have all of the logic server-side and let the client only predict what happens. This would prevent the current situation with CSM and would make hacked clients pretty much useless.

Did we read the same post?

[Linuxdirk]

Did we read the same post?

The question is: did you understand it?

[Byakuren]

Did we read the same post?

The question is: did you understand it?

There's nothing about not letting the client load their own mods that implies making the server authoritative. You could even have an authoritative server and still have client-provided mods.

The issue with CSM hasn't been that the server isn't authoritative, but that:
A) The server sends information (nodes) that can be used to gain an advantage when playing with a modded or modified client compared to playing normally or how the server operator would like you to
B) CSM makes it easy to read and display this information

The server being authoritative or not has nothing to do with the current CSM problems because the problematic client mods do not try to modify any game state in a way that clients usually are not allowed to. Hiding information from the players could solve the current CSM issues and client modifications that achieve the same effects, but it is a separate issue from how much of the game is decided by the client.

[texmex]

And here I had such high hopes for minetest_modbase! Guess I can try to maintain it, although it'll be painfully clueless.

Good luck on your endeavors, burli. I hope I'll get to play your new project one day as I'm sure it'll be good.

[Fixer]

burli

https://github.com/SirCmpwn/TrueCraft (in C#, mc beta 1.7.3 clone)

[texmex]

burli

https://github.com/SirCmpwn/TrueCraft (in C#, mc beta 1.7.3 clone)

Actually yes, this comes quite close to what burli describes!

[u19503]

burli what about you mob ai you were working on?
Is that disconnected :(

[burli]

burli

https://github.com/SirCmpwn/TrueCraft (in C#, mc beta 1.7.3 clone)

Actually yes, this comes quite close to what burli describes!

Looks interesting, but looks more like a MC clone. I don't want to clone MC. I want to make what MT's original purpose was: an engine for voxel games. If someone makes a MC clone with it... that's ok. But you should also be able to make a game that works completely different.

But I'll take a look at the code. Maybe I find some solutions for my project

toby: I stop any project for MT because it costs me too much time to maintain two projects, sorry

[texmex]

I don't want to clone MC. I want to make what MT's original purpose was: an engine for voxel games.

Being able to successfully replicate MC in it is probably a good measure on how the engine performs, though I can't speak to its flexibility.

[burli]

I don't want to clone MC. I want to make what MT's original purpose was: an engine for voxel games.

Being able to successfully replicate MC in it is probably a good measure on how the engine performs, though I can't speak to its flexibility.

Can't try it myself. It crashes. But as mentioned on Github: it is an implementation of Minecraft and is compatible with Minecraft beta 1.7.3 clients and servers. And is seems that is uses the same data structure of chunks, file format and so on.

[Linuxdirk]

Can't try it myself. It crashes. But as mentioned on Github: it is an implementation of Minecraft and is compatible with Minecraft beta 1.7.3 clients and servers. And is seems that is uses the same data structure of chunks, file format and so on.

Which isn't a bad thing at all and allows gigantic huge worlds compared to the relatively small 64k³ worlds Minetest allows.

[burli]

Which isn't a bad thing at all and allows gigantic huge worlds compared to the relatively small 64k³ worlds Minetest allows.

I found my own chunk format. It allows worlds with the size of Minecraft and the height of Minetest ;-). I can easily address everything.

Btw: the TrueCraft Server is running. I can create a world and start to play. It takes really long until the chunks around the player are generated. Too long

[Linuxdirk]

I found my own chunk format. It allows worlds with the size of Minecraft and the height of Minetest ;-). I can easily address everything.

So ... you just started developing something from the beginning on and you already have a system allowing to address 222,620,160,000,000,000,000 nodes (59999968*59999968 in size and 61839 in height)? Wow.

Did you hear that, MT developers, DID YOU HEAR THAT!? :)

[Sokomine]

Can't try it myself. It crashes. But as mentioned on Github: it is an implementation of Minecraft and is compatible with Minecraft beta 1.7.3 clients and servers. And is seems that is uses the same data structure of chunks, file format and so on.

Servers as well? That sounds intresting. MC does not run on my machine. There's one MC server I would love to visit for sightseeing once. Isn't there some authentication in the background with the server asking some MC account server weather it's a legitimate account or not? That certainly wouldn't be extended to third party programs.

So ... you just started developing something from the beginning on and you already have a system allowing to address 55,655,040,999,999,992,904 nodes (29999984*29999984 in size and 61839 in height)? Wow.

Come on. You sure know that making plans is one step - managing to implement all, especially the details, is where the work starts. It's no big deal to invent a system that covers a far higher volume. Just add a few digits here and there. To make it run smooth, to deal with all the corner cases...that's where it gets tricky.

[Linuxdirk]

29999984

You answered when it still wasn't fixed. I always forget that its +/- the value :) Which makes it even more impressive. I don't even know what the number is named.

Come on. You sure know that making plans is one step - managing to implement all, especially the details, is where the work starts. It's no big deal to invent a system that covers a far higher volume. Just add a few digits here and there. To make it run smooth, to deal with all the corner cases...that's where it gets tricky.

Of course. But it's still awesome. Just a few days and an addressing system was created allowing to address a bazillion of nodes and MT devs trudge around with a few thousands in each direction ...

I hope so hard that one of the backwards compatibility breaks with 1.0 release will be a massive node addressing system update allowing to have virtually unlimited nodes in a world.

[burli]

(59999968*59999968 in size and 61839 in height)

Actually I can address much larger maps in theory, if I did the math right. But it is ridiculous to have a map of the size of 14 light seconds on each axis ;-)

[burli]

Come on. You sure know that making plans is one step - managing to implement all, especially the details, is where the work starts. It's no big deal to invent a system that covers a far higher volume. Just add a few digits here and there. To make it run smooth, to deal with all the corner cases...that's where it gets tricky.

You are right. I actually have a basic implementation of the map and I need 80 Bits to address a node. I think it is really efficient to access the chunks and the nodes in a chunk.

I also have some ideas to manage chunk borders. I have to figure out some details, but I'm pretty sure that it will work.

[Sokomine]

Afaik most of the problems with increasing the amount of nodes was with the resolution of positions of entities and with drawing the whole scene.

[burli]

Afaik most of the problems with increasing the amount of nodes was with the resolution of positions of entities and with drawing the whole scene.

That's correct. I had to use decimal instead of double for entity positions. Would be possible, but again, ridiculous.

[Linuxdirk]

But it is ridiculous to have a map of the size of 14 light seconds on each axis ;-)

Do it! DOO EET NAO! :)

[burli]

Currently I think about using Unity as client. This gives me time to develop the server and I have a tool to test it. I can write my own client later

[Linuxdirk]

Currently I think about using Unity as client. This gives me time to develop the server and I have a tool to test it. I can write my own client later

If it works natively on x86_64 Linux without having a shit-ton of Gnome- and Ubuntu-Specific dependencies it's all fine :)

[azekill_DIABLO]

too bad. cya around for stuff.