search

minetest / serverlist

The global Minetest server list server
GNU Lesser General Public License v2.1
52 stars 28 forks source link

More honest points calculations #33

Closed MoNTE48 closed 4 years ago

MoNTE48 commented 5 years ago

In continuation to the last discussions. Just a suggestion, let's discuss it here.

1) Ping. What a fool came up with it? He considers the time between the server list and the game server, not between the client and the game server! If I am in Russia and the server has a ping of 10ms for me, why should it be lowered? Or if I'm in the USA and the server is in a neighboring state. The problem is even more relevant for Asian players, where the ping to Europe can reach 200ms. I think this is discrimination. MultiCraft is working on client-side ping. I plan to do PR in Minetest, if it works well.

2) clients_max. What is the reason for this limitation? If the developers do not believe in their code and the performance of my hardware, I do not have to suffer from it. I registered 121 players on my server. And a few thousand mobs. Nobody heard about server processors, Postgresql, RAID SSD and good optimization?

3) guests. now 1/8 per guest is considered for 0.4. servers, it would not be fair to consider them full-fledged players for old versions and guests for new ones. Now players of 0.4. servers are not considered 1/4 by default. It means nothing in any way, see the code! This is only the server position at http://servers.minetest.net/ ! Client 0.4 will put server 5.0 down the list and vice versa. So this is just for the web page!

ghost commented 5 years ago

first time I can agree with MoNTHE48: (but not at nr 3) 1.) ping is discrimination - as based the one server "ruling" the list, and NOT see the real ping between client and server and so gamer will never learn to use more closed servers for much better game fun !

2.) but it is used with 200 now, I have not seen a server that handle 200 gamer or more yet, the max at Multictaftserver so far I saw was 161, and on my minetestserver was 141 with lag of 0,25, and as the penalty starts with set max OVER 200, --> line 348 and the penalty of real gamer then at 160 (= 80%) we could live with this settings until new proof ! not guessing ! --> line 337

3.) never forget to mention exactly about what you write: --> line 325 ff As long we cant proof where this "(r"[A-Z][a-z]{3,}[1-9][0-9]{2,3}", name)" gamer come from (APP), and they are flooding servers with there low behaviour, I am OK with it. I know this name123 gamer now since about 2 years, and they are still no advantage for us. Even at obsessively they are not allowed ? to use a free own name, standing too often around as might be of ads break, and even on same IP I have seen near 5 diff names ... This app doing that to gamer, have to be stopped - and most of all server had added that rules (? with no_guests) to ban them. Causing a lot of data transfer every time login in, for most nothing.

MoNTE48 commented 5 years ago

@Minetest-One Consider number 3 not relevant. I misunderstood for which version of MT this code

MoNTE48 commented 5 years ago

I think that a penalty for indicating too much maximum number of players is something that is not needed. Even if I specify 100500+ on a server with a Core i3 processor, I will very quickly overload the server and the players will leave my server. This restriction is created to prevent this. But making a server penalty is stupid. Let's wish a penalty for a big lag!

Thomas, instead of stupid discussions, I suggest focusing on the development of the Minetest. We can spend our energy on useless disputes, or on something more worthwhile.

ghost commented 5 years ago

max_user penalty - the max of 200 isn't a problem yet, you/we can go up to 160 and this is still fine.

Yes your are right, the problem is not the code of mts, it is hardware and unlucky choose mods ... and an very old burned in thinking in the core mt coders.

MoNTHE48 I know ... after 30.06. I am off, out of minetest. As realized ... worthless. Ok, tell me what might be wore worth-full - I keep a bit time to discuss short and hard.

sofar commented 5 years ago

I would not accept a client-side ping implementation - it would send needless packet spray and effectively be a DoS tool. A better solution is pr #29 and build on that in the client instead.

ghost commented 5 years ago

ping, wrong here, and first might have to search for all possible solutions, and then discuss about them or other possible solutions, NOT direct blaim on something, without absolute full explain technical or other fact and there advantages or disadvantages - as also a own ping of OUR client (do we have, more than real 300 gamer in one time, even if everyone would ask for refresh list ping - I think that is far from any DDOS comparing, and also why clients cant store the result and a renew might be after a week or one of the top 20 is new not measured yet ?) I hate this direct complain, without any facts about what could be used, how often it might happen, and how long can ping results be used after a use. F''' my firewall has def more complains from bad servers than this might be. ping: client-side

MoNTE48 commented 5 years ago

sofar, we are not talking about the fact that ping on the client side is good. We say that when a server located in Europe speaks to a client in Japan about ping is bad! I agree with Thomas: you should not do ping all the time. Besides, I can’t imagine a situation where the owner of the server says "I want to move the server to another part of the world"! Ping once a week is enough. Ping only the first 5 servers, and further as you scroll through the list, this is my idea. Another idea: We can use your code! I am sure that there will be examples and formulas somewhere. I'm talking about getting the distance between the client and server in kilometers and making calculations. This is something that requires discussion in the minetest engine repository. But not here.

sofar commented 5 years ago

None of you read the ping code in the serverlist.

The code currently penalizes servers with a 400ms or higher ping.

There is no place on earth that has a base ping of over 400ms from the master server list. Any server with a ping value that high is severely lagging or having network issues. Last time I checked there was one server with a 635ms ping and I'm sure it was having some sort of connectivity issue. No other server had a ping value over 300ms or so.

No server is put higher up the list for having a sub-400ms ping.

(I'm not talking about how the ping value is represented to users, this would be a discussion for the engine since that is where it is represented).

IOW the ping code here is entirely reasonable.

So :-1: from me on removing the ping code section.

MoNTE48 commented 5 years ago

@sofar About ping. You're right. Your opinion about the penalty for clients_max?

MoNTE48 commented 4 years ago

2019-8-3 15-31-0 Do you think my 12 core server processor can handle 200 players? ;)

SmallJoker commented 4 years ago

@MoNTE48 Minetest is bound to a single core, what you need is high single core performance to reach 200 players. 10 of your cores will just idle unless there are more applications you're running in parallel.

MoNTE48 commented 4 years ago

@SmallJoker 1) https://github.com/minetest/minetest/blob/master/minetest.conf.example#L3010 2) do you understand that in single-threaded tasks, server processors significantly outperform even the most powerful cpu on home PCs?

sofar commented 4 years ago

Given that your server is listed HIGH already (nr. 1 spot), I don't think that the negative rating for servers with a high max users has any effect. If any, the data you post suggest we should increase the penalty even further.

MoNTE48 commented 4 years ago

@sofar Then server owners will be forced to pass the slightly lowered max_users parameter. Although it will be strange to see 250/200 :)