adding multiplayer

[Alshain]

4 minutes ago, technicalfool said:

Not always, and certainly not unless you're paying a large sum of money each month.

As for different clients having different results, the server can choose randomly (or based on a benchmark result) which clients should be authoritative when physics bubbles cross and players get close.

Really, dumping all the math on the server for a game like KSP would be incredibly demanding, perhaps beyond even a powerful host machine if multiple server instances are expected to be running (say for instance, on a commercial hosting service). Pushing as much as is feasibly possible to client-side is probably going to be the only way anybody can run their own server. Possibly the only way a multiplayer KSP could be done, unless someone has really cheap access to a datacentre.

I think you overestimate KSP.

[Lelitu]

5 hours ago, technicalfool said:

Problem with KSP being that it really isn't a game where you want to dump all the math on the server. Right now I can run a Minecraft server and several DMP server instances on the same micro-instance VPS with its single virtual CPU and massive (heh) 1.5GB RAM. Even the Minecraft server is more of a RAM hog than any kind of a CPU eater, at least with the small number of users it gets.

If DMP was primarily server-controlled, you're upping the required machine stats to something on the order of a multi-core dedicated host, which most people just aren't going to afford. While the simple message-passing DMP server has its issues, the only extra brains that really should be added to the server (and probably client too) is whatever is required to ensure proper consistency between clients. Asking the server to calculate everything in a multiplayer KSP game with probably thousands of parts active in multiple physics bubbles is going to be a very big nope.

In terms of actual hardware, server class hardware is definitely much more capable of handling the math load than any consumer grade hardware. KSP would not be somthing that would be suitable to running on a 30th of server(or less), like your DMP instance. However, there's an insoluble problem with trusting the client in all multiplayer games.

People cheat. A lot. I don't think KSP is a game for battles, but for those that do like to stage space battles, would you really want to lose because someone decided to make their ship invulnerable and the server believed them?

People also troll, and cause destruction just for fun. I don't know about you, but I'd hate to see a 60+ kerbal combined fleet headed to Sarnus to colonise being completely ruined by one griefer.

Making the server side authoritative makes many kinds of cheating *much* more difficult, or impossible. 

Edited July 2, 2016 by Lelitu
bad words filter is silly.

[technicalfool]

1 minute ago, Lelitu said:

In terms of actual hardware, server class hardware is definitely much more capable of handling the math load than any consumer grade hardware. KSP would not be somthing that would be suitable to running on a 30th of server(or less), like your DMP instance. However, there's an insoluble problem with trusting the client in all multiplayer games.

People cheat. A lot. I don't think KSP is a game for battles, but for those that do like to stage space battles, would you really want to lose because someone decided to make their ship invulnerable and the server believed them?

People also troll, and act like amazing persons just for fun. I don't know about you, but I'd hate to see a 60+ kerbal combined fleet headed to Sarnus to colonise being completely ruined by one griefer.

Well, it's not like entire (and very server-controlled) Minecraft instances haven't been utterly ruined by one guy with a hacked client before.

And yes, a fully dedicated 32 core, 128GB server could be capable of running a rather large multiplayer KSP match. It'll also cost you more than the average wage in many countries just to keep alive.

A few-bux-per-month micro-instance VPS on the other hand, while it can handle running a bunch of DMP instances and a Minecraft server quite well.. I would shudder to think of how awful the results of trying to calculate several thousand parts in a bunch of physics bubbles at once would be.

Griefers gonna grief, and it doesn't matter where you're doing the calculations. This is what whitelists, backups and banhammers are for.

[Lelitu]

1 minute ago, technicalfool said:

Well, it's not like entire (and very server-controlled) Minecraft instances haven't been utterly ruined by one guy with a hacked client before.

And yes, a fully dedicated 32 core, 128GB server could be capable of running a rather large multiplayer KSP match. It'll also cost you more than the average wage in many countries just to keep alive.

A few-bux-per-month micro-instance VPS on the other hand, while it can handle running a bunch of DMP instances and a Minecraft server quite well.. I would shudder to think of how awful the results of trying to calculate several thousand parts in a bunch of physics bubbles at once would be.

Griefers gonna grief, and it doesn't matter where you're doing the calculations. This is what whitelists, backups and banhammers are for.

yes, griefers are going to grief. But why make it easy for them?

Whitelists, backups, and banhammers also only go so far, and don't at all help with the coming back to find the whole mission you just spent ages on completely wrecked. just make it a bit easier to recover. 

[AlamoVampire]

On 6/4/2016 at 1:50 PM, dsonbill said:

It's amazing how people still claim certain facets of multiplayer won't work with KSP, despite the fact that DMP successfully solved all of these.
<snip>

You cannot possibly plan or control every aspect of what multiplayer is. This is one game where Multiplayer is just a flat out bad idea. The game is too young. You can sabre rattle all you like, but, the fact remains, you cannot have factored in every single possibility no matter HOW slim of a chance it has at happening. I had a funky experience a few days ago with time warp and 2 vessels in a SINGLE player environ that showed me how bad it can be in a multiplayer environ. I set a ship to a reentry course via mechjeb and swapped to a different ship. The remote controlled ship on autopilot kicked on autowarp and I was stuck unable to ANYTHING while the other ship warped.

What about mods? What about part count? Player count? Trolls? the list is as long and varied as there are stars in our known universe and beyond. When this game is 10 or 15 years old, then MAYBE multiplayer can be stock. But now? With soo many issues that NEED PRIORITY, like crashing, or the resurgence of this cannot deploy while stowed bs <as examples that are still  priority> must be dealt with FIRST. Not to mention the risk to single player content. No, multiplayer isnt ready to be a thing yet. Not now, not in the near future. Sorry, but brutal truth is brutal truth.

[technicalfool]

1 minute ago, Lelitu said:

Whitelists, backups, and banhammers also only go so far, and don't at all help with the coming back to find the whole mission you just spent ages on completely wrecked. just make it a bit easier to recover. 

Well that's the thing - as mentioned, shifting the calculations to server side isn't going to prevent griefers, and isn't even necessarily going to make it hard for them. If anything, it could make it easier for a griefer to make a probe-bomb, throw ten thousand parts into orbit (like some do right now with DMP) and slow map view down to a crawl. Only you'd be slowing down the entire server.

Whatever shape KSP Multiplayer takes, it needs to be something that normal people can run on a normal budget. Demanding a dedicated host is not a "normal people" budget. Not only that but it'd also make setting up any kind of official Realms-like host an expensive proposition, even with customers paying for access.

[Alshain]

Mojang has made absolutely zero effort to combat that sort of thing anyway.  That's why there are so few vanilla Minecraft servers out there.  A good modded server with World Guard or Grief Prevention is pretty secure.  If it were controlled client side, those mods wouldn't be able to do anything about it.

From a personal standpoint, I'm less concerned about that (my servers are private, I know all the players personally) but it is an issue for some operators, so that is all the more reason to keep it server side.

Edited July 2, 2016 by Alshain

[Lelitu]

11 minutes ago, technicalfool said:

Well that's the thing - as mentioned, shifting the calculations to server side isn't going to prevent griefers, and isn't even necessarily going to make it hard for them. If anything, it could make it easier for a griefer to make a probe-bomb, throw ten thousand parts into orbit (like some do right now with DMP) and slow map view down to a crawl. Only you'd be slowing down the entire server.

Whatever shape KSP Multiplayer takes, it needs to be something that normal people can run on a normal budget. Demanding a dedicated host is not a "normal people" budget. Not only that but it'd also make setting up any kind of official Realms-like host an expensive proposition, even with customers paying for access.

Shifting the calculations server side won't stop all griefing. Never claimed it would.

But it does make it harder to do things like simply ignore collisions, or fuel requirements, because the client no longer has control over collision detection, or burn calculations.

It makes some of the more blatant cheating much, much harder to do. 

[EBOSHI]

mistaken

Edited October 22, 2016 by EBOSHI

[technicalfool]

12 hours ago, EBOSHI said:

A game should surely prepare "a synchronization mode" because Kerbin time is the important information.

What you're describing seems to be almost exactly like the DMP mod's existing "subspace" method of handling timewarp. Players can warp when they like, and if you want to do something together, you can synchronize your time forwards to match a player in your future. In DMP though, you can only sync forwards in time. Allowing players to go back in time causes paradoxes, and is generally to be avoided. In a kind of "what happens if you go back in time and knock the docking port off a station you docked to in the future" kind of way.

[EBOSHI]

mistaken

Edited October 22, 2016 by EBOSHI

[EBOSHI]

mstaken

Edited October 22, 2016 by EBOSHI

[Mjp1050]

Here's an idea for the timewarp problem:

What if time warp was limited to the SOI? For instance, if someone was on Mun, they could timewarp independently of someone orbiting Kerbin, and vise versa. So someone could warp for years in Munar orbit, but only a few hours would pass in Kerbin orbit.

To make things simple, the orbits of the bodies themselves would only be affected by timewarp that was outside the SOI; e.g. someone warping inside Mun's SOI would not affect where Mun was in its orbit, but someone warping in Kerbin's SOI would affect Mun's orbit. This is to keep transfer windows available.

However, someone warping in Minmus's SOI would affect the rotation of Minmus, but not its orbit. This wouldn't apply to tidally locked bodies, as that would be impossible to manage.

Switching SOIs would stop all time warps for both SOIs, and if multiple players were in the same SOI, then warping would have to be agreed upon by all parties within the SOI.

Atmospheres and surfaces would count as different SOIs than orbits, mainly so base-building and station-building do not interfere.

So, here is a chart of what affects what:

Kerbol SOI/ Tracking Station Anything you can do or see while in Kerbol's SOI. Warping here will not affect where moons are in relation to the planets, however.
Planetary SOI Anything you could do or see while focused on a planet in the Map view. Warping here would adjust the rotation of the planet, and the orbit of the moons. Warping here will not affect where the planet is in its orbit, nor will it affect where stations are if the station is orbiting a moon. Nor will it affect anything that is on a surface.
Planet Atmosphere/Surface Warping here only affects ships that are on the surface or in flight. It will not adjust the rotation of the planet.	Moon SOI Anything you could see or do while focused on a moon in the Map view. Warping here will affect the rotation of the moon. Warping here will not affect where the moon is in its orbit. Warping here will not affect anything that is on the surface of the moon.
	Moon Surface Warping here only affects ships that are on the surface or in flight. It will not adjust the rotation of the moon.

 

[Ourworldalpha1]

I understand the argument but what other game let's you build your own Starfighter and you own Star Wars oops I mean Kerbol Wars

Edited July 10, 2016 by Ourworldalpha1

[Firemetal]

Why should KSP be violent? I believe that it would be nice for each player to have their own Solar System.

Each player could invite one another to help them build or fly. This way Ksp stays unique.

 

Edited July 11, 2016 by Firemetal

[Gaarst]

9 minutes ago, Firemetal said:

Why should KSP be violent?

Go tell that to all the folks who only play KSP with BD armory installed (and seeing the number of jet fighter planes in the Spacecraft Exchange, I'm sure they are a lot).

Edited July 10, 2016 by Gaarst

[Firemetal]

4 minutes ago, Gaarst said:

Go tell that to all the folks who only play KSP with BD armory installed (and seeing the number of jet fighter planes in the Spacecraft Exchange, I'm sure they are a lot).

Ha ha. Well they miss the point. Its Kerbal Space program not Kerbal Army Program.  Sure it is sandbox and you can make airplanes, spaceplanes rockets and gantic cruisers (I'm looking at you Shadowzone) but 50% of those guys who just build aesthetically pleasing fighter jets can't make it to the Mun and back.

 

[Alshain]

4 hours ago, Ourworldalpha1 said:

I understand the argument but what other game let's you build your own Starfighter and you own Star Wars oops I mean Kerbol Wars

StarMade

SpaceEngineers

(This does not invalidate the argument for multiplayer, but since you asked.... )

Edited July 10, 2016 by Alshain

[EBOSHI]

mistaken

Edited October 22, 2016 by EBOSHI

[dsonbill]

On 7/1/2016 at 11:18 PM, AlamoVampire said:

You cannot possibly plan or control every aspect of what multiplayer is. This is one game where Multiplayer is just a flat out bad idea. The game is too young. You can sabre rattle all you like, but, the fact remains, you cannot have factored in every single possibility no matter HOW slim of a chance it has at happening. I had a funky experience a few days ago with time warp and 2 vessels in a SINGLE player environ that showed me how bad it can be in a multiplayer environ. I set a ship to a reentry course via mechjeb and swapped to a different ship. The remote controlled ship on autopilot kicked on autowarp and I was stuck unable to ANYTHING while the other ship warped.

What about mods? What about part count? Player count? Trolls? the list is as long and varied as there are stars in our known universe and beyond. When this game is 10 or 15 years old, then MAYBE multiplayer can be stock. But now? With soo many issues that NEED PRIORITY, like crashing, or the resurgence of this cannot deploy while stowed bs <as examples that are still  priority> must be dealt with FIRST. Not to mention the risk to single player content. No, multiplayer isnt ready to be a thing yet. Not now, not in the near future. Sorry, but brutal truth is brutal truth.

Sorry for the incredibly late response.

Most of what you're mentioning here has, in fact, been solved in DMP, and we're still client-authoritative. It even fixes some stock bugs in order to function properly. Having a fully server-authoritative system will solve most - if not all - cheating possibilities, raise the server requirements to ungodly levels, and create an overall better experience, which is what Squad should be attempting. Additionally, the movement to a new system should solve bugs inherently - most things will get a look-over and re-implementation in the move to a server-authoritative system.

As for part count, the new version of KSP should help with that, but I haven't really tested it with DMP. Remember though that slowness with DMP won't translate over to a server-authoritative system, because things would work completely differently - not just with the multiplayer code, but with how the client itself works. Ships also probably need some kind of simplification for graphical representation and placement or something. Maybe individual parts are only simulated on the server-end?

DMP does work with mods, usually with no changes at all. Some will require server-side plugins, but not many, and there's only a handful of plugins in existence. In the move to a server-authoritative system, mods would need to be made a bit differently to work with multiplayer, and you'd need client mods most likely too, but things would be a lot more controllable on the server side. The new system would need to have objects and game-code added on the server side as well for simulation. Single player mods would also likely need major changes depending on their functionality, as Squad would likely make the single player game work on an internal server.

DMP's main issue is that we do the position and rotation stuff really weird. This could actually probably be solved with a combination of a bit of simplification/allowing the client to lie a little bit more, and increasing the update frequency. Honestly, all it would probably take to get DMP working nicely would be for me or Chris to try at that problem for a day or two. We're just burned out.

Edited July 22, 2016 by dsonbill

[Jatwaa]

The problems people see with multiplayer are still the same ones always mentioned. As far as warp goes, I am all for one persons warp being another persons teleport then resync. Would it be chaotic, possibly. Basically, it's the only issue I foresee being a problem. For lift offs, easy, Kerbal Konstructs shows you can add runways and launchpads. Also, for mods, most games that have mods have a mod checker. DMP does it well. Basically...I can't see many reasons why not to have MP options. Only thing they need is to sort timewarp and let modders handle the rest if they don't have time.

[dsonbill]

2 minutes ago, Jatwaa said:

The problems people see with multiplayer are still the same ones always mentioned. As far as warp goes, I am all for one persons warp being another persons teleport then resync. Would it be chaotic, possibly. Basically, it's the only issue I foresee being a problem. For lift offs, easy, Kerbal Konstructs shows you can add runways and launchpads. Also, for mods, most games that have mods have a mod checker. DMP does it well. Basically...I can't see many reasons why not to have MP options. Only thing they need is to sort timewarp and let modders handle the rest if they don't have time.

DMP actually changed the way warp sync works. Instead of simply warping to their time in the future, your vessel doesn't get changed at all, and moves to the new time slot. Since there's no real advantage to doing this, it seemed appropriate.

[Jatwaa]

I have the new DMP installed but havent tried it in a bit. Perhaps it is time to revisit with a centralized server...

[AlamoVampire]

DMP can't plan for everything. And if its as good as y'all suggest leave mp to that. No reason to have squad waste time on it.

[Wjolcz]

5 minutes ago, AlamoVampire said:

DMP can't plan for everything. And if its as good as y'all suggest leave mp to that. No reason to have squad waste time on it.

I say drop the idea of MP for KSP and leave it for a sequel (or rather prequel) with airplanes only. Seriously. MP is not worth it if it's going to be in the KSP we have now.

It's more fun to race, fight and fly with others than warping and syncing all the time.