HarvesteR shares his thoughts on KSP2

[moeggz]

1 hour ago, Scarecrow71 said:

So I watched the 2 minute section (or however long it was) in the video...and I'm not sure what to make of it.  HarvesteR sounds like someone who created something and then left, and now doesn't know if he has an opinion or not about what is happening to the sequel.  I've been pretty outspoken about what I think of KSP2, but I don't get the feeling from that 2 minute section that HarvesteR is disappointed or at a loss of what to say.  Then again, none of us are inside his mind, so unless he comes out directly and says what he's thinking, this is all merely speculation.

And now I gotta go check out Kitbash and see what that's all about.

I can see that point of view. To be fair, I should have been more clear earlier that I don’t think he hates KSP2, wants it to fail or anything like that. So I read him as more critical than you, but far from a “hater perspective.” The lack of any real praise is telling, to me at least.
 

The main point I’m trying to make is that he has a system that solves wobble that he was unable to implement in KSP1 but has in his new game. To me, it’s not a jump that he would’ve solved wobble in KSP2 in a similar way to the way Kitbash has. 

And yeah I was disinterested in his last game but this game is making me nostalgic of old PC battlebots games and I think I will have to check this one out.

 

@regex I have remained respectful of your opinion and discussed with you without mocking you or your intelligence. If you want to discuss with me in a respectful manner than please do.

If my views are so ridiculous to you that you feel there is very little discussion to be had than you are welcome to ignore my profile so that I do not upset you anymore, as that is not my intention.

[Lisias]

22 hours ago, RocketRockington said:

That HarvestR made one bad call, therefore all his opinions should never be listened to? 

He probably did a lot of bad calls, and if he learnt from it, we call it "experience". And if he's able to use this experience into new projects with success, we call it "wisdom".

This doesn't means that he will not make new mistakes - only that he will not repeat the old ones, and that he will probably make less new mistakes than someone else.

 

16 hours ago, Lyneira said:

The gameplay footage demonstrates that this implementation works and things break off with no noticable framerate hits when gun shots or obstacles are hit. During the car battle segment, there was a tendency for car-car collisions to cause both cars to break up into lots of little pieces. I'm sure the level of "disintegrate into thousand pieces" from collisions could be tweaked, though maybe this is a bit of kraken showing through here as well.

Yes, but we need some context: that model was fine tuned to be used for a few kilograms "vessels".

KSP2 will need to handle multi-stage, multi-tons and docking capable crafts. It's a problem some orders of magnitude higher than what Kit Bash had to handle.

Additionally, we still have the vessel's "internal physics". How it will be modelled? Using Graphs (and so how to avoid the terrible pitfalls - do a search for the Salesman Travelling Problem)? Or the same Tree used on KSP¹ and probably on KSP2 (and then we will have the very same problems we have now, but encapsulated on a second physics engine)?

It's interesting to note that @HarvesteR avoided to do a direct answer about KSP2. He wisely avoided giving an clear opinion without knowing exactly what's inside KSP2 first, besides I'm pretty sure he have a very good idea about what is going there.

That said, and to avoid another unfortunate flame-fest, I'm not saying that the Kit Bash solution will not work on KSP2, I'm saying it will not be enough, and some of the current pitfalls will be reproduced on it the same, because it will be the same problems with similar solutions - unless we decide to dumb down KSP2's physics models to a RC model level (that it's perfectly fine for Kit Bash, but I doubt people are buying KSP2 to fly RC balsa airplanes on Laythe).

 

13 hours ago, regex said:

Unfortunately it just outlines how little discussion there is to be had with you here because you consider Intercept, a pretty big team, to be so incompetent they can't figure this out on their own, to their own satisfaction.

A team is so smart (or stupid) as the guy paying their bills. Without privileged information, we can only speculate about but there must be a reason for that layoff that happened recently on P.D.

What follows is pure speculation based on my own experience on the software development business:

Team Leader: "We are going to rewrite the core business of the Company's Flag Ship, as maintaining it is consuming more man/hours than introducing new fetures!"

3 months later, the new components are sent to Q/A and, as any new component, get a lot of issues opened.

Manager sees that as a regression, because the core components are working on PRD as is, and shut downs the [current] development, diverging developers from an other teams to code that new features, on the most "9 pregnant women will give birth 1 child a month" mentality as possible.

9 months later, no new features could be finished and all the old ones that were working started to misbehave due the huge technical debits that were accumulated not only from the older core business components, but [also] from the brute force injection of that components into the new code base after ripping off that new code that the Q/A guys flagged as regression.

Looks familiar to you?

 

11 hours ago, moeggz said:

The lack of any real praise is telling, to me at least.

Exactly. The questions he didn't answered sounded a lot louder than the ones he did. Silence can be deafening loud.

 

Edited July 26, 2023 by Lisias
Kraken damned grammars.

[regex]

27 minutes ago, Lisias said:

Looks familiar to you?

Doesn't look even remotely the same. If anything I'd expect them to contact Unity support if they wanted help rather than HarvesteR, a guy working on a game of his own and not related to Intercept in any real way, who has no idea what the current codebase looks like (or likely even what the current KSP1 codebase looks like).

[Alexoff]

19 hours ago, moeggz said:

I will get claims that I’m lying if I give any text other than a word for word transcript.

You're lying! (The world collapsed due to a logical contradiction)

 

Comments on the KSP2 video have changed a lot since March

[moeggz]

1 minute ago, Alexoff said:

You're lying! (The world collapsed due to a logical contradiction)

 

Comments on the KSP2 video have changed a lot since March

I’m not prepared for paradoxes today 

And yeah. And my opinion I think has followed the tide of most. I was very hyped and positive at the beginning and here we are now.

[Alexoff]

HarvesteR is visibly struggling to find words, apparently deep in his soul he wants to say something else, but still tries to keep up appearances.

[Vanamonde]

41 minutes ago, Alexoff said:

HarvesteR is visibly struggling to find words, apparently deep in his soul he wants to say something else, but still tries to keep up appearances.

Perhaps he is about to burst into song. 

[Spicat]

Also, to add to the "They should take advices from Harvester", you know that former SQUAD employees are working at Intercept? This is a much better solution to ask them because they know ksp2 pipeline and code.

[Alexoff]

19 minutes ago, Spicat said:

Also, to add to the "They should take advices from Harvester", you know that former SQUAD employees are working at Intercept? This is a much better solution to ask them because they know ksp2 pipeline and code.

It would be great if Harvester was one of those employees

[jost]

5 minutes ago, Alexoff said:

It would be great if Harvester was one of those employees

He wasn't even a Squad employee since 2016 (!):

If it would have been impossible to develop KSP1 right without him the last version would have been 1.1: 
https://wiki.kerbalspaceprogram.com/wiki/Version_history 

Sorry, not sorry: Many of the things which really made KSP1 great were implemented after he left Squad.
And to make it clear: This is not to trample on HarvesteR: His idea of a rocket simulator was a stroke of genius and I have a lot of respect for his work. 
But to claim that he was sole responsible why KSP1 was a better game than KSP2   seems slightly exaggerated to me and unfair to the other KSP1 and KSP2 developers.   
And I'm not even sure whether KSP2 will reach it goals in the end. 

[razark]

Why should they ask him?  What they need to do is hire one of the folks around here that seem to have psychic powers and know everything about everything.

[Alexoff]

49 minutes ago, jost said:

If it would have been impossible to develop KSP1 right without him the last version would have been 1.1

In 2016, many left, and as far as I remember, not because of the good atmosphere in the company.

51 minutes ago, jost said:

Many of the things which really made KSP1 great were implemented after he left Squad

What is this, for example? After 1.2, there were small steps forward, various improvements, new parts, but I would not say that 1.2, released a few months after his departure, is so different from 1.12. And the KSP had the maximum hype in those days, judging by the views of the videos on YouTube.

I would also add who should not be involved in the development of KSP2, but I will try, like Felipe, to start singing some popular song in my head

[asmi]

1 hour ago, Alexoff said:

It would be great if Harvester was one of those employees

Absolutely NOT in my opinion. He made so many wrong calls, that KSP1 started progressing in a right direction only after he left. if you talk off-record over a beer or two to any of late KSP devs who has seen Harverster's late days in the team, you won't hear many good words about him. That's not to take away his achievement of the very creation of KSP, but at some point things turned for the worse. That just shows once again importance of leaving at just the right time before your star has finally set for good.

[Klapaucius]

What struck me in the video: How much more interesting the terrain in Kitbash is than KSP2.  Granted, I have not flown over the whole surface of Kerbin, but in general the terrain is just not as engaging.  That small bay in Kitbash had so many interesting elements--even if you remove the people, highways and houses, you still have a lot of canyons, interesting waterways etc.

I realize that putting that level of detail on whole planets and moons is probably too much, but it would be nice to have at least a few areas where the features get more attention.

[PDCWolf]

Gotta love how most of a certain viewpoint is comprised of people wanting to have an opinion on HarvesteR's sayings without watching the video.
 

Quote

 

Matt: You know I can't not ask you what your thoughts on KSP2 are, as the person who created KSP1.

Harvester: It's a weird feeling I have to say, I've been looking at it and it looks pretty, it looks good. I've been following on the release and how that's been going. I know what they've been going through, we went through a lot of that back in the day too. It's hard, but yeah, I don't know if uhh... I don't know how to feel about it is the truth. KSP It's like having a child, and that child has grown up, has moved out of the house and it is hanging out with this new people you don't really know but uhh, I'm still proud of "him" I guess.

Matt: Because you left KSP1 before that game was fully done didn't you.

HarvesteR: I did as we delivered 1.0 back in 2016, when I felt everything I wanted to see in game was there in some degree, and the other guys still had ideas they wanted to see implemented before saying they were done. Uhh, but yeah most of the original team, we left back then, after 1.0.

Matt: Final KSP2 question, because this is a kitbash video, but I've got one more. A lot of people are worried about KSP2 and its future, are you worried about KSP2's future? Because, do you think what you've seen so far is a solid enough foundation and think that, as they get to work it'll be the game that it was promised to be.

HarvesteR: Honestly, I don't know. I haven't heard from any of the devs of KSP2, at any point, so I know as much as you do I think, or maybe you know more as you've been looking at it more often. I feel like.. yeah, I would like to see it do well, but yeah, to say that I fully believe that it's gonna go smooth is guesswork in my part. I guess we'll see, cause uh, yeah, you know, uhh, how that kind of thing can go.

Matt: It's nice getting your thoughts not only as a gamedev but also as... KSP would have not existed without you.

HarvesteR: It's a weird feeling now that it's out of my hands. It's weird to see, it became it's own thing right? I may have started, but KSP is like a living thing at this point, that's got its own life cycle. Very much like having a child that left the house and is doing its own thing, it's weird, and you worry about it.

Regarding wobble

HarvesteR: I don't know if you've noticed, but in Kitbash we don't have wobblyness. We've moved on from... instead of making every part its own rigidbody like it was in KSP, here we have got a single rigid body for the whole vehicle, but we have an internal simulation happening to calculate  internal stresses. This is something I thought about doing since the early days of KSP, and here we got a chance to try it out. Now, if I shamelessly aim and shoot at you and you see things break, when you take a shot or get hit, or hit something, there's this internal physics lattice that simulates stresses. Kinda how in KSP the parts do that by themselves, but it is separate and runs in its own reference frame, it's completely turn-off-able. If we turn the influence of that to 0 we get fully rigid physics. Imagine you with your 1000 part vehicle meeting your friend with 1000 parts vehicle, that has to perform at least to some degree, so uhm, we're able to turn this off, this level of physics simulation for vehicles that aren't yours, cause they're being simulated on other people's ends, so it's gonna happen anyway.

 

Now, inferring from hearing the audio: HarvesteR at no point said, or felt like he would say, or hinted at saying that KSP2 is good, in a good place, has future, etc. What he thinks is impossible to know, but he at no point casted anything but doubt about KSP2.

3 hours ago, regex said:

Doesn't look even remotely the same. If anything I'd expect them to contact Unity support if they wanted help rather than HarvesteR, a guy working on a game of his own and not related to Intercept in any real way, who has no idea what the current codebase looks like (or likely even what the current KSP1 codebase looks like).

Ah yes, the KSP2 codebase, very different, much new.

[Superfluous J]

1 hour ago, Alexoff said:

What is this, for example? After 1.2, there were small steps forward, various improvements, new parts, but I would not say that 1.2, released a few months after his departure, is so different from 1.12. And the KSP had the maximum hype in those days, judging by the views of the videos on YouTube.

Let's see, 1.2 had the comms network and kerbnet. 1.3 added localization and I can understand an English speaker (I'm one myself) not caring about that too much but it's a big deal to a lot of people. 1.4 didn't have a big groundbreaking change, you've got me there. 1.5 revamped the burn timer which seems small but I remember it being a big improvement. 1.6 added delta-v information which IMO is the biggest improvement in the game since the atmosphere rework in 1.0. The next few versions didn't add much but that ended with 1.11's construction mode which again is a huge change. Then 1.12 itself had the maneuver node creator.

So no, 1.12 was not all that different from 1.2 if you spoke English and had Remotetech, Better Burn Timer, Kerbal Engineer, Kerbal Attachment System, Kerbal Inventory System, and Precise Node installed.

[MechBFP]

34 minutes ago, PDCWolf said:

.What he thinks is impossible to know, but he at no point casted anything but doubt about KSP2.

it is sad that saying nothing immediately implies doubt for you.

[RocketRockington]

13 minutes ago, MechBFP said:

it is sad that saying nothing immediately implies doubt for you.

 Nothing.

[MechBFP]

3 minutes ago, RocketRockington said:

 Nothing.

Sorry, actually he said the game was pretty, so not nothing even.

[RocketRockington]

2 hours ago, MechBFP said:

Sorry, actually he said the game was pretty, so not nothing even.

[Strawberry]

I don't see the point in everyone in this thread trying to assert the true meaning of harvesters words when he's pretty explicit on his feelings? Like, his feelings are complexed, he understands the struggles that the game is going through and think it has potential but he's also worried about it because KSP was a big part of his life and there's a lot he doesn't know. Trying to analysis how his resonance in his stuttering means that Nate Simpson is holding him at gun point behind the screen and he's upset by this is just kinda missing the forest for the bushes. He makes it pretty clear himself that even he isn't entirely certain on his own feelings.

[Lisias]

11 hours ago, asmi said:

Absolutely NOT in my opinion. He made so many wrong calls, that KSP1 started progressing in a right direction only after he left.

You see, this is the main reason I like to talk to people that previously maintained/create any code I maintain (both on hobby as in Day Job®).

There're only a few ways on doing things right, but a huge amount of other ways to do things wrong. The more I know about the later, better may chances on the former.  

However… My personal experience on modding KSP (and I have some reasonable knowledge since 1.2.2, as I choose to keep backwards compatibility on everything I do), is not soooo kind about the team (or at least, part of them) that took over after HarvesteR departure. There're a lot of huge, huge mistakes and bugs lingering there for almost a decade, and they failed to correct fix them - worst, they created worst bugs and unnecessary collateral effects by naively trying to tackle down some of that bugs. There're a few decisions on the thing that really made me mad over the years.

I'm not complaining about bugs happening - bugs are unavoidable consequences on doing new things. I'm complaining about they not being diagnosed and/or fixed.

 

11 hours ago, Klapaucius said:

What struck me in the video: How much more interesting the terrain in Kitbash is than KSP2.  Granted, I have not flown over the whole surface of Kerbin, but in general the terrain is just not as engaging.  That small bay in Kitbash had so many interesting elements--even if you remove the people, highways and houses, you still have a lot of canyons, interesting waterways etc.

Other thing that caught my attention is he explaining how he found the need of doing a more detailed city to play at - since the crafts would be essentially toy sized, things around should be way more detailed than if the crafts were normal sized. People are going to look at bricks, not at walls while playing!

Another thing that I found pretty clever: the use of Cities: Skylines as a Scene Editor for the KitBasch's city, taking advantage of an add'on that implemented Open Street Map I think. I'll buy this game as soon as it is on sale (and the thing works on MacOS!)

 

6 hours ago, Strawberry said:

Trying to analysis how his resonance in his stuttering means that Nate Simpson is holding him at gun point behind the screen and he's upset by this is just kinda missing the forest for the bushes. He makes it pretty clear himself that even he isn't entirely certain on his own feelings.

Take in consideration that he signed a (probably pretty nasty) NDA when he left, and I'm almost sure that NDA is still in force. So, there's a chance he was feeling exactly like that (besides I doubt Nate would be the gunner).

Edited July 26, 2023 by Lisias
Tyop! Surprised?

[Lyneira]

11 hours ago, Lisias said:

Yes, but we need some context: that model was fine tuned to be used for a few kilograms "vessels".

KSP2 will need to handle multi-stage, multi-tons and docking capable crafts. It's a problem some orders of magnitude higher than what Kit Bash had to handle.

Additionally, we still have the vessel's "internal physics". How it will be modelled? Using Graphs (and so how to avoid the terrible pitfalls - do a search for the Salesman Travelling Problem)? Or the same Tree used on KSP¹ and probably on KSP2 (and then we will have the very same problems we have now, but encapsulated on a second physics engine)?

It's interesting to note that @HarvesteR avoided to do a direct answer about KSP2. He wisely avoided giving an clear opinion without knowing exactly what's inside KSP2 first, besides I'm pretty sure he have a very good idea about what is going there.

That said, and to avoid another unfortunate flame-fest, I'm not saying that the Kit Bash solution will not work on KSP2, I'm saying it will not be enough, and some of the current pitfalls will be reproduced on it the same, because it will be the same problems with similar solutions - unless we decide to dumb down KSP2's physics models to a RC model level (that it's perfectly fine for Kit Bash, but I doubt people are buying KSP2 to fly RC balsa airplanes on Laythe).

The parts may be bigger and the mass higher, but the math of the physics works all the same on it regardless, and it will only need to act on a single rigid body (until something breaks). There will certainly be some differences, like KSP having staging and docking (counterpoint: Parts breaking off in Kit Bash is already a working example of vessel separation), but nothing that makes KSP2's physics an "order of magnitude" more difficult for a single-rigidbody model. If anything, KSP2's current spring joint model is the one that adds the order of magnitude more problems to deal with.

As for the internal physics, Kit Bash shows it can be done with good performance. There is another thread on these very forums with a proposal by @Kavaeric that is very similar to what is being demonstrated here:

You mention "dumbing down KSP2's physics model to a RC model level," implying that because an RC model craft simulator uses it, it must be dumb. The model may be simpler (because it only has one rigidbody until stuff breaks) but it produces accurate results with no side effects. As a "not-dumb" example, I would say flight simulators have more in common with Kit Bash's model than they do KSP2, because they do not simulate an airplane as a collection of wobbly parts connected by springs but rather as one whole body.

[Periple]

3 minutes ago, Lyneira said:

The parts may be bigger and the mass higher, but the math of the physics works all the same on it regardless

You would think that, but in reality game physics will have loads of fudge factors that would make your eyes water if you knew about them. They're added to make the game feel right, realism be damned. Link in TotK for example weighs about as much as 10 apples!

[Lisias]

1 hour ago, Lyneira said:

The parts may be bigger and the mass higher, but the math of the physics works all the same on it regardless, and it will only need to act on a single rigid body (until something breaks). 

No, you are unfortunately misinformed. The bigger (or the smaller) the magnitudes of the attributes involved on the computations, the more likely they will end up giving you Infinites or Non-A-Number in the middle of your physics engine. Believe me on this one, I was one of the guys that diagnosed KSP's CTD due physics problems back in the 1.4 era.

You can really never get rid of these pesky problems, but you can mitigate them by using binary representations with better precisions (i.e., doubles are sensibly less prone to these computations mishaps than floats), but since floats are usually adopted by physics engines to allow faster net-computations inside the silicon, the sad true is that we need to carefully trim down our models to this problem, instead to the thing from Real Life© you want to simulate.

 

1 hour ago, Lyneira said:

(counterpoint: Parts breaking off in Kit Bash is already a working example of vessel separation),

If and only if each new part injected into the World are fully capable "vessels", and not only statics being thrown into the game engine.

I'm not saying you are wrong, I'm saying you don't know if you are right about the matter without more information.

 

1 hour ago, Lyneira said:

but nothing that makes KSP2's physics an "order of magnitude" more difficult for a single-rigidbody model.

You are misguided. Vessels on KSP use stages, do docking, hold cargo and fuel (and consume it) and a lot of things that will force you to decompose or compose different vessels into a single high level entity. How do you think we should compose and decompose these vessels? How do you think they should be implemented on the Physics Engine?

Kit Bash don't have to cope with any of that.

 

1 hour ago, Lyneira said:

As for the internal physics, Kit Bash shows it can be done with good performance. There is another thread on these very forums with a proposal by @Kavaeric that is very similar to what is being demonstrated here:

I suggest you read my posts there, I'm one of the original proposers of the idea.

 

1 hour ago, Lyneira said:

You mention "dumbing down KSP2's physics model to a RC model level," implying that because an RC model craft simulator uses it, it must be dumb. 

The thing bends under stress? Misbehaves aerodynamically if you pumps fuel to the wrong fuel tank? Things just blow up out of the blue when abused, or they visually warns you that you are abusing the frame due aerodynamic forces?

If no, it's dumbed down.

A RC airplane model, in essence, is a dumbed down version of the original craft.  If you don't need a pilot brevet to pilot the thing, it's because the thing was dumbed down so you don't need to take years of piloting classes to pilot it.  

 

1 hour ago, Lyneira said:

because they do not simulate an airplane as a collection of wobbly parts connected by springs but rather as one whole body.

Because they do not simulate crafts made by the user. Every craft on MS Flight Simulator and X Plane were custom made by software developers, carefully trimmed and optimised to behave as specified - some of them costing more than the game itself.

You are misunderstanding the problem with the effects. The excessive wobbling on KSP2 is not the root problem, it's the most obvious manifestation of the problem.

We had replaced spars, spines and structural reinforcements for attachment points on a tree data structure, but we still want the thing to behave like it had spars, spines and structural reinforcements. Some abstractions will leak, some compromises are unavoidable.

You move the current solution for a "lattice" physics engine for each vessel, and you will still have the same problem happening but with a different collateral effect: heavy crafts will still behave weirdly, the wobbling will be replaced by some other undesired consequence.

I want my vessels to bend and ultimately fail under excessive stress - otherwise, what would be the point of KSP2? Being a gold platted Juno rip off?

Edited July 26, 2023 by Lisias
Tyops, but you already knew it.