Resolved Is DOTS being abandoned?

It really isn't "bigger team" stuff. I've been using ECS and task scheduling in Rust for my own engine for some time now and it actually severely decreases workload when implemented right. At the moment, I handle everything with Legion but I'm currently implementing my own system that has more robust task scheduling and system management. It has actually been a fundamentally massive timesaver because once you understand how data flow is handled, it becomes much easier to debug and refactor.

Data oriented frameworks are pretty great, but having used Legion for a fair amount of time now, the clear thing is more that Unity's implementation has been pretty junky.

 

Sure you can use it but Unity depends on a lot of things for games like animation, audio, physics, navigation, networking, particles and all that jazz. It's doable but IMHO quite a bit more work than mono land for the same (lower performing) game due to missing functionality. Tis why you'd want a team around to plug all those gaps.

 

This is precisely the problem, but the problem there is that this is a problem we have been telling Unity about since the dawn of the DOTS packages. Their lack of interoperability with Unity's core systems is the biggest problem it has, followed shortly by the incredibly lacking interface for working with components and systems, then followed by the boilerplate overload, itself followed by the extremely poor documentation.

But the problem with all that stuff is that this is kinda endemic to loads of parts of the engine, not just DOTS. Unity's ECS and data oriented in general implementation is poor and largely confusing and that's Unity's fault.

That's what I mean when I say it's junky. It's been years and it feels like we're still ages off from anything practical.

 

The simplicity of Legion(Rust) and flecs(C++) makes DOD so interesting. Note that all the DOD I learnt is with Unity. I don't understand the need to defend the "string them along" behavior of Unity anymore. You have described my anger with Unity very well.

 

They can take as long as they want sure, but the issue is with their transparency and not the time taken. Acknowledging that the communcation ball was dropped and then continue to have no communication just feels like they don't care about us (its even a red flag in relationships). And the "updates are coming soon" promises are such an overused trope in this industry that it just makes developers jaded, I just want to see what is actually being worked on, and perhaps this is what other people also want.

 

I've sort of given up hope with DOTS...there are many ways to defend it and some are definitely valid, but its just been too long (like 4-5 years?). And we are far away from a remotely usable version. Seriously, this is just taking too long.

Let's be honest, DOTS will never be ready for another 2-3 years easy.
Was DOTS always meant to take this long?
I'm just sitting down here thinking...

I wonder how long it will take to actually finish DOTS. A decade perhaps...

 

I find this to be an interesting statement as multiple games have been released with it already so it's clearly usable.

Not saying progress hasn't been slower than desired and there are definitely a lot of features that I wish it had.

 

They do not care about DOTS early adopters. If they did, it would be obvious by their actions.
Unity has become a large public company, they don't need us anymore and that's okay.
Not worth losing sleep over

If DOTS 1.0 becomes a huge success, everyone will want to use it anyway.
Even after this year, it seems people are still very excited about DOTS' possibilities.
After all, Unity brought us the GameObjects/MonoBehaviour workflow.
Maybe DOTS 1.0 will be just as awesome.

That being said, caring goes goes both ways.
It seems the days of people defending their beloved Unity engine are long gone.
Surely many people here would happily switch to a better & more open engine.

 

Why you guys treat DOTS as single package feature? Of course it won't be ready, as DOTS is a combination of nearly dozen of individual features. Burst and jobs are fully usable for long time. ECS is still in preview, but successfully used by many devopers, in contexts of DOTS.

Reading such comments, it makes me feel again, even recently mentioned, people are still confusing what is DOTS, and thinking about ECS package by default.

Dots includes animations, physics, audio and many more packages. So is obvious, these will not be ready within year or even 2.

People who use DOTS are great problem solvers. Others are just waiting for some magic tool, seems thinking, that will gain performance with simple toggle. I can assure you, it will never be the case. There is no shortcuts, how to write highly performant code.

Learning current basics packages of DOTS to satisfactory level, will require year of practicing. It is either jumping on the wagen and start learning, or not expecting DOTS that it will be complete any tim soon.

It would be silly to postpone making game/application, just because some engine feature is not ready.

Other than that yes, community officials do lacking in communication.

 

When we complain about current features, the answer is often, "oh that will be fixed / improved with DOTS".

Audio things are outdated and kinda suck ass? They refuse to touch them. Wait for the DOTS stuff.
Animation tools are slow and / or outdated? Wait for the DOTS equivalent.
And didn't the Input System (which released in a half baked state), stop getting updates for about a year, while they were working on the DOTS prototype?

Everything in Unity is on pause / maintenance mode until "DOTS is ready" whatever that means. You can't blame people for thinking it's a monolithic thing when Unity does the same and they also present it as the magic solution to all our complaints, while in the meantime we are more or less paying to use a decade old tech they refuse to touch.

 

Cool, it's been nearly five years now and it's not just ECS/DOTS that's the problem though. System interoperability in Unity is honestly absolutely dire since they decided to start targeting packages especially, in fact. For instance, UIToolkit or UIElements or whatever it's called now? Doesn't work with the new input system despite both of them being either verified or built into the engine. In fact, neither does Cinemachine unless you follow a third-party guide!

DOTS being ECS and Jobs and Burst doesn't really change a single thing being said here because ECS really does live at the absolute core of the Data Oriented Tech Stack. ECS itself is data oriented design in architectural form. This is a needless "ummm actually" for a system that still requires you roll your own skinned mesh support, a ridiculously common thing in 3D games, with the only Unity-lead solution being a single example that does the Unity Special: only works with a single version of ECS and doesn't work very well there either.

And these have been problems for years that they just don't talk about and we get next to no progress updates on. It's not just communication, but the rate at which updates happen in the first place and Unity's long-standing history of just shrugging and giving up, from an end user perspective.

 

We all know how things are with Unity. So I don't know what is anyone expecting from DOTS, to be any different. My best hope is that we get entities package release 1.0 sometime in 2022, and that's being optimistic.

Anything else will be just dream. Our DOTS community hash showed, we are are able to make better solutions, to almost anything Unity DOTS made so far. Mostly tailored for individual projects needs. And this is our fault, showing, we don't need Unity solutions. Chances are, we may never see them get finished. So I personally never would take a breath, to all these fancy semi techs, made by Unity.

I personally only expect from DOTS, that entities package is improved. Anything else is, meh. At least from what I see. For anything else, I just watch space for Unity 2022, hoping DOTS core packeges will be fixed and working as advertised.

 

I do remember this being a thing back in 2019. But I'm pretty sure all those teams have backtracked by then. Could you reference where someone from Unity has said something along these lines within this last year?

As someone who has written a custom skinned mesh implementation but also made the test code work with the Hybrid Renderer's skinned mesh implementation for comparison, what are you talking about? Are you conflating skinned meshes with the DOTS animation package?

Honestly, if everyone who was complaining instead spent their time collaborating on these community solutions, we would have our own community 1.0 sooner than Unity.

 

Unity themselves have been calling the DOTS stuff + SRP "New Unity". People are expecting that at some point the editor and all tools will be built around DOTS and that our grand kids

I think the idea was to make a Unreal 3 to Unreal 4 style upgrade and break and build up from a new codebase, which isn't a bad idea, but it's absolutely incompatible with the subscription model and release schedule they chose.

Instead we're left with multiple yearly releases where the old features "are-not-really-abandoned-really-but-they-are" and the new features are "you-can-use-these-to-do-stuff-it's-not-too-early-but-also-don't".

IMO, they should have kept a small team improving the Unity as it was, and then launched a Unity NEXT, which would be experimental for a while, but that we could maybe install once a in a while and get an idea of what they are building towards.

The current Unity, with mix of old and new and where none of it feels at home, feels like it's going in circles (or rather a downwards spiral).

 

Have they said something in the last year that contradicts what they have previously said?

I'm not seeing any improvements to audio. Are you?
Or animation?

The only one that is picked up again is the input system.

 

Yes they have. That started when they announced NetCode for GameObjects back in the summer of last year, which they followed through on.

I've seen more work done for GameObjects than for DOTS. With that said, I've mostly just seen bugfixes. So I think Unity is just struggling with audio in general.

Animation rigging?

 

You mean the netcode they are working on since 2018 or potentially even earlier than that? That shows them changing their priorities?

I hope you understand that this is more or less an excuse.

You mean the package they have been working on for years (wasn't the preview release in 2019?) and that finally released, although it totally isn't ready (it didn't work with humanoid rigs last time I tried it, which, uhhhh...), while they recently they cancelled / postponed everything else related to animation to FOCUS ON DOTS? ( https://forum.unity.com/threads/animation-status-update-q2-2021.1125482/ and https://forum.unity.com/threads/animation-status-update-q4-2020.1023445/ )

Wake me up when they provide a replacement for Legacy Animation, an Animator that doesn't suck, and a Timeline that isn't designed by aliens.

 

Unity DOTS team didn't focus on individual features pas year, from what I gather, on verious post of 2020 and early 2021.
But mostly on fixing things in Unity core itself.
That includes suposidly compatibility with core DOTS packages.
But 2022 is yet to prove, if that is indeed the case.
My understanding is, Unity DOTS development was in conflict with Unity core. Which had to be addressed internally. Also DOTS individual teams seems worked independly on various features, causing conflicts on the inbetweene, and hindering general progress.

In my understanding, all that lead general DOTS progress to stand still, untill is rectified and fixed.

However, how accurate is my statement, would need to be validated. Yet, I am not spending time on searching individual posts to do so. So yes, please free to comment on any ot that.

 

I mean there is an alpha of 2022 out, not much has changed there.

In any case, I am not arguing they have abandoned DOTS, I am arguing they have abandoned everything else, while they are fighting to cram a DOTS shaped peg, into a Monobehaviour shaped Editor, which takes more time than it should, and in the meantime the users don't really get much benefit.

 

Nope. It was a different one. Shortly after that announcement, they adopted MLAPI and morphed it into something more integrated and usable, complete with simple demo projects. That all happened in about 6 months, which was pretty impressive. But prior to that, they were a bit lost and had all their eggs on DOTS.

I hope you understand that there is a difference between the following statements:
"Unity is spending all their time building new features for DOTS and not working on GameObjects features."
"Unity is not building anything useful for me for GameObjects."

Unity tends to work on lots of useless things, and they take forever to finish them. I disagree with the first statement, but agree with the second for many things. And audio falls into this category. They aren't doing anything new for audio, neither for Game Objects nor for Entities.

I read those posts, and have no idea how you came to that conclusion. They say they are not making changes in order to keep the animation tech stable because it has served many games well. And instead they are building new technologies on top of it like animation rigging. There are probably some engine-level animation programmers that don't have a lot of work left to do and are helping out with DOTS.

 

https://github.com/Unity-Technologies/Graphics/pull/5842
https://github.com/Unity-Technologies/Graphics/pull/6071
& others
Seems like DOTS is planned for 2022.2, maybe 2022.1, just going off based on pulls

 

My only wish for DOTS 1.0 (or Entities 1.0, whichever comes first) is dynamic entity chunk size. Fixed 16KB is a complete waste for a persistent gamestate singleton. Yes, there's Latios excellent package extension but dynamic size would also greatly assist in manual vectorization in Burst. No need to worry about handing remainders when you set array size (entity count) to a multiple of 4 manually.

 

My wish is for them to release DOTS 1.0 and then instantly abandon it (to chase a new market “Unity for Asshats”) like they’ve done with all their features that hit 1.0, but have people disagree with me when I point that out because some team half heartedly adds half a feature while hiding from management.

 

Then all your avatar misses is a nice shapely pair of buttocks. That said, I bet if Unity surprises us all, it will be either:

a) a really welcome surprise - we see the final form of megacity style data management and rendering
b) we see great streaming going on, and it's kinda solved big scale

Or maybe come bullet points you can think of that are positive.

OR

Unity is bad! they are the new daz but with more money!

 

c) nothing gets announced about DOTS but in eight months they tell us that they're deprecating their new multiplayer solution for some reason

 

Sir DOTS was introduced alongside FLUTTER!!!!!! If DOTS was open sourced I am pretty sure you, @Antypodish, @hippocoder would be top contibutors. DOTS would be standing so far above any other game engine in the market that clones of the tech wouldn't even matter because of the community.

They are wasting our time. Yes it's true we can't all make the Next Big Project TM but we all hope to. 1 engineer from there team is enough to make a DOTS library a community can build on. But there strategy has been to deliberaltely cripple releases. It's all business and you are defending there bad practices.

 

Honestly, I'm sick of people coming to the DOTS forums complaining that Unity should stop working on DOTS and focus on MonoBehaviours. Their arguments do not make good business sense from the consumer's (our) perspective if you really think through the practical implications. That's the only aspect of Unity I am really defending. If you look at my framework, you can clearly see I do not agree with a lot of things the DOTS team does. But rather than continue to make a stink about it (I usually bring up my initial concerns in case they want to address them), I just do things the way I want them to be done.

Regarding their quietness as to future releases, what they are doing is super risky, but until we see the result I can't say whether or not it is a good or bad risk. But my particular concern doesn't seem to be shared by the broader community, as I learned from a survey I conducted. So I don't push the issue much on the forums. Instead, all I will say is that if Unity cares about me and what I am doing, they should be able to avoid some of the worst mistakes they can make.

But until then, I'm going to keep doing things the way I want them done and maybe if I am lucky more people will join me on this journey to make something that is actually good.

 

I follow your framework, your survey was a great help two years ago. Yet here we are - You don't know and I don't know. But let's keep polishing this turd of the unknowing.

 

1) The survey was this year. What from 2 years ago are you referring to?
2) Unless you actually know what my biggest concern is (which I have stated publicly multiple times), what my goals are, and what my alternatives are, then please do not accuse me of polishing a turd of any kind.
3) If you believe I am saying or doing something wrong, please suggest what I might say or do instead.

 

1) whenever it was, you helped me understand that there wasn't as much community contribution to it as you expected.
2) It's about not knowing the next big drop which might impact all your goals in a good or bad way.
3) You are doing good. Unity is trying to make a Buildbox version of DOTS and we are allowing them more and more time by defending there unknown progress. Asking for a minimum viable DOTS PR release is what we all need.

 

You have exactly zero idea what Unity may or may not be trying to do.

No one is defending their unknown progress. Attacking or defending their unknown progress does make exactly zero sense. It is unknown, there is nothing to attack or defend on it.

It is you may need. I certainly don't. They said end of the year. It is not the end of the year yet. If they don't release anything around end of January/early February and they don't communicate what's happening, I'll pickup the pitchfork too, until then, there is nothing to fight about. They said they're working on it. That's it.

 

Me no idea. A Lurking Ninja ate my brain.

Yes I need it. It's not night time vigilante acitivity for me. You are always right oh great Ninja there is no need to fight.

 

"Hoped" rather than "expected". I try to not expect things outside of a professional relationship. But for the number of clones of the framework I see on GitHub, the community has been about as quiet about it as Unity has been about what they are doing with DOTS. I would have expected a lot more questions and bug reports with the volume of traffic I see. So while Unity's decision to stay quiet doesn't make sense to me, everyone in the community doesn't make sense to me either. No one makes sense to me, and I've accepted that.

That's actually not what I am the most concerned about. Perhaps there is a way to twist your words to make it account for my concern. But your statement is quite broad and my biggest concern is much more specific.

Well I am not going to be the one who asks for that, because it is highly unlikely such a PR would contain the information I would want to know. Here's what I actually want to know:
1) I want to know if Unity is solving a problem I'm solving, so I can not waste time repeating their work and instead solve a different problem. So far this hasn't happened yet.
2) I want to know if Unity makes a change that prevents me from making hooks into the system update, ordering, and world state logic, as these are what my tech relies on. There was a close call with generic jobs, but they brought it to our attention and after some back-and-forth it is in a much better state. They also broke one of my hooks involving system ordering in 0.16, but remedied it by giving me a new API that could resolve it in 0.17. I'm concerned about ISystem being very problematic in this regard.
3) I want to know if Unity is going to release a new feature (like their environment system) having a hard dependency on Unity Physics. So far the interdependencies have been pretty soft.

If Unity is doing none of those things and instead doing the things that they have already said they are working on as well as making conversion a better experience for both designers and programmers, then I'll have nothing to truly be upset about.

 

To be fair, we (and me) neither. Most of us just don't claim that we do. Well, I'd be a very bad zombie because I like goths more and "It Can't Brain All the Time".

Obviously I respect that. But you are not "we all".

 

The scenario I'm afraid of is in between those two. I'm afraid it will launch in a somewhat promising but clearly undercooked state, the jaded among us will claim it's DOA, while the naive / asshats / pathological fans among us will focus on the good parts and will keep repeating that it's a good start and it just needs some iteration. (and somehow they don't know that when you mention iteration to Unity management post release, they break out in hives).

 

Well the fact is in order to avoid a repeat of that, Unity should probably keep DOTS and mono Unity as far apart as possible, and away from the public (but keep it available for those who want to do business and can afford to do so)

This cycle of doing business with bigger customers will battle test, mature and flesh out DOTS while the rest of us wait for it to be completed. When it is complete, Unity can basically do a brand new Unity 2.0 with an incredibly strong hand and we won't need to accuse them of over promising and under delivering.

Fact is Unity does need more time, so what will we do knowing this? Keep hammering them over DOTS? What is the right thing to do? In my case I am practical and am asking for more mono performance meanwhile, better features, a stronger higher performing SRP with some of hybrid's best features etc etc.

The waiting for DOTS is nicer if classic Unity users are served a drink and some snacks meanwhile.

 

I posted this earlier in this thread.

 

How exactly is that different from what they’re doing now with the preview packages?

 

Because they are trying to make everything work in the same codebase and often they make core changes for some package to work properly, which breaks old features, and then they have to fix said old features.

Packages are far less independent from the rest of the engine than you might think.

 

I am doing what @Murgilod is doing, Bevy Rust is my solution for DOD for now, WebGPU is cutting edge tech sitting on top of Metal and Vulkan, supporting all the advanced GPU stuff.

Most problems I face I rely on MonoBehaviour veterans more then DOTS leaders.

 

Which things from dots packages broke existing features?

 

This discussion goes nowhere. Its the same argument of "Hurry up Unity! I want my cool free stuff faster!"

UT already mentioned it a couple of times - they're working on it.
If you're unwilling to wait - make your own solution.
As rough as it is - Entities are good enough even now, if you ignore conversion.

There are technical limitations and issues, but they all can be overcomed.
Moreso - sources are available to be modified.

Supplying early version of Entities did a great job of improving games for me and our company.
It provided an ability to make games faster with much more higher entity count, which is insane boon.

DOTS in general improved performance (for both runtime and editor) and provided API for things that couldn't been done say like in 2017 or prior.


My suggestion is to lock this thread and similar, and just wait for official annoucement.

 

No, it's "for the love of god, actually say something Unity. You do this every time."

Because they do. This is a tale as old as the canvas UI system, repeated by two networking systems, and just a running theme with the engine. Poor communication exacerbates the fact that this whole tech stack was announced near five years ago. We want Unity to do more than gesture towards feature delays in one-off segments of videos and forum announcements that get a handful of follow ups that amount to "we hear your concerns."

 

I do not understand how I could be asking for free stuff when I'm paying for Pro subscriptions.

I wish Unity never had a free tier so I could stop hearing this stupid argument.

 

Is reffering to the works of core DOTS, which requires core Unity changes. Which in the end, my affect "break" some existing none DOTS features. But yet I never heard, nor experienced of such hypothetical scenario.

On other hand, I understand core change require long and slow changes. And if executed right, it may bring lots of benefits in long run. But won't be anything measurable for a while.

The funny thing is, while there is less noise from DOTS team on the forum, while not completely silent, as other DOTS packages had some responses, the various Unity DOTS git repos are good and dandy. Kept fairly regularly updated. I think one of them is DOTS training git, if I recall correctly.

Also DOTS team is separate department, from main Unity core team.
It is not like they took people of it, to work only on DOTS.
DOTS team has I believe around 50 people (correct me please if I mistaken), if that hasn't changed since last year. I don't expect 50 people been fired suddey from Unity this year.
At best they may have been transfered, if anything.

 

You are paying for what you got and what you agreed for.
Not for future features. All though some of your money goes toward that as R&D.
But expecting new features right now is a bit silly argument.
Who de heck makes app based on what it may be, rather on what is available right now?
I see some major flow in a decision making.
If not, then there is literally no reason to complain. Other than expecting critical bug fixes I suppose.

 

I paid for a good lightmapper and fast start up times, but that's not what I'm getting, because the version that has those cannot build for mobiles any more.

 

That reminds me, I've been putting off making a thread to see if anyone has any solutions for the 2020 LTS startup times. I'm booting it off an NVMe and it routinely takes 1-1.5 minutes to start up even on empty projects.

 

Its current features that you're paying for. And license to publish applications.
Not some experimental tech.

Highly suggest following Joachim then. Or paying attention to the DOTS forum.

Unity is a public company now, and communication is probably gonna be even worse.
Communication is a different problem.

Also, I don't see how creating "hype" by constant annoucements would change a thing.
Hyped products or technologies usually do not live up to the expectations.

Personally I don't need a dev blog. I'd rather prefer a product.

 

Just tested, it takes me 39 seconds to load our 150GB project and 11 seconds to load a script only library on 2020.3.16f1 so no idea why you'd be so slow, especially on an empty project.

That said, even if it took 3x as long it would hardly bother me. It's (hopefully a) once a day thing. Go make a coffee?