animats
-
Posts
6,144 -
Joined
-
Last visited
Content Type
Forums
Blogs
Knowledge Base
Posts posted by animats
-
-
11 hours ago, Profaitchikenz Haiku said:
little Garbage-collection forcer
That's only needed if you really need to know how much memory you have left. Which, in turn, is only useful if you can take corrective action. Releasing some historical data, such as a list of past visitors, would be an example.
-
1
-
-
16 minutes ago, Paul Hexem said:
These days I just keep hoping LL will port some of the stuff @animatshas been working on into the viewer.
This isn't rocket science. It's basically modern game dev from a few years back, plus some less common stuff such as a Rend3->WGPU->Vulkan->Rust pipeline and many threads. Some of the stuff Unreal and Roblox are working on goes way beyond this.
The most useful new idea here is that SL lacks a crucial step in the asset pipeline, as mentioned above. We all know that. But that doesn't mean, skip the step. It means, find some other place to do it. Like BOM bake time, or daily, or when the pathfinding mesh updates, or something like that. This is where SL, Open Simulator, and metaverse systems have to part company with the classic game asset pipeline to get anywhere. SL does not have the building full of overworked junior artists optimizing content of an AAA project.
-
1
-
-
Chin Rey and I argue about this now and then. Rey builds very well, and pays attention to lower levels of detail. So her content has low LI and looks good at distance. I have some of her chairs and furniture.
The trouble is, SL has all that existing content. Much of it looks good, but is inefficient to render. There are hundreds of thousands, perhaps millions, of work hours in that content. Re-working it by hand would take many years.
So I take the position that the system has to cope with what we have - overly complex high LODs, and poor quality lower LODs. It just needs to do a better job.
Until recently, actually getting LL to do anything about this seemed hopeless. But there are now faint indications that the new VP of Engineering, Mojo Linden, is serious about pushing on performance. At the last creator user group, we heard that performance measurement tools are being used on the viewer's main render loop, and things that don't absolutely have to be done in that loop, holding up the render, are being moved to other threads. This is a step forward. Previously, efforts in that area were considered "too hard".
There's quite a lot that can be done. I've mentioned occasionally that I'm working on an experimental SL viewer not based on LL code. (Don't expect this to be usable soon. It's an attempt to push the technology of large metaverse-type systems and find out how fast you can make them go. Right now I can log in and cam; that's it. Message me if you're interested in technical discussions.)
One of the things I'm doing is loading textures in the order of how much screen space they take up. This is a big win. Everything right in front of you loads in a second or two. No more waiting for signs, vendors, and artwork you're looking right at to load. The overall effect is that the world seems to be much higher rez. I've been amazed at how much detail shops in New Babbage have. This also makes shopping events much more pleasant. You can actually see the displays as you approach them.
There's some talk from the Lindens of putting that in their viewer.
So that's just one thing that's possible and works. There are many other things.
Here's a short list of possible speedups:
- Large impostors for distant areas. Pictures of entire sims, taken every few days from multiple angles, by something like the SL map maker. That's what you'd see beyond draw distance. This is how games such as GTA V do it. When you're looking at mountains clear across San Andreas, you're looking at pre-rendered flat pictures of the distance. This is tricky to do well, so that the seams don't show, but routinely done by games today. The difference for SL is that you have to rebuild those pictures occasionally. If you want to see for kilometers, this is the only way to make it work. (An annoying problem with this is seeing all the junk in the sky from far away. Probably ignore everything more than about 128m above ground level when impostoring to avoid seeing all those skyboxes and build platforms.)
- Vulkan. There are more efficient rendering systems than OpenGL. With Vulkan, the GPU does more of the work. I'm using it. LL has looked at it.
All this stuff is more or less standard game dev technology, modified for SL.
SL lacks an important step in the asset pipeline. In game development, artists and modelers create models. Those don't go directly into the game. They go to people who use tools to simplify them and make them more efficient, working on both meshes and textures at the same time. SL has a strong separation between meshes and textures. You can mix and match after uploading. So some basic optimizations used in game development are unavailable.
And then came Bakes on Mesh. At the moment you change your avatar's clothes, all the textures come together for the first time. They go out to a baking server, get composited down to a few textures, and become assets ready to use. That was a big win. I'd like to see that extended to baking the meshes themselves, along with the textures, down into something like a one-piece game character. Like BOM, this would happen when clothing was changed. This is the moment when all the pieces of an avatar come together, and that's an optimization opportunity. This is a complex subject and I'm doing some experiments in Blender to see if this can work. Roblox is doing something like this, which is what gave me the idea. They're adding a SL-type clothing system, and they had to do something so that multiple clothing layers would not choke their system. Ready Player Me also seems to have something similar; you dress your avatar on their site and an optimized model comes out.
So, when someone says it can't be done, point them at the people who are already doing it.
-
6
-
I've said before that Social Island 10 is the Port Authority Bus Terminal of Second Life. (Those familiar with New York City will get this.) It's a transit point. Most people arrive, choose one of the portals to somewhere else, and leave. You find the same kinds of losers hanging around you find in a big bus terminal. The lost, the lonely, the crazy, the crooks, the bad musicians.
There are two big new-user problems, as anyone who spends time at Firestorm Help Island quickly sees. "What do I do now?", and "How do I fix this %$#*& clothing problem". Social Island 10 offers no help with either.
I'd suggest sending all new users to Firestorm Help Island or Caledon Oxbridge or London City for a while and see what happens. Measure retention rates.
By the way, if you go down enough stairs, there's a cute little bar at Social Island 10. Nobody goes there.
-
1
-
-
A must-read for LL management: So you want to compete with Roblox.
This is the best article in some time on the business problems SL faces. Some excepts:
Minimum Requirements to not Fail Right Away
You should consider the following as mere table stakes:
- High quality multiplayer support for user creations out of the box. (SL has that.)
- High performance servers with excellent reliability. (SL needs more server side performance.)
- Powerful, user friendly, and joyful creation tools. (Hmm.)
The Cloud Cost Trap
Why doing everything on AWS costs too much and hurts you competitively.
Tools are Hard
It is not easy to make good tools. And not just for the normal software engineering reasons – ie, you'll inevitably become too familiar with them and become unable to see their flaws, you'll forget to dogfood them and build them around speculative users rather than actual things creators want, etc. No, tools are hard because they take time. Not just time to build, refine, debug, and test, but time for creators to learn, adopt, and build them into their lives and workflow. This is much harder than a typical user acquisition problem for a simple game or CRUD app, it's more like inventing a new instrument and expecting musicians to start writing and recording music for it.
The Real Problems
The big three problems you're facing are:
- Chicken-or-the-egg deadlocks
- Platform dynamics
- Ownership and trust
Chicken-or-the-egg deadlocks
- You need players
- Players won't show up without content, so you need creators
- Creators won't show up until you have players
Content creators value audience size above nearly everything else, closely tied with fat stacks of cash money. It doesn't matter how easy it is for creators on your platform to make mind-blowing experiences if you don't also have an audience eagerly waiting to play them. (SL's flat user count is not good enough.)
Platform dynamics
For the purpose of this article I use "platform" to mean a digital ecosystem whose owner gets other people ("creators") to build stuff for players to enjoy, and the owner makes money off of this activity somehow. (SL is in reasonably good shape on this front.)
Ownership and Trust
Platforms tend to follow a certain kind of life cycle, and there's no better primer than Dan Cook's Game of Platform Power. In it he outlines how platforms transition through "Growth" and "Engage" phases where they are friendly and generous to the creators who produce value on their ecosystems, before maturing into the "Extract" phase where they leverage their size and power to lock-in users and capture as much creator-produced value for themselves as possible. Once upon a time, platforms said things like "Don't be Evil" and "users will never be required to use a Facebook account to use an Oculus" and we've seen what that's worth. (We see SL slowly moving into the "Extract" phase. SL isn't a giant monopoly and can't get away with that.)
Good Luck, You'll Need It
Look, I know this all sounds a bit crazy, but you're the one who decided to compete with Roblox. I'm just here to make you fully aware of the magnitude of the problem, coming up with an exact business model that threads that needle and adapts it to your specific situation is your job. The entire venture is crazy to start with, and my point is simply that the riskiest thing to do is play it safe.
If you do something wild and ambitious and off the wall you might still fail but at least you'll have stood a chance. Even better, you'll greatly increase the odds you'll discover a weird new opportunity along the way and pivot away from "competing" with Roblox, accomplishing something much better instead. (SL is already past that point. This is a huge edge.)
Now, all this is encouraging. The take-away from this is that nobody is positioned better than Second Life to compete against Roblox and build the Metaverse. It's going to be all uphill for new startups. Despite that, SL is blowing its opportunity.
They've called out SL's three big problems:
- Too hard.
- Too slow.
- Not enough users.
Read the article in full if you have any interest in the business aspects of all this.
-
3
-
1
-
1 minute ago, Coffee Pancake said:
Usability demands 30, gamers demand 60 and we wont be taken seriously unless that's achievable.
Might be able to get there, especially on gamer-grade GPUs. A reasonable goal is 20-25FPS on the lowest hardware supported, with some rendering features turned off and the draw distance pulled in a bit. 60FPS on an NVidia 3xxx GPU in "Ultra" mode.
-
I am ever so slightly encouraged. Mojo Linden, the new VP of engineering, wants to improve performance. Today, at Creator User Group, we got a few hints that deep down inside the viewer's rendering loop, frame rate bottlenecks are being found, using a performance measurement tool, and being fixed.
To be in the game, SL needs to get up to a consistent frame rate of 30FPS or better. Anything lower is just seen as broken.
(Incidentally, no, it's not impossible with user-created content. Anyone who wants to argue that, IM me and I will convince them that it's a technology problem.)
-
1
-
-
The Washington Post has some Metaverse coverage today. It's mostly about Tim Sweeney talking about Epic. Epic is making lots of noise about an "open metaverse", because they are annoyed with Apple trying to take a 30% cut of their revenue. But Fortnite takes a much bigger cut from creators.
Someone commented on the Post article:
QuoteThe metaverse does exist exactly as it's been described in this article... Second Life. http://www.secondlife.com
-
2
-
-
6 hours ago, rasterscan said:
Theres a new line opened goes through some log home and camper regions and the run is an absolute delight to use. The moles have excelled in the scenery creation. Rez at Randlesham station - http://maps.secondlife.com/secondlife/Randelsham Forest/150/105/69 and travel either direction for some stunning views
Also Ive made an SLRR experience teleporter to belli slrr rez zones based on a map another fan made its on the mp for free, yes copy yes trans for passing onto friends
Oh, good, a map of Belli with rez zones. We need more of those.
-
2
-
1
-
-
12 hours ago, Aethelwine said:
little wall to protect traffic from the banlined parcel in the corner with the zombies.
Ah, yes, I know that banline very well.
-
11 hours ago, Aethelwine said:
If I remember right there is a resident owned railway you can use from Yeeowler to head North into Satori.
The owner of that has reportedly been ill, and that project is unfinished.
You can get from the south end of that railway to an unmarked water rez zone, if you can find it,, and then, with difficulty, get out to open water. squeezing between obstacles and ban lines. It's also possible to make the trip in an amphibious vehicle with good off-road capability. But it's not easy.
-
1
-
1
-
-
2 hours ago, Coffee Pancake said:
Lets just add add avatars and chat to blender if that's where we're all going to be living from now on
You can do that! It's called NVIdia Omniverse. It's a way for people to work collaboratively in 3D tools including Blender, Maya, and Photoshop. Anyone tried this? It could be useful to creators who collaborate remotely.
-
1
-
-
3 hours ago, KecskeShajt Ichibara said:
Now I was able to pass hrough the switch of Gully Wash without dragging my tram manually, but just to the eastern branch, the switch still doesn't work just they rearranged the guides.
This is why the SLRR needs a track geometry car. Something which reports all such problems in a useful way. Something both users and moles could use.
-
2
-
-
Given the limitations of the Bellissaria rail system, maybe most rail vehicles should be trams, rather than multi-car trains.
The famous PCC streetcar. This is a 3D model from CG Trader, potentially convertible to Second Life use.
It's a good streetcar for Bellessaria tracks. Although long, it has a short wheelbase and a surprisingly tight turning radius. San Francisco still runs these. It's a good fit for a suburban neighborhood.
The rare double-ended variant. No need to turn it around. Nice for track systems which lack turnaround loops.
-
3
-
-
On 9/27/2021 at 2:28 PM, Mr Amore said:
The first half is just a rehash of the metaverse concept, anything of note begins at 34 minutes in. Aside from the politics and regulation, the only real takeaway is the idea of multi-metaverses where users can carry their avatar and clothing(and presumably inventories) across metaverses.
There's much interest in asset portability. The trouble is, the interest is from people who want to sell assets, not those who own them. There are several approaches.
- Outsourcing. Ready Player Me is an avatar system where the company behind Ready Player Me sells avatars and clothing, and there's an API which can be put into games to render them. All clothing changes are done via the Ready Player Me site. They have this working on about a hundred second and third tier games, but the major game companies are thus far uninterested.
- Non-fungible tokens. NFTs are technically interesting. They're basically a way to make no-copy objects and track who owns them. Or at least being able to detect the copy. Unfortunately, the entire field has been taken over the make-money-fast crowd and the Ponzi schemers.
- Assets as a service. Some big central service (Amazon? Valve?) stores assets for multiple games and virtual worlds.
They're really copy protection schemes.
On 9/27/2021 at 2:28 PM, Mr Amore said:Pixar's USD has been adopted all over this year, it's basically a format for sharing scenes(and live collaboration) between modelling software, game engines and anything.
Yes. I haven't looked at USD format that deeply, but Blender supports it now, and some SL creators I know speak well of it. NVidia and Khronos are suporting it. Disney and Pixar defined it. It's powerful enough for Hollywood. This is one of the more promising developments.
-
1
-
1
-
On 9/26/2021 at 6:46 AM, Mr Amore said:
Nick Clegg (former deputy prime minister of the UK who moved up to becoming a Facebook VP) will on Monday set out the tech giant’s vision of a virtual world where you can work, shop and ‘live’.
Today is Monday, and the announcement is out.
Facebook announced the "XR Programs and Research Fund", a $50 million fund that it says will help it develop the metaverse "more responsibly".
So, no big product announcement.
-
4 minutes ago, Ardy Lay said:
How about that XBox controller? I use one to fly the SL camera around. It can move avatar too but that seems awkward to me.
There's a little more control over the avatar. But anything driven by an LSL script only gets fake arrow key inputs.
-
While you can hook a joystick to SL, it doesn't do much. All it does for vehicles is activate the arrow keys. There's no proportional control. You can't fly any better with it.
-
I didn't think the new mesh optimizer was in yet. I know it's not in Firestorm 6.4.21. Maybe in some project viewer.
-
I went to Governance every few months, for the same reason I'd go to a city council meeting once a year or so. It's good to keep an eye on the people who think they're in charge. You could discuss things such as just how annoying do mainland ad boards have to get before they should be removed. Governance was boring. That meant things were working.
-
1
-
-
There's been discussion of age verification for SL recently, so SL could get on Twitch. Roblox is preparing to do it.
Age Verification comes in two stages. First, an ID document check, and then a selfie match. On the Roblox app, users start by scanning their ID card, driver’s license, or passport. Roblox uses advanced image processing technology to validate the document’s legitimacy.
From there, the user is asked to capture a selfie to check for “liveness” and “likeness.” Liveness ensures that the person is a living, breathing individual taking the selfie (rather than just a static image of a person), and likeness ensures that the person in the selfie is the same person as the face on the ID document. The entire process is automated and usually takes seconds to get a result after images are captured.
It will be available globally in over 180 countries on both mobile and desktop for anyone 13 years of age or older with a government-issued ID or passport.
Verification will initially be required for creators and voice users.
Roblox is mostly kids, but they want more adults. This is how they're going to keep them separate.
Tencent has been doing this for a while now. They validate against China's national ID. That's how the time limits on gaming for kids works.
Not sure if it's a good idea, but it seems to be the coming thing from the big players.
Should SL do this?
-
2
-
-
What we have so far is a statement by Mojo Linden that there will be, for now, a focus on "performance". OK. That's reasonable enough. SL is so slow that it's laughed off gamer sites for that reason alone.
Near term, performance could be visibly improved by having the viewers target a frame rate, automatically reducing view distance and quality settings if the draw cycle is using up too much the CPU time. Never dropping below 20 FPS is a reasonable goal. That's mostly cosmetic, but easy to do and useful. New users are not going to tweak the graphics settings manually. Nor should they have to.
Beyond that, it gets harder. There's much that could be done, but few easy fixes.
-
2
-
-
On 9/19/2021 at 1:13 AM, Nika Talaj said:
I don't know if a real thunderstorm would fall within the capabilities of current experience tools - I'd have to review the experience API.
Start with a boat at the dock at Rigamarole, and head due south to the new Fantasy regions of Bellessaria. The weather effects are impressive.
-
I''m curious to see what the new guy does first. What he says publicly means little at this point.
On the technical front, I don't see measuring complexity for land impact purposes as the crucial issue. It's more useful to deal with complexity by automatically adjusting level of detail of both textures and meshes to maintain roughly equal detail per screen pixel. I've demoed this experimentally for textures. That approach deals with legacy content.
(Wrong topic to discuss the technical issues. I've been discussing this on GameDev, which is a more technical forum.)
-
2
-
Facebook
Instagram
Twitter
YouTube
Flickr
SL is up. Facebook, Instagram, WhatsApp are down.
in General Discussion Forum
Posted
Big outage in Facebook land. Facebook, Instagram, WhatApp, and Messenger are down.
News coverage in all major media.
SL is fully operational.