animats

Water reflections for a pool! Kama City city center monument This is a nice effect. It's in downtown Kama City, which is several meters above the surrounding ocean. You need region powers to do this, though. This was done by raising the water level for the entire region to just below the surface, then terraforming a pool. Unfortunately, land owners in Zindra can't terraform, so the nearby nightclub with a pool can't use this effect. It's swimmable water, of course. (Incidentally, if you bought the ZiSwim swim HUD, get the new version redelivered. The old one is frantically trying to phone home to some server too many times and getting HTTP 500 errors. The new one works better.)

Uh oh. Check the US Federal Trade Commission's Sweepstakes Rule. If you have to do something of value to the sweepstakes operator to enter, it's a lottery. There are a whole series of specific rules for a sweepstakes in the US. California is even tougher. This has been coming up a lot lately with "promote us on social media for a chance to win" schemes. Is this really from Linden Lab? They have all those compliance people they hired for Tilia. Did no one look at this? The legit way to do this is to just send every responder something. A SL T-shirt or coffee mug. I just received an unexpected coffee mug from a game development conference, and all I did was sign up for a free virtual event.

Is this the same "SSOC" that has the big Japanese urban sim near Mopire?

That's the only real justification for going through all the headaches of using a blockchain. While you could do this for SL only, it's pointless, except that you get to shout "NFT" a lot and maybe charge way too much for stuff.

I've mentioned before that someone could implement this in LSL. The object would check on rez, or once a day or so, to see if the blockchain ownership info matched the in-world info. If not, you'd get a warning, and it would return to inventory or be deleted. Then you could have something like an inter-grid version of CasperVend that would allow you to get redelivery of your objects on OpenSimulator, and mailboxes where you could send your objects to other grids. Then you could move unique objects between grids. I don't see much of a use case for this, but if someone really wanted NFTs in Second Life, they could implement that without LL help.

Go to Cocoon when Japan is awake and see what people are wearing.

The concept of "fitmesh" is that it Just Works. The clothing responds to all the avatar sliders. With regular mesh, the customer has to fiddle with the sliders to get clothing to fit. If fitmesh doesn't fit perfectly, it's the creator's fault. Thus, it's unpopular with creators.

"The Second Life Status Feed is down." - from https://support.secondlife.com/?lang=en-US No indication of outages on the status page. Will check tomorrow to see if this clears up.

First, deal with the RL problem. If you're in the United States, the The National Domestic Violence Hotline is 1-800-799-7233. They also have a web site at https://www.thehotline.org/.

Good point. SL motion needs a lot of cleanup for VR to be tolerable.

This is encouraging. It's not impossible to make VR work for SL, just hard. And expensive. If you could get a viewer running at a solid 60 FPS, and then use a headset's GPU hardware to shift, rotate, and interpolate to 120 FPS, it could work out.

Very nice! I just went to look at the castle. New Babbage is getting huge. 15 regions now.

That's a displacement map, a standard technique. The next step up from "bump maps". You can do that in Blender and see the results in Blender,, but Second Life can't use a displacement map. Bump map, displacement map, both. See the difference? How often do you get that close? Examples from the Blender manual. The GPU does the work in systems which support this, and it slows down rendering. It's neat, but not near the top of the rendering improvements SL needs. My rendering list would be roughly 1) physically-based rendering, so shiny objects don't turn white, 2) subsurface scattering, so skin looks more like skin, 3) more accurate shadows, so avatars don't look like they're floating above the ground and walls don't leak light at the base.

Please go on. You may be on to something. How should walking be different in SL? Right. You can hook up a joystick, but all it does is operate the arrow keys. There's no proportional control. It would be tempting, as an experiment, to send joystick info. The "control" event sends 32 bits from viewer to simulator 30 times a second, and LSL can read all of them. But only 12 bits are used, leaving 20 free for expansion. If you sent 4 bits each from 4 axes of joystick, you could make a standard two-joystick game controller talk to SL. Each joystick would have 16 positions, so it would feel smooth. You'd have to script a vehicle to read that data. Someone should put that in a third party viewer and try it out. I'd like to see one of Kelly Shergood's helicopters set up for that, so you could operate cyclic, collective, and throttle properly.

Hm. I've watched some of Rosedale's videos, and I spoke to him in world once, but don't really know him. From his talks, he has a vision of SL. That vision does not scale much beyond present levels of engagement. It's a vision of a good virtual world built by users with a commitment to that world. We have that now. It's a niche. It's probably a bigger niche than SL occupies at present, but it's a niche. As Rosedale points out, it requires a commitment to the virtual world. As such, it appeals to people whose RL life is unsatisfactory. He doesn't go all the way to using the word "losers", but that's what he's edging around saying. This is a problem if you want to scale. Now, there are ways around this, but they have their own problems. One is a two-class system, where there are creators, who get revenue for what they do, and casual users, who just consume content. Roblox explicitly has that. Roblox is YouTube for junior game devs. Post a game, get viewers, profit. Fortnite has some of that, although it's a minor part of the user experience. Ready Player One has that - most of the protagonists are casual players, and Aech is a content creator. In all those examples, the content is a game, with a start, an end, a goal, and a winner. The world is open-ended to creators, but episodic to casual users. SL lacks an episodic component. Attempts to add one have been failures. From Paleoquest to Zenescope, they're just not that much fun. SL's tough onboarding means there are not many casual users. So this approach doesn't really work for SL. What could work? That's a hard problem. LL certainly doesn't have an answer.

Don't force garbage collections unless you have to. They're slow. The idea is to wait until it looks like you're running out of memory, then force a GC and recheck before taking other actions to avoid running out of memory. Unless you have a script which really can do something useful when nearly out of memory, don't force a GC. A good use would be a greeter script that kept track of recent visitors. If llGetFreeMemory indicates it is close to running out of memory, it should remove the oldest entries.

Here's an essay on that topic. Yes, those people are trying to sell their avatar system. But what's worth noting is that they went all the way to working photorealism and backed off. They discuss why. This is relevant to Second Life, where the avatars approach photorealistic, but don't quite get there. This puts them in the uncanny valley. With face tracking on the official SL roadmap for 2022, this becomes more relevant. If we have face tracking, what should avatars look like? The uncanny valley. The Polar Express movie (2004) was where Hollywood bottomed out on this scale. Facial motion capture was used in a mainstream movie for the first time. The result was considered creepy. The overall look was weird - too realistic to be a cartoon, not realistic enough to be live action. The lesson was that once you get deep into the uncanny valley, you have to up your game in all areas to get out. The rendering has to get better. Especially for skin, where you need subsurface scattering, the effect that makes skin glow slightly. Motion has to look right. Hollywood has managed, at great expense, to get out of the uncanny valley. Games are getting there. (See the Unreal Engine 5 Matrix demo.) SL is part-way into the valley. A question is whether facial motion capture moves it deeper down or starts to go up the other side to photorealism. I'm not against motion tracking. I think it would force LL to up their game in other areas as well, though. Comments?

So far, the metaverse crowd that makes a lot of PR noise hasn't done much about portability. NFT portability is, so far, just talk. There is something called Ready Player Me, which you can look up. It's more outsourcing than portability. Game developers can outsource avatar appearance management to another company. All appearance changes happen on their web site, they have total control over avatar appearance, and own all clothing and models. It's mostly used by small game developers who don't want to build an avatar system.

Terraforming takes some practice, but you can practice. There's a small terraforming practice area at Builder's Brewery. There's also an entire empty Linden sim for terraforming practice, but I don't remember where it is. Note that if you have a subway below ground, the rest of your ground has to be well above the water level, so you have to figure out how the shores of your region(s) are going to work. Go spend some time in New Babbage, which has this well worked out, with underground tunnels, docks, and canals.

Wide view. Who needs VR goggles? Ultrawide monitors are becoming cheaper. The same processes used to make 4K TVs work to make these. This may be the practical answer to how Second Life competes with VR.

There used to be a thing for 4x4m parcels as bases for giant billboards along roads. The Circuit de Corse in Corsica, which is a popular drive, used to have several. I haven't seen those in a while. I ARd one of them years ago, and probably other people ARd the rest of them. As ads, they weren't working anyway. About half of them were "rent this billboard". There are still some little brightly colored ad things on 4x4 parcels. Some say "Taxi" on top, so maybe they have teleport links. Never clicked on one. Ugly, but harmless.

I'd like to see an SL viewer on NVidia's GEForce Now service. There's a free tier that's usually busy, and you get kicked off after an hour. More useful is a $100/year tier that lets you stay on for four hours. The viewer runs on one of their servers, and sends video to the client, which can be anything with enough power to stream movies. Why GEForce Now? - Minimal effort to convert a viewer to it. LL or Firestorm might be able to make ti happen. - Moderate cost. - NVidia doesn't get involved in payments, so Linden dollars will still work. Valve's Steam, Google's Stadia, and Apple's store insist on games using the store's payment system. Downsides: - Resolution is limited to 1080 lines. - It uses a lot of bandwidth. About 25Mb/s. If you use this on mobile and don't have unlimited bandwidth that's really unlimited, it will cost you too much.

Assets have been on AWS, front-ended by Akamai's Ngnix servers for years. Assets work just like web assets, accessed through HTTP, while the sim software is all custom. None of that changed during "uplift".

If you want to understand Nanite, go watch this SIGGRAPH video. If you're into rendering theory, it's fascinating. If you have no idea how computer graphics really works, you'll just get frustrated watching this. So go watch the UE5 Matrix demo instead. Nanite is very clever, but relies on content having quite a bit of repetition. A key idea of Nanite is that a mesh can have submeshes which are replicated. You can model a chain link fence in Nanite and see every bend in every wire, plus the little bumps on the wires from the galvanizing. But there are only a few different parts. Miles of fencing will not result in a huge mesh file, because of that heavy replication. The replication happens at multiple scales - one wire bend, one vertical fence wire piece, one fence panel with post... That's how they get super detail without super file size. Great for terrain, roads, dirt, sidewalks, and repetitive buildings. Not so great for SL content. SL has little repetition. That's forced by the many-creator model - you mostly use your own parts, or pay others for theirs. Game development uses big, common asset libraries. I was looking at a walkthrough of the Cyberpunk 2077 world a while back. The same items keep showing up. The same railing and the same trash bags show up in quite different areas and contexts. Now look at New Babbage in SL. Every railing is different. SL has no LI penalty for having 10 different chairs instead of 10 copies of the same chair. SL content has maybe 10x the number of unique meshes as game content. This slows both rendering and downloading. Could SL use UE5? Probably not. Nanite requires heavy offline preprocessing. But I suspect that we will see UE6, Metaverse Edition, with better support for dynamic content, in a year or two. Game content goes through much preprocessing and optimization between artist and shipped content. Unreal Engine does all that in the development tools. Some of that could potentially be done in backend servers. Yes, everything can change in a dynamic environment, but in practice. most of it doesn't. That has to be exploited for performance. SL does a bit of this, with Bakes on Mesh, and needs to do more of that in the avatar area. Here, as I see it, are SL's biggest problems in development: They have a big engine with a big problem and a small staff. They have a huge legacy code problem in several dimensions. SL's code is all unique to SL. Few libraries are used. They have their own everything - their own vector and matrix math, their own networking, their own data formats. So new hires won't be productive for a year or so while they learn the innards of the system. Going to work for Linden Lab can be career death for a good developer. Nothing learned at LL is directly transferable to a new job. They're not getting the buzzwords employers want to see on resumes. They won't have Kubernetes or UE5 or Unity or any of the web stuff. Anyone who can make real progress on LL's code is probably good enough to get a job at Google or Facebook or Amazon, make $350K a year, and have a future. SL has a bad reputation for being obsolete, obscene, and full of losers. A worse reputation than the NFT clown car. This makes hiring tough. More of a people and money problem than a technical problem. Roblox and Unreal have dev budgets 10-100x bigger.

Yes. The consensus is that for 24/7 servers, AWS costs about twice what colocation costs. (Colocation means you own your servers but rent space in someone else's data center building.) AWS is most useful for variable loads. Before the AWS transition, Oz Linden was saying it would reduce costs. After the AWS transition, he said it had increased costs. I won't speculate on why the move was done.