There really should be an SL2. Not Sansar.

[Blush Bravin]

Forgive me if I am in error; I am not a techie by any means. But, I always though the problem with SL isn't the code but constantly having to be sure they don't break content. Backward compatibility is what makes advancing SL so dang difficult. Isn't it?

[Selene Gregoire]

2 minutes ago, Blush Bravin said:

Forgive me if I am in error; I am not a techie by any means. But, I always though the problem with SL isn't the code but constantly having to be sure they don't break content. Backward compatibility is what makes advancing SL so dang difficult. Isn't it?

In a nutshell, yes.

That and the floating point error that's been there since forever.

[Coffee Pancake]

15 minutes ago, Blush Bravin said:

Forgive me if I am in error; I am not a techie by any means. But, I always though the problem with SL isn't the code but constantly having to be sure they don't break content. Backward compatibility is what makes advancing SL so dang difficult. Isn't it?

To a point, but not so much as you would think. An old prim and a new prim are exactly the same. So as long as you don't break prims, both will work just fine. Tinkering with physics has to be done with extreme care, but only because scripters will have made assumptions about how the physics worked at the time they wrote their code. And so on ... old content is a concern, but it's rarely an albatross 

12 minutes ago, Selene Gregoire said:

That and the floating point error that's been there since forever.

That's not an error, that's how floating point numbers work.

 

[Selene Gregoire]

Just now, CoffeeDujour said:

That's not an error, that's how floating point numbers work.

So you are trying to say floating point errors don't happen and the Ariane 5 didn't explode due to a floating point error? Ok.

https://itsfoss.com/a-floating-point-error-that-caused-a-damage-worth-half-a-billion/

[Coffee Pancake]

You really should have taken the 10 minutes to watch the video....

[Gabriele Graves]

7 hours ago, chibiusa Ling said:

Hmm....this is purely my own personal wants and desires but my version of SL 2.0 would have to include :

-Prims....I don't care what you all say I love in world building. But enhanced, so dividing faces, editing points, edges etc etc. Basically rudimentary "mesh prims". Limitations on how many divisions can be applied etc etc

-Mesh importing BUT with limitations on the complexity of the items that can be uploaded or perhaps a restriction on MB size (see region section further down)

-Maybe texture limits on uploads to keep textures at an acceptable size

-Sound clip lengths extended to 60 seconds or more

-Enhanced LSL 2.0 based on a powerful but simplified language/syntax along the lines of Python, Swift or Lua etc. With other extra features such as custom user function libraries, more enhanced debugging such as stepping through code, code cleanup, predictive function display as you type etc etc

    -Added on to the above, more useable controls. In the present day Up, Down, Left, Right, Mouse, Pg_Up, Pg_Down and Shift + Left or Right.....is not enough

-High quality fully customisable starter bodies that actually look decent that ANYONE could create for instead of the elite few.

-Clothing that can be created in world. One idea could be along the lines of current system clothing but more enhanced with more options, way more clothing styles, customisation and rudimentary baking for shadows and occlusion.

-Master user accounts and under that you create your alts which all share inventories.

-Enhanced regions, much larger in scale, with maybe voxel or similar terrain that allows you to tunnel, re shape and remove terrain to your own specifications. Maybe "land count" and prim limitations removed and a switch to restrictions by Megabytes. That way you could possibly have more in your sim than the original 20k prims would allow so long as the size of each item was low enough. Also perhaps enhanced terrain design features that allow you to generate trees and grass. Also the terrain split up into sectors that allow you to apply different textures to different areas...idk stuff

 

I do have more but I am too tired right now and no longer want to type

Also remove the existing Linden builds from old mainland and put newer builds in their place to showcase what could can be done.
Move the old builds to a set of museum islands to preserve them if it is deemed important, otherwise their time has past.

Edited February 8, 2019 by Gabriele Graves

[Coffee Pancake]

You know ... if someone came up with some concept art and a proposal, I'm sure they could get the labs attention

[Paul Hexem]

Popular games aren't popular because they have simplified graphics. 

They're popular because they're easy to learn, easy to use, and allow a player to log in, accomplish clear objectives with minimal investment, and log out, all inside the 30 minutes of free time they may have after work/school. 

[KanryDrago]

On 2/7/2019 at 7:04 PM, Bree Giffen said:

I think you're right. SL2 is Sansar. What I'm suggesting to fill the void is something else entirely.

Sansar is not SL2, neither is your suggestion. Though Sansar and your suggestion do seem to have something in common. Judging by the replies a lack of interest from current SL residents

[Coffee Pancake]

4 hours ago, Gadget Portal said:

Popular games aren't popular because they have simplified graphics. 

If anything, a more stylized look allows for an overall lower poly experience that can both be ported to anything with a screen and ages well. It does however require a strongly defined and enforced design aesthetic which is impossible to accomplish in a free for all mess like SL.

There is nothing stopping anyone from creating avatars and clothing in that kind of style, SL's render engine could easily make such content really pop.

[KanryDrago]

30 minutes ago, CoffeeDujour said:

If anything, a more stylized look allows for an overall lower poly experience that can both be ported to anything with a screen and ages well. It does however require a strongly defined and enforced design aesthetic which is impossible to accomplish in a free for all mess like SL.

There is nothing stopping anyone from creating avatars and clothing in that kind of style, SL's render engine could easily make such content really pop.

You are forgetting however it is not enough that it can be done it is also necessary that all are compelled to use it. I am sure if the Labs created the perfect world posited then there would be a clamour to shut down SL from certain quarters when the perfect world failed to entice people over. 

[Coffee Pancake]

Just now, KanryDrago said:

You are forgetting however it is not enough that it can be done it is also necessary that all are compelled to use it. I am sure if the Labs created the perfect world posited then there would be a clamour to shut down SL from certain quarters when the perfect world failed to entice people over. 

I'm pretty sure that if SL had a constitution the first line would be "Expletive you, I do what I like."

[ChinRey]

10 hours ago, Gadget Portal said:

Popular games aren't popular because they have simplified graphics. 

They're popular because they're easy to learn, easy to use, and allow a player to log in, accomplish clear objectives with minimal investment, and log out, all inside the 30 minutes of free time they may have after work/school. 

I gave you a well deserved like for that post but I still have two serious objections.

One is that you can't achieve much in 30 minutes if half of that time is spent downloading too complex graphics.

The other is that the graphics of Second Life is actually too simplified because it is too complex. Jellydoll avatars, collapsed LoD models, short draw distance, items failing to render properly (or at all), the sparse vegetation, the limited use of high quality shaders and surface maps - those are all de facto graphics simplifications and they are all caused, at least partly, by models with nominally high poly and pixel counts.

Edited February 11, 2019 by ChinRey
Typos

[Coffee Pancake]

1 hour ago, ChinRey said:

One is that you can't achieve much in 30 minutes if half of that time is spent downloading too complex graphics.

It wouldn't matter what the fidelity of the meshes and objects were for SL, it has always involved sitting around waiting for assets to be fetched and textures decoded, even if everything was already cached, the bulk of the work still has to be done. Games can do it as  everything a player needs is loaded long before it's seen.

A better cache will go someway towards fixing this, but still, there is a huge difference between pulling in 100 small files and just decompressing one pre-packed ready to go blob into memory.

Quote

The other is that the graphics of Second Life is actually too simplified because it is too complex. Jellydoll avatars, collpased LoD models, short draw distance, items failing to render properly (or at all), the sparse vegetation, the limited use of high quality shaders and surface maps - those are all de facto graphics simplifications and they are all caused, at least partly, by models with nominally high poly and pixel counts.

The real kicker is that we're sinking huge amounts of expensive CPU I/O time into loading and rendering junk. Junk looks ugly prompting the user to crank their LOD though the roof to try and circumvent the entire process, only creating a self defeating feedback loop in the process as creators work to this jacked up LOD standard.

The mesh package is processed in a very ideal-world-logical-way that's great on paper. The idea being lowest LODs load in one after another finally stopping at the highest suitable LOD for the things distance from the camera and the clients LOD settings. LOD is ment to act like a 'quality' slider, slow PC or poor network, turn it down .. but that's not the way we abuse use it.

For rezzed objects we get  junk -> junk -> mostly junk -> the thing, half the time spent is sunk into meshes that aren't worth seeing and probably wont stay on screen anyway due to jacked lod settings.

For worn assets it's really bad as no one cares about Li, so we get junk > 98% full rez thing -> 99% full rez thing -> the thing -> the rigging. End result, the viewer is loading and displaying almost the full mesh, almost the full mesh again, the full mesh finally, the rig and only then manages to place it on an avatar. This is why attachments float around and take ages to load.

Downloading files is expensive, loading files from disk is expensive, unpacking files is expensive, loading stuff onto the GPU is expensive, lets do that lots and throw most of it away the moment it's finished. Lets make this mess worse in high load situations by trying to get every thing or avatar to load in at about the same time.

🤦‍♀️

I 'm really starting to think the current mesh structure needs a do-over. Scrap letting the user make LOD meshes, we can't be trusted. Make the mesh package just the rigging, the final mesh and some weights to guide automated LOD generation. Let the viewer do the lower LOD magic and save the result clientside for next time. This puts LOD rendering in the hands of LL, and while it wont be perfect, it can always be improved upon retroactively making literally everything we look at better as the technology in the client improves. Win win win.

</rant>

On the subject of graphics simplifications, the lab really need to take a step back evaluate our common ugly hacks and come up with some ready made solutions that help us tick those boxes without us killing performance in the process.

For example, while we should NEVER be allowed to craft our own shaders, a pallet of effects shaders vetted and provided by LL, so we can do water and glass and fog and whatever else without needing to resort to crazy expensive alpha blended trickery, they could even be provided with there own LOD.

 

 

Edited February 11, 2019 by CoffeeDujour

[ChinRey]

41 minutes ago, CoffeeDujour said:

it has always involved sitting around waiting for assets to be fetched and textures decoded, even if everything was already cached, the bulk of the work still has to be done.

"Has always" does not necessarily mean it has to be that way.

There are a lot that can be done about the streaming cost - in theory at least, I can't imagine it will ever happen in SL:

• For data transfer:
  • Coalesced linkset data. I mentioned this in the Idle Thoughts... thread as a way tor educe the assets server load. It would also drastically reduce the data transfer by combining a lot of micro packages into fewer (still quite small) ones.
  • More procedural objects, less polylist meshes. (This is the big one actually.)
  • More sensible texture use (that's mostly up to content creators and users of course).
  • A more efficient assets adressing system. An UUID is 16 bytes. That may not seen like much but there are so many of them. LL did go for more copact addresses for mesh files but unfortunately upgrading the system to do the same with older assets classes is out of question.
  • A dedicated compact sculpt map format. Sculpts are still very much a part of SL and will remain so in the foreseeable future. Today a sculpt map requires at least 16 kb of data, in extreme cases it may well be 4 MB. Only 3 kb of that data is actually needed, the rest is nothing but wasted bandwidth.
  • A more streamlined transfer protocol.
  • Reduce the amount of unnecessary data updates. There seems to be a lot of that going on.

When ti comes to reducing the cpu load, more efficient data transfer will help here too and of course better use of multi-threading will improve the performance on modern computers. There should be a lot of other ways to lighten the load here too. I'm not familiar enough with that link in the chain to go much into detail but fitted mesh and sculpts in particular seem to requrie far more preprocessign than what is reasonable to expect. I also have the feeling that the 21 years old jpeg2000 format isn't quite what we need today. The way it handles low resolution "previews" in particualr, leaves a lot to be desired. This only affects subjective, not objective load time but that is important too.

But it's all dreams of course, it's not going to happen. Then again, substituting fantasy solutions for harsh realities has become all the rage in certain RL nations these days (no names mentioned) so why can't we do the same in a virtual reality?

Edited February 11, 2019 by ChinRey

[Coffee Pancake]

The only way we will ever have effective multi threading is if the entire client is pitched over the side and rewritten that way from the ground up. Retro fitting it is exceptionally difficult to do and requires a huge amount of time profiling every little change on multiple systems .. made worse by how difficult SL is to benchmark.

[Paul Hexem]

3 hours ago, ChinRey said:

I gave you a well deserved like for that post but I still have two serious objections.

One is that you can't achieve much in 30 minutes if half of that time is spent downloading too complex graphics.

The other is that the graphics of Second Life is actually too simplified because it is too complex. Jellydoll avatars, collapsed LoD models, short draw distance, items failing to render properly (or at all), the sparse vegetation, the limited use of high quality shaders and surface maps - those are all de facto graphics simplifications and they are all caused, at least partly, by models with nominally high poly and pixel counts.

SL is the only game where you gotta wait around that long for stuff to download. Most others deliver content way ahead of time or far more efficiently.

The simplified/complex mess really is a mess. I don't have a response for that.

[Bree Giffen]

The main point of my suggestion was actually to address the missing link that seems to exist from watching young kids cutting their teeth on creative mode in Minecraft and building homes in Roblox. I just can't see these kids growing up in just a few short years and going to Second Life the way it is right now and Sansar, High Fidelity, etc. I watched some videos on Sinespace and creation seems to be just as complex. This isn't for SL residents either. It just feels like an opportunity is being wasted. I do acknowledge that even thinking up a simplified 3d building interface isn't something that can be done easily. It will take some kind of truly inspired thinking. It's easy to say something should be made but really difficult to do.

[rasterscan]

At 16 years old now SL one is getting more and more cranky.  They tinker around the edges, usually making things worse instead of better. The earnings the game bought LL should have been poured into improving it, not wasted on a project like sansar. Not even 3D glasses took off, never mind strapping binoculars to our head to play. Aint ever gonna take off big style.

[Coffee Pancake]

17 hours ago, rasterscan said:

At 16 years old now SL one is getting more and more cranky.  They tinker around the edges, usually making things worse instead of better....

Citation needed.

[Selene Gregoire]

36 minutes ago, CoffeeDujour said:

Citation needed.

If anyone had bothered to keep a record of the number of times LL has screwed things up (and sometimes royally screwed up) the list would be pretty long. For a while there, they had to roll back everything they did because it was too buggy. The result of pushing things out before it should be.

One of the worst disasters, way back in 2005/6 or so, ended in the grid being down for several days (a week if I remember correctly). There have been other, not quite so drastic, disasters. All of the incidents I remember were a situation where LL had tried to improve on something but something went wrong and things were really worse, instead of better.

That's not to say things didn't get fixed. They did. Mostly. Even that old permissions bug where perms just arbitrarily change themselves (even in a vendor) rears its ugly head now and again. I know that has happened because it happened not only to me, but to my RL other half as well.

Edited February 14, 2019 by Selene Gregoire