Second Life starting to show it's age again

It's not easy to keep things up to date in terms of rendering engines and the like but SL is, in my opinion, really falling behind. Admittedly, Windlight added some great atmospherics to the game but it fell short of realtime shadows (the problem with which might be that most builds aren't water tight and might leak shadows in places, resulting in something very ugly).

We're still using trees we were used to seeing in the times of Everquest, there has been no mention of SpeedTree recently, but again, the problem as I see it is that the current rendering engine is incapable of handling overlapping transparent layers properly, as most of you know from messing about with alpha maps. It'd make rendering trees a very awkward affair with the priority of certain leaf materials becoming mixed up.

There's also an air of snobbery surrounding Second Life critics and I'm starting to find myself feeling the same way. Functionally speaking, SL is great though not without problems. The real issue is the aesthetics. When you have a bunch of video game critics get together in 2008, the first thing they look at is the graphical quality. The general consensus is that the base animations for walking and the like are sub par, to say the least. We can speculate that something may be done about it in the future but looking at how little has been done about it in the past, we shouldn't be holding our breath. Perhaps LL are working on implementing euphoria physics, something that'd allow the client to create animations on the fly.

What are the other problems, inhibiting future progress?

Would SpeedTree put tree sellers out of business?
Would Euphoria physics put the animators out of business?
Should we care?

It'd be nice if there was a Linden devoted to handling these queries, the ones about graphics I mean. The great thing about Second Life is that it's just a database and the only thing that the engine does is rez assets with attributes drawn from it. Yes, it does other things like prediction, blah blah, but does it do so much that jumping to a different engine entirely would be impossible?

I find that doubtful. I'm just wondering what direction LL is planning to take SL in. Yes, "graphics aren't as important as gameplay" I hear some of you say, but the fact is that they are as important. Second Life needs to continue to impress people if it's to achieve the goal of becoming as widespread as a standard internet browser (I personally believe this to be impossible but whatever), though I wonder why LL aren't making use of engines like Unity 3D to create a browser based client for Second Life.

There're many areas LL ought to be looking into, and I'm sure a few of it's employees are, but, it has to be said, SL is slow to make any real changes. What needs to change in order for it to keep pace with the newer technologies?

Opensim is coming along nicely, though every MMO has it's own version of Opensim at some point. People use packet sniffing to try and emulate the server processes and eventually mimic everything completely. The only difference with SL is that LL are actually helping with the project and not trying to shut it down.

SL continues to grow but things can change very quickly. While Playstation Home is not a threat, it being on a different platform altogether, it will inspire developers to challenge SL in the race to create a real standard for 3D worlds. Does SL have to lose some excess baggage to keep up?

I just realized yesterday that there are lots more things going on behide the scene in SL client coding. I've seen photos of real-time shadowing in SL already, by a developer friend. It look great to me and he said it drop his fps down by 5. Mean my poor graphic card won't keep it up on a continuous basis, but it would be available nonetheless. We can be expecting real-time shadowing option in a couple of months, he said. And from that, I'm sure there're many more projects going on that us normal user don't know about.

I'd check out the SVN myself, but I'm not client programmer.

If the client included some sort of functionality for texture baking it'd make things easier. Let me give an example:

You create a floor using a 10 x 10 prim and in the center you put a cube with a light source to one side of it. After you texture bake the floor prim (the entire prim as opposed to one surface) you see the shadow cast by the cube appear, but no one else does. You should then have the opportunity to upload the lightmap to the server. The lightmap can be a very low resolution, it doesn't need to be crisp. To start with 128 x 128 would suffice and would create some incredible looking scenes.

I've already texture baked a few prims, although I used 3D Studio Max and Vray to do this. It's far too long winded and complicated for the normal user base to learn but the results are what I'd like to see from a similar function inside SL.

http://zante.myzen.co.uk/Second_Life/screenshots/Bake_001.jpg

More than agree. And I wish there is a new "snapshot" option too. One that re-render the whole scene in a max-out quality and without LoD in calculation. They can even put in ray-trace shadowing in there for all I care. I don't need extreme lag while I move, but I can wait a while for a beautifully rendered scene! =p

Can't argue with that, anything that makes screenshots look better can only be good for business.

I doubt it will happen too. That being said, all those high graphic games are aimed at a niche market, and even though SL is as well, the bleeding edge gamers are an even narrower demographic. They tend to have new , up do the edge equipment, which they update, and tweak constantly. If SL were to focus on this group as their primary user, a good portion of us, those who aren't gamers or computer enthusiasts would be left in the dust. In order to appeal to as broad a range as possible, LL has to develop and improve at a rate to keep the Siliconheads happy, but also to keep the Luddites interested.

Here's a video of the shadow draft client, I don't know where to download it :[

http://youtube.com/watch?v=fHjibHylkkU&feature=related

One of the biggest differences with SL is that it is an OpenGL platform, which is user content generated. Raytracing demands a high level of computations and would have to be client based.

The 'difficulty' comes when users create content on the fly. Simply rezzing a prim, would require all the clients present on that sim, and the child objects (avatars in adjoining sims) to be broadcasted the information, in order for them to compute the effects on their screen, irrespective whether they can see the effects or not, on the basis they may fly closer. Consider this in the corner of a sim, with 3 adjoining sim corners.

Suddenly 20 people all drop linked primed objects, such as a house, complicated tree, a milti primed cuddly bear even flexi based prims, all having an effect on the potential shadows people will see. This will cause a freeze of the system (lagg) as it tries to broadcast the basic information to each client to render the various shadows. We all know the effects of 20 ppls tping into a sim!

In the vast majority of games, the scenes are pre-created and light maps are rendered along with the multiple light maps to make the rendering adjustments for predictable action (blow up a building, chop off a tree branch etc). All those events are predicable, whereas the ilke of SL is not. DirectX and platform specific games, all have the specific light maps and pipelines that can cope with the effects, it just appears to be dynamic.

In SL, it is not the platform programmers that predict, create in advance, what the users create on the fly by clicking the 'Build' button, therefore they cannot render light maps in advance and supply that data stored on the clients system. They could.... perhaps create some predictable light maps, but on what predetermined basis, would it be for the basic prim shapes? Hardly worth it, unless we live in lego land with set sized shapes and not user interaction other than perhaps placement.

Consider that the SL client also allows each user to set the time of day and environment, the permutations between clients would be enormous and therefore lessen any potential predictability. Having to rely soley on the users system capabilities in a OpenGL platform, which has only seen minor development over the past few years, is hardly going to create an atmosphere akin to the likes of HalfLife, this side of neutron based computers with a few trillion instructions per second. I'm not saying it cannot be done, because clearly it is on the horizon, but it will never compete with other non-creator intervention platforms.

As one poster pointed out, shadow bleed and interaction with prim face refractions linked to the environment, would be the issue to obtaining such perfection of the predictable platforms, when even overlapping alphas sometimes stress and confuse the client. The vast majority of GC advances these days have been the computational leaps in the direction of DirectX for windows and XBox systems. Obviously, increases in memory size and speed, combined with increases with core speeds, helps with the computations of the OpenGL pipelines, but they cannot take advantage of the specific code written for the DirectX pipelines.

As a Machinimast working within the SL and OpenSim platforms, I would welcome dynamic/unscripted shadows for sure. At the moment, I have to create and predict every shadow and its interactions, by hand/eye. The illusion is somewhat muted when an avatar walks into the scene. I wait with bated breath to see what additional features graphically, hit the SL Platform.

Wow that's crazy cool.

I love the effect, but to my machinimast's eye, the shadows are incorrectly orientated. Call me a perfectionist/anal retentive, but I work hard to consider multiple light sources and direction/reaction. I admire your work to at least give SL what it currently lacks.