Transparency + shiny, why not?

This is a feature I've been waiting to see for years now, but LL doesn't seem to have any thoughts on implementing it and I cannot understand why. There are many objects (such as water and glass) on which people always wanted to enable reflections and have transparent as well, but the LL viewer does not render Bumpiness and Shininess on a face that has any Transparency bigger then 0 enabled (or an alpha texture).

I don't believe it's that difficult to enable this in the viewer. What could it take to make reflections support being transparent as well? As for bumpiness I really can't understand why that gets removed once the object / face becomes transparent too.

I checked a Shadow Draft beta client a few months ago and even with the new shadows and lightning enabled (deferred rendering) turning transparency on still kills bump and shiny. So I just wonder, can't LL fix this already or anyone make a viewer patch to resolve this issue? Even old games like GTA3 support transparent reflections on windows, and it's a feature many are probably starving for years to use on SL builds too.

I already have a jira open for this... here is the link if anyone wants to check it: http://jira.secondlife.com/browse/VWR-4478 . What are your thoughts on this? Why isn't it implemented and what would it take to get it working?

It isn't LL's choice. It is a factor of how OpenGL works. As with the alpha sorting glitch, or the limited number of local lights, the only way to "fix" being able to do alpha and shiny at the same time would be to drop OpenGL and go to some completely different rendering system. And doing that would pretty much require starting a whole new grid from scratch, because so much of the content here that was designed for the way that OpenGL works would break. Also, the few alternative 3D real-time render engines that I have heard of are not cross-platform, and would not be available to Mac and Linux users. A Windows-only solution won't fit with our installed resident base.

Is there really no way in SL's OpenGL system to do it at all though? I play many games which run on OpenGL and they have many such things. I'm not that good into the client code but I think there should certainly be a way to enable this one way or another.

And it's already done actually, this works on water. The water in Windlight now reflects its environment, is transparent and even distorts what's under it at the same time. Maybe the on-prim shiny system would need remaking a bit or use the same one as the water with other types of reflections, but this should certainly be doable as far as I understand.

Also, since you mentioned the limited number of local lights, these are already fixed with Shadow Draft. If you get a SD client and enable deferred rendering, you can see all lights without any limit, so it's not all impossible even if a lot of people thought they could never be fixed.

I don't think that's quite accurate, Ceera. I don't know for certain, but anecdotal evidence strongly suggests the issue is in LL's implementation, not in OpenGL itself. In other 3D modeling programs which are OpenGL based, such as Maya, for example, I make things that are both shiny and transparent all the time. And as Mircea pointed out, we see it in OpenGL games regularly. SL's really the only program I've ever run across that can't do it.

I've always assumed the reason LL doesn't allow it is because it's not an easy thing for which to determine good rules. If I make a texture that puts an opaque figure in the center of a transparent field, and I make it shiny, was my intention for the figure to be shiny, or the field, or the whole thing, or some other variant?

Really, the right way to control it would be to allow for an additional channel in the texture, to serve as a shininess map. Since we're five years out of beta now, and we still don't even have a customizable bump channel, I'd say the chances that we'll get a shine channel any time soon are just about zero.

Three of four years ago, I distinctly remember LL talking about implementing a full blown material shader system to replace the circa 1995 texture system we have now. In their words, it was to be "similar to the material shader system in Maya". Were that to have happened, I'm sure we wouldn't be having this discussion right now, as we'd be able to use custom maps for all kinds of things, and we'd all look back on the old texture days and laugh. But sadly it's just one more notch on the Linden belt of right ideas that were never implemented.

I'm really afraid the situation might not ever improve, to be perfectly open about it. My feeling is the Lindens wer so taken aback by the small but extremely vocal contingent of Windlight haters, that they've concluded graphics shouldn't be a priority. I really hope I'm wrong about that.

Lindens, if you're reading, I hope you understand that those same people will complain about WHATEVER you do. Please ignore them. Don't sacrifice the quality of SL just to try to appease the eternally negative. Thanks.

You can do this with invisibility prims.

That's true, Soap, but it's only because invisiprims don't use "real" transparency. As I understand it, they exploit a (deliberately unfixed) bug which reverses the draw order of certain things. As far as SL knows, invisiprims are not actually transparent. Therefore, the system has no problem alllowing them to be shiny.

The invisiprim texture itself looks like an opaque cloud pattern, no visible transparency in it at all.

Also regarding invisiprims: word on the street is that these will no longer work at some point in the future--presumably a point after transparent avatar textures, which should be Real Soon Now.

QFT

I noticed that in some office hours transcripts I read. There will be much gnashing of teeth followed by a shoe buying bonanza that day.

I fully agree... keeping the graphics at the same level forever is a horrible choice. I know some people hate Windlight (and I cannot understand why, WL is one of the most wonderful things that happened to SL) but improvement on the graphics needs to go further as well. That is very important, and I strongly hope Linden was not convinced otherwise. The lack of a modern rendering system, mesh support, etc. had gotten me thinking on using a game engine for making my own virtual space at some point, though I found out about RealXtend and was much happier

On LL I am upset as well now, and cannot lie about the fact I still consider them extremely incompetent. There are open source games out there (where the developers get no money at all except a few dollar donations) in which much more effort was put then in the SL client to have an up to date rendering system with advanced framerate improvement systems, distortion maps, offset mapping, etc, which I think should cause LL to ask theirselves a few questions and poke their cheeks, given what a big project they have and how much money they monthly ask for a sim.

At the same time I have to agree that the client changes they -have- made over the last years have always been generally good, though I believe they are not enough. Leaving some things old (like ancient rendering methods) upsets many residents including myself. Some changes are very much needed, although sadly, LL listening to their residents on implementing a new feature would be as much of a miracle as Jesus's second coming.



I really hope not, for the sole reason that this would brake dozens of current avatars using them, although I do encourage an alphamap avatar hiding system very much as well.

The best strategy I think would be projects on the scale of Windlight or Havoc 4, rather than developing smaller new features independently. The reasoning is that if you upgrade an entire section of how SL works at the same time, it's much more efficient and will work better.

Thus, it would be better if they developed an entire new materials system; improved texture mapping with perhaps custom UV coordinates and multiple texture layers per prim, normal and bump mapping (and maybe others such as specular/shine), and perhaps shading control, along with improved shiny.

However, to do this will probably require borking invisiprims. And so before anything like that happens, they need to create a viable alternative (which they are doing in the form of alpha masks for avatars) and wait for that alternative's use to become commonplace, and solutions used for legacy content. So the groundwork is being laid now.

In the immediate future, invisiprims do not work well with shadows, and so alpha masks are needed for the same reason.

It's just a matter of getting the foundation down so that future improvements can be made without breaking content on a wide scale (and there's tons of avatars and attachments that use invisiprims.)

As for large projects, it seems to be the case that LL is sort of embarking on a Havoc 4 scale project towards fixing bugs and improving SL. LLNet is a huge part of that.

It seems to me that careful use of textures might emulate shininess in transparent objects. All you need for an outside window is some kind of irregular horizon line and make it darker and maybe more transparent below that with a hint of sky and clouds above the horizon, then a hot spot for the reflection of the sun or moon or whatever -- the same way we might do a painting of chrome or glass in real life. You don't have to let the computer render everything for you.

After spouting off all that, I haven't tried this, but I'm sure someone has. I may try it right now in fact.