RC5 Windows & Mac Next Week

[Ernest Burden III]

Now we are faced with the scenario that 1.0 may never be seen, especially if people start perusing legal action...

 

Which brings us back to what I and others have been saying since November, RC1-4 being a distraction. It's really hard to tell if we're even going forwards. Tyrone Marshall has been very positive about even the RCs, and I respect his opinion very highly.

 

So its roadmap or more of the same.

 

Another thing that may be useful is to compile a list of the 'must-have' features. Different disciplines have their own, for us its sun + glass probably above all others. If material layered coatings don't work, I can somehow live another day. Others feel the exact opposite. So if we can put together a list, by poll perhaps, of these features it will give NL some clues about what they should concentrate on first and at risk of failure of 1.0

 

The Maxwell project will not simply be abandoned. NextLimit has three options:

1- complete 1.0

2- go out of business

3- admit defeat, refund everybody (see #2)

[AdamT]

One thing to consider is that sunlight through glass *may* not be a total showstopper. If standard emitters work through glass it could be workable.

[yog]

One thing to consider is that sunlight through glass *may* not be a total showstopper. If standard emitters work through glass it could be workable.

I would respectfully disagree with this.

IMHO the only thing that substantially sets Maxwell apart from the competition is the quality of lighting produced by the physically correct sky.

 

Other renderers (specifically Vray) are begining to replicate MW's light distribution by brute forcing their own path tracing technologies. Also the new materials with coatings are nice, but renderers like Mental Ray through XSI, (I find MR in MAX much more limiting), can do a good approximation.

[_PopArt]

I think he was saying that if they can't get the sunlight working with glass, at least with a distant emitter one could still replicate sunlight, so that option would exist as a workaround, at least it would be usable. Of course, as you point out, the current sunlight system (which will hopefully be working) has enormous advantages, both in set up time speed and image quality.

[Devin Johnston]

I agree, the physical sky setup is a must, without it your back to faking everything which is the exact opposite of what Maxwell is all about.

[Ernest Burden III]

I think he was saying that if they can't get the sunlight working with glass, at least with a distant emitter one could still replicate sunlight

 

That was the point. But a distant small light is not the same as sun/sky, and its loss would indicate a significant failure of the promise of Maxwell. OK, so there could be a way to get a not-so-physically-accurate image out of physically accurate Maxwell. Up is down, war is peace? When you can't hit your one big selling point, there's a problem. And then what? it's back to 'wait for the next version'? If NL couldn't work it out in their three years of work so far, why would they be able to later? And then they either have to give us all a free next version or ask us to pay again for what we already paid for but didn't get.

 

I need a life.

[yog]

That was the point. But a distant small light is not the same as sun/sky.

That's my point. With Maxwell's physically correct sky it's not just the sun that produces the light, but the whole sky. This is also a major fact because Maxwell doesn't rely on fake specularity, but on different levels of diffuse reflections, so anything that reflects an outside environment is dependant on the physically correct sky.

[_PopArt]

That was the point. But a distant small light is not the same as sun/sky, and its loss would indicate a significant failure of the promise of Maxwell. OK, so there could be a way to get a not-so-physically-accurate image out of physically accurate Maxwell. Up is down, war is peace? When you can't hit your one big selling point, there's a problem. And then what? it's back to 'wait for the next version'? If NL couldn't work it out in their three years of work so far, why would they be able to later? And then they either have to give us all a free next version or ask us to pay again for what we already paid for but didn't get.

I need a life.

 

Me too. Anyways I agree with you on all points, but I'd just add that I doubt the algorithms that the skylight system are staggeringly complicated - in essence we're talking about algorithms creating a type of high dynamic range lighting with some pretty simple sky data (azimuth, ozone, et. al) and a distant source as the sun. I suppose the fact that the sun is considered as an absolute point might be causing some of the problems with glass and the path tracing algorithms. Or maybe the fact that it is considered so mathematically distant that it just makes rays passing through the glass painfully slow. It is definitely a good idea (everyone else is implementing it now), but often with hdr lighting you can get similar effects. Strange, though, a few years have passed since they've been trying to fix it:)

[AdamT]

Yes, near as I can tell the sky algorithm is based on this paper: http://www.cs.utah.edu/vissim/papers/sunsky/

 

I think the problem they've been having has much more to do with the sun than the sky.

[_PopArt]

Agree. And yes i think they are using the CIE sky model standard for irradiance calculations. RADIANCE has a good scene description plugin which does the same.

[bricklyne]

Yes, near as I can tell the sky algorithm is based on this paper: http://www.cs.utah.edu/vissim/papers/sunsky/

I think the problem they've been having has much more to do with the sun than the sky.

 

 

.....yeah, this might or could be supported by the fact that the physical sky seemed to work fine with the RCs' preview window, except with no Sun settings or Sunlight. And the fact that Victor consitently refuses to directly address this issue when it's brought up at the forums.

 

I'm beginning to believe that the Sunlight-dielectric issue might, after all, not be solvable; and to some degree NL probably know this and have been trying to find work-arounds themselves by implementing various Sky systems and even resortig to re-writing the entire code after the Beta. It might even be at the crux of their whole catalog of fiascos since then, as they've tried to resolve it.

 

I believe that eventually they might be forced (much to their chagrin) to adopt a "cheat" around this problem such as by implementing a distant emitter object ( small plane or sphere), at a fixed distance from the model and of relative incandescence to the rest of the sky using an algorithm to control the intensity and hue based on the selected time of day, for example. I once saw a render that someone (- I can't quite remember who exactly it was) did using the alpha version ( before the dielectric problem reared it's ugly head) and with an emitter as his sun as opposed to the actual Sun, and if I remember correctly his render cleared up much faster that way.

 

I think that currently NL might be so hung up ( and consequently, stuck in a rut) on the whole concept of physical "accuracy" and the "unbiased" nature of their renderer that they are dead set on solving this problem as opposed to pursuing more practical or applicable, yet 'inaccurate' alternatives that might work just as well, or if not better. Even though, doing so, would be an admission of defeat of sorts for NL.

 

Given the number of cracks that they've had at trying to tackle this issue without success, I just can't see them resolving it at all, eventually - particurlarly, given the extensive research that other developers have already performed on MLT methods as well.

 

In any case, I just speculating; particularly given the volumes deductible from what they are not actually telling us,or avoiding telling us. But like I said before, RC5's stability and functionability ( even without Sunlight) will reveal volumes as to the viability of the project's eventual success.

[Devin Johnston]

[quote=bricklyneI think that currently NL might be so hung up ( and consequently, stuck in a rut) on the whole concept of physical "accuracy" and the "unbiased" nature of their renderer that they are dead set on solving this problem as opposed to pursuing more practical or applicable, yet 'inaccurate' alternatives that might work just as well, or if not better.

 

In some of Victor's quotes he mentions that Maxwell is for use in Arch Viz as well as other scientific uses. I don't truly understand what "scientific purposes" means but I assume that it must be one of the primary reasons they are so dead set against using cheats to fix Maxwell. I personally don't care if it's 100% physically accurate as long as it delivers the same quality.

[AJLynn]

As best as I can see, emitter light and skylight through glass work but sunlight doesn't - and never has, regardless of changes to the core, complete overhauls and new materials systems.

 

WTH? I never understood this. Why is sunlight special and why is this so hard for them? I agree completely with Ernest - for architectural work, I need sunlight through glass.

[_PopArt]

In some of Victor's quotes he mentions that Maxwell is for use in Arch Viz as well as other scientific uses. I don't truly understand what "scientific purposes" means but I assume that it must be one of the primary reasons they are so dead set against using cheats to fix Maxwell. I personally don't care if it's 100% physically accurate as long as it delivers the same quality.

 

Well on the Nl site they state this:

 

"This technology can be applied to other fields related to optical simulation like:

 

 

 

Manufacture: lens simulation, jewel design, surface painting

 

 

Medical: visual perception, contact lens design

 

 

Defense: visual and infrared signature, camouflage design, sensor simulation"

 

I guess the defense industry doesn't appreciate workarounds:p

[AdamT]

I guess the defense industry doesn't appreciate workarounds:p

Yeah, and they have loads of supercomputers, so Maxwell should be quite speedy for them. :/