Luxrender direction...

[Carbonflux]

I thought I would like to very gently raise a subtle issue, which is where is luxrender going.

I have heard a lot of comments that luxrender development is slowing down or stagnating yet at the same time luxrender is being again and again recognized as one of the leaders in photo-real CG.

Are we too going to be poisoned by the growth metric which has trashed so much of the world?

Speaking as a artist I have almost everything I want from luxrender, I could use the current dev version for 10 years and still not be finished with the art I have in mind. I could use blender 2.49 in conjunction with this and never miss anything, the only reason I am switching to 2.6x is because I am to lazy to maintain the 2.49 exporter myself.

This is a complex issue relating to insecurity and the shinny factor but what does it have to do with art? Also, what does popularity have to do with interest in the technology?

When you work on the luxrender code base are you doing it because you are interested in the math or because you want to be rewarded with recognition and validation?

I don't see anything wrong with slowing down and just enjoying working on beautiful stuff.

With Cycles being in the main blender "trunk" I think we need to ask ourselves why are we doing this? Are the interest based values good enough or do we want to be recognized as being elite? I found the response to the SIGGRAPH thread very positive in that no one wants to go and talk about SPPM. To me this is almost Zen and like drinking clean water because it does suggest that the interest based value system is still intact.

So I close with the question, why is luxrender dev "slowing down" bad or even a valid idea?

Personally I am sick of this endless growth/popularity metric, lets think about doing art for a while.

[jeanphi]

Hi,

There are a few factors to take into account:
- for the past few months I've not had as much spare time as I would have liked, and some other devs were in the same situation
- to be able to add new features, you sometimes have to stop, reorganize the underlying infrastructure and move on on better grounds, that's what is currently happening with the overhaul of the material subsystem

So it might look like development is stagnating but the new layered materials are just the beginning of a broader evolution. We will also look into improving moving objects son, I'm working on cleaning up the light sampling schemes to offer more options and be more consistent between rendering algorithms...

Jeanphi

[Carbonflux]

Agreed, I have had a lot of ideas related to the UI but keep ending up getting distracted by "real life" issues. Must of the time when I allocate a day to work on lux I get distracted and pulled away by stuff going on around here. Also I have to admit the economic "downturn" stuff means I actually have to work sometimes, I hate admitting I have to work

I guess my point here is that I think its ok to slow down a bit, it does not mean we are dead.

[moure]

As a bit of an "outsider" since im not a core developer
I have noticed a bit of "slow down" in new features addition, BUTi think thats something it must happen.
Personally i dont think that Luxrender lacks many features comparing even with other commercial projects. I believe what it needs is optimization of the current core code and features. Make them more stable, faster, more clear to the mid-low user. And the latest tries with the material system and the other changes are in that direction

To sum it up i dont think that luxrender needs another 10 integrators, or 15 filters or 20 accelerators
on the bigger image it just needs to be as "simple" and as fast as possible, without adding new features all the time. And im happy that the developers move towards that road

[dougal2]

I'm too very busy with real life recently. I've been trying to keep my hand in with a few small contributions here and there - however the bigger picture of LuxBlend25 and Blender 2.6x ++ is that there is an expectation that LuxRender will eventually have the full integration like Cycles. I don't mind working towards that expectation, after all it'd be a damn cool thing to have (and to have helped create) - however the reality is that at present, Lux core isn't capable of such integration and also the new Blender Render API is still what I would consider too unstable to build on (anyone remember when lb25 first came to life and the perpetual headaches we had?!) and at the present time poorly documented.

I agree with CarbonFlux with regards to what we have already though - it works, it's reasonably fast, it's fantastic quality, go and create some images...

[Carbonflux]

I think its interesting that luxrender dev has "slowed" at the same time the global economic crisis has emerged, it seems we all must face the reality of having to generate money to support our real interests, I posted this thread in part in the hope that we might think about how larger macro-economic issues might effect the appearance of luxrender progress.

My ideal is that luxrender will always be here, that we can work on both art and interest moving forward but outside of linear time, when someone wants to contribute fine, but that we don't evaluate luxrender in terms of a metric that is based on a linear time growth/popularity measure.

I submit we should look at development as a refinement of a valid physical simulation driven by interest in physics and art.

[Dade]

I would split the "problem" in few parts:

- time: it is not trivial for people to find time to work on something for free. It is an intrinsic problem of any free project. It is something we have always faced, we are facing and we will always face. Not a big deal, it is one of the rules of the "game". It is also physiological for people spend more time on the project during some period and less on some other one.

- perspective: if you consider when the first line of the BiDir integrator was written and when it was "finished", you realize that major features often require a lot of time to be finished (i.e. years). This is somewhat relate to point discussed above. I would like also to add that some of the thing people have written for LuxRender are not exactly trivial and/or easy to write.

- stable Vs experimental: LuxRender is a bit double faces, stable code is what is supposed to be used to produce beautiful renders, etc. The development of experimental stuff is instead, very often, erratic and a bit random.

- fatware: LuxRender may even have too many features, I wouldn't worry too much if we don't add 50 new features every year (i.e. it is not like we have to sell new licenses).

On the personal side, since the development of my personal interest on GPUs, I have become a bit minimalistic. I have worked on a new toy recently (i.e. a mini real time path tracing demo) and I must admit, every time I have some free time I end to work on the new toy instead of hybrid BiDir at the moment.

[patro]

- time: it is not trivial for people to find time to work on something for free. It is an intrinsic problem of any free project. It is something we have always faced, On the personal side, since the development of my personal interest on GPUs, I have become a bit minimalistic. I have worked on a new toy recently (i.e. a mini real time path tracing demo) and I must admit, every time I have some free time I end to work on the new toy instead of hybrid BiDir at the moment.

Dade.
will you present the new toy asap ?

[SATtva]

I must admit, every time I have some free time I end to work on the new toy instead of hybrid BiDir at the moment.

Sad.

[paco]

I think the slowdown is mostly due to all the low-hanging fruit being picked already.

A lesser issue from my experience in the core is that the deviations from PBRT and the lack of documentation make things very difficult for new devs. I think its inevitable that the current devs will drift on to new projects (good luck Dade!). If jeanphi left then we'd be sunk because I don't think many (?any) others understand what happens in the core anymore - and hence anything involving it hangs around for him to fix - which is a large drain on his time.

Still, as pointed out, Lux pretty much does everything and I personally don't see the slowdown as a problem.