@David So even > 100k per week.
@Mareck Transforms on the view are actually a very basic byproduct of art and design thinking. It goes like this:
"OK we need to create an application for manipulating RGB pixels in a creative way. Blurring. Painting. Lighting. You name it."
"So we need a linear model?"
"Yes, otherwise all of our math is broken and the results are wonky."
"OK. So why don't we just linearize everything via the sRGB EOTF and call it done? 0.0 to 1.0 and everything is linear right?"
"Sure. But that is display linear. We want to manipulate photographs. So we need scene linear."
"OK. So we use scene referred linear."
"But we need to manipulate things. I can't manipulate things if 0.18 is so low on a curve now. If we make it nonlinear, I can."
"But if we make it nonlinear, we are back at the second point!"
"Oh. OK. So is there a way for us to have a scene linear reference space but see our work and have all of our controls nonlinear at times?"
"We can apply the transform on the view."
That is essentially why view based transforms rock.
Even better, and this is the precise point Blender is at a juncture to, we can support all workflows with educated pixel pushers. Consider the following example:
David is working on look development using Filmic. He wants to tweak his curves so that they are 1:1 with the view; he wants middle grey at the Filmic Log Encoding base 60.60% along the X axis 0% at the low value and 100% at the far right to be the 16.291 scene referred value.
Mareck on the other hand is tweaking normals for the wood floor. He needs a purely scene referred linear set of UI values.
Cegaton is working on a depth buffer from another application. This depth buffer is linear, but in the range of 0.0 to 50.0, so he wants his UI to reflect that range and not be constantly fighting it.
How can we support all three imagers carefully crafting work?
The answer is to permit the imagers to select their transforms per UI.
Based on the "suitcase" of configurations they are currently using in their small collaborative effort.
We aren't there yet with Blender. We could be if the culture picks up the core concepts in Filmic and thinks it through just as above's little discussion flows.
(And yes, the very problem with curves currently is precisely as above. The previous answer to the problem used to be "Use a transform and transform to the destination, do the work, then be careful to transform back" but that is horribly inefficient. Easier to have the UI tool to the mapping for you, and the reference never changes from scene referred linear.)
