[Bf-cycles] Remove old Sky Model

[Sergey Sharybin]

Hi,

I'm not convinced in "getting rid of legacy algorithms" policy. There are
surely other algorithms which are more than a decade old in Cycles which we
simply can't get rid about. So where to draw a line?

Well, it's simple. Cycles is still a BF project and BF is primary
developing it for Blender and we do have quite strict policy in there: no
compatibility breaking without user-measurable good outcome. If Cycles ever
becomes a project on it's own you can re-consider this rule, for until then
just stick to it.

Thomas, now replying to some points which i find wrong.

> If we would get a more realistic Glass shader (at the same performance
level), no one would complain that the render looks different (aka
backwards compat breakage)

No one would complain for a new projects he (or she) is working on.
However, if you replace shader completely, then you'll most likely have a
complaint.

It does happen when during movie production when you need to go back to a
shot a re-render some frames. If they'll look different you'll be screwed
to re-render full shot. This even happened during open movies here in the
studio.

> I can add it back for 2.7x (to keep compatibility, as Ton suggested)

That's what we agreed on in IRC i thought

> but later I really like to remove it

Make it properly that time. I never told we don't have to simplify
settings, but cutting stuff away immediately you have a developer's
response in a list where we don't have enough Blender users is not a good
approach.

Make it a list of things which will be removed / replaced, get feedback
form actual artists. And only then make decisions.

There are quite some settings which could go into that simplification list.

> I don't want to end up with a Blender Internal 2.0, which has gazillions
of buttons and options after 5 years

Amount of buttons is not the issue of BI. If you ask artists, they love
buttons and knobs. The real issue of BI is that it's design stressed quite
a bit with all the extra effects which were added on top of simple
rasterizer.

You are not avoiding BI destiny by removing something which perfectly fits
into design.

----

Stefan,

I'm not sure what are we debating here about. Such a removal was quite
against our 2.7x roadmap. That's just given and nothing to debate about.

And at this point it's definitely not better for Blender to break things
(surely sometimes you can't move forward without breaking something, but in
that cases we are usually delivering something better).

On Tue, Apr 5, 2016 at 7:51 AM, Stefan Werner <stewreo at gmail.com> wrote:

> Hi,
>
> I know we’re getting into a philosophical debate here, but in my opinion,
> what was good clean code two years ago, will still be good clean code ten
> years from now if you don’t touch it - code doesn’t rust. For what it’s
> worth, should the Preetham model get removed, most likely we will keep it
> in alive in our Poser version of the Cycles source code, so that we don’t
> take away features from the user. If it’s better for Blender and its users
> to remove the Preetham model, we’ll be able to work with it.
>
> -Stefan
>
> > On 04.04.2016, at 18:11, Thomas Dinges <blender at dingto.org> wrote:
> >
> > Hi,
> >
> > The benefit is "get rid of legacy algorithms". If we would get a more
> realistic Glass shader (at the same performance level), no one would
> complain that the render looks different (aka backwards compat breakage).
> We kinda have the same argument here. Even in the Hosek paper, it directly
> compares with the Preetham model.
> >
> > I can add it back for 2.7x (to keep compatibility, as Ton suggested),
> but later I really like to remove it. I don't want to end up with a Blender
> Internal 2.0, which has gazillions of buttons and options after 5 years.
>
> _______________________________________________
> Bf-cycles mailing list
> Bf-cycles at blender.org
> https://lists.blender.org/mailman/listinfo/bf-cycles
>



-- 
With best regards, Sergey Sharybin
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.blender.org/pipermail/bf-cycles/attachments/20160405/f8839cd8/attachment.htm