Consider removing mandatory color stop in interval [0,1]

[drott]

I don't think we need to enforce that there is a color stop defined between 0 and 1.

[rsheeter]

Revise to just there must be at least one color stop?

[svgeesus]

Right. Zero stops is undefined. One stop gives a solid fill and it doesn't matter where the stop position is. Two or more stops, the stop positions matter (but still need not be in the 0 to 1 range).

[PeterConstable]

@drott Current wording in 5.7.11.1.2.1 is:

A color line is defined with at least one color stop within the interval [0, 1].

Should the requirement to be within [0,1] be removed?

[PeterConstable]

@drott We need to close on this.

[behdad]

@drott We need to close on this.

Is clear what needs to be done.

[PeterConstable]

Conceptually, it doesn't matter. But is there any issue with what existing libraries can handle?

[davelab6]

@PeterConstable existing libraries for colr v0?

Colr v1 isn't finished speccing, so any existing colr v1 libraries should reasonably be expected to support whatever the spec finally says, imo

[PeterConstable]

But we were trying to accommodate (to some extent) existing graphics libraries. (Sorry, that's what I meant, not colr v1 libraries.)

[behdad]

But we were trying to accommodate (to some extent) existing graphics libraries. (Sorry, that's what I meant, not colr v1 libraries.)

Either do some reasoning yourself, or accept that the rest of us have done that.

This pattern of behavior of yours is what I have been objecting to since last year:

• The rest of us discuss and agree on something, then you show up and keep throwing roadblock after roadblock after roadblock in front instead of just Writing the Damn Thing Down.

• When you do write things down, you go ahead and CHANGE THINGS IN SEMANTICALLY SIGNIFICANT WAYS. When I spend my valuable time and find those and file issues and ask that they be reverted, you keep asking question after question and demand to be explained to.

I'm not going to accept and tolerate that behavior. If you have suggestions, you discuss it with others, on github, like the rest of us do. Only if there was consensus, you write them down.

Moreover, when I propose things like self-contained offsets, which solve REAL problems (eg. #293, #11), you just ignore.

You didn't understand why this pattern is problematic a year ago. And looks like you still don't.

So yes, I can explain to you why virtual gids are a problem. But I won't, because I do not condone and won't tolerate the above behavior.

[PeterConstable]

@behdad Your prejudice toward me is getting in the way of reason. The requirement to have at least one color stop defined within the range [0,1] was in the initial document commited by Dominik; I have left that as stated until now. Dominik raised the question in December; he, Rod, Cosimo and I discussed various aspects of the gradient definitions along the way and agreed on various changes, but we never closed on this issue. I have not changed anything; I have not created roadblocks; I am trying to discuss this now on GitHub as you suggest should be done, except when I do you object. This is not reasonable conduct.

[rsheeter]

Psychological safety is hard to achieve when we are forced by external factors to eschew face to face interaction. It will make everyones life easier if we assume good intent in others.

Taking virtual gids as an example, I didn't find the harm of them obvious either. Further, one of the things I've always deeply appreciated from the font community in general and @behdad in particular is courtesy around noob questions. It feels OK to not know things, to make basic mistakes, and that makes it dramatically easier to learn. I'd hate to see us lose that.

[davelab6]

But we were trying to accommodate (to some extent) existing graphics libraries. (Sorry, that's what I meant, not colr v1 libraries.)

Ah yes, right - so is what you are asking, is anyone aware of any graphics library that requires that there is a color stop defined between 0 and 1?

[justvanrossum]

Skia's and Cairo's color line semantics are quite different from what COLRv1 prescribes. In blackrenderer we calculate new points so we can convert COLRv1 to what the graphics libraries need. This recalculation of points does not require any color stops to be within [0, 1] in the COLRv1 model, even though the graphics libraries want everything to be within [0, 1].

So therefore, the [0, 1] restriction in "A color line is defined with at least one color stop within the interval [0, 1]" is IMO unnecessary.

[behdad]

But we were trying to accommodate (to some extent) existing graphics libraries. (Sorry, that's what I meant, not colr v1 libraries.)

Ah yes, right - so is what you are asking, is anyone aware of any graphics library that requires that there is a color stop defined between 0 and 1?

That's the wrong question to ask. If any graphics library has such limitations, implementation of COLRv1 on top of that library needs to add extrapolated points as necessary.