Minor nitpick: The Web Content Accessibility Guidelines are precisely that - guidelines. You should follow them until you have a good reason not to. And contrast ratios arent everything.
This article https://www.bounteous.com/insights/2019/03/22/orange-you-acc... is a good example of how the "most accessible" colour pairing, by the "mathematical proof", is the least accessible and preferred by users.
There's a rather lengthy discussion of this on the WCAG issue tracker [1].
Look into the correspondence between the space Myndex uses, the space contrast ratio is measured in, and then again at the W3C docs: TL;DR Myndex's approach doesn't account for reflectance off the screen
WCAG has been adopted as the standard for a variety of regulatory purposes. For example, with Section 508/504, the US Access Board requires procurements to be compliant with WCAG 2.0 AA. Germany’s BIVT is harmonized with WCAG 2.1 AA, California requires state entities and procurements comply with WCAG 2.0 AA, etc.
At least in the US, the main sphere where WCAG has not been specified as the regulatory standard is with Title III places of public accommodation.
However, Federal Courts have repeatedly upheld that places of public accommodation must make their websites accessible, they just won’t say how to accomplish that (nor is it their job to).
In the case of GitHub, I can tell you with 100% certainty it has been provided to, at least, the Federal government as part of a procurement, is subject to Section 508 statutory requirements, and WCAG 2.0 AA is the required standard.
This is to say, it might not be a good idea to arbitrarily decide that color contrast sufficiency is just a suggestion.
The European Union requires its member states to comply with WCAG 2.1 AA on _all_ public websites. Every branch of government, every agency.
Anyways, I do think this is an important point: the math of contrast ratios is a model that’s intended to reflect what’s going on with people’s vision; if a large-scale study of people indicates that the white/orange button is easier to read, the math is just wrong: we have to be careful not to force reality into our models, otherwise the modes lose their utility.
The goal always to produce software that is accessable and usable by people, not to satisify a mathematical constraint. WCAG and 'constrast ratios' are a means to that end, but they're not the end goal itself.
Anyways, we’re not really disagreeing as far as I can tell.
It doesn't really matter because most displays aren't color-calibrated.
> While contrast ratios aren’t [everything](linked it here), they are a simple way to quantify the difference between two colors.