Noam Rosenthal
@nomster.bsky.social
Web platform engineer @ Chrome
In other other words, 0x80 black on a white background *substracts* 0x80 from all the 0xff channels of white, so (0xff - 0x80 = 0x7f).
February 3, 2026 at 9:57 PM
In other other words, 0x80 black on a white background *substracts* 0x80 from all the 0xff channels of white, so (0xff - 0x80 = 0x7f).
The formula is:
(foreground * alpha/0xff) + (background * (1 - alpha/0xff)
foreground = 0, white=0xff
So
(0 * 1) + (1 * (255 - 128) = 0 + 127
In other words, the max is 255 and not 256, and what's multiplied is the background with the inverse, so you get 127 and not 128.
(foreground * alpha/0xff) + (background * (1 - alpha/0xff)
foreground = 0, white=0xff
So
(0 * 1) + (1 * (255 - 128) = 0 + 127
In other words, the max is 255 and not 256, and what's multiplied is the background with the inverse, so you get 127 and not 128.
February 3, 2026 at 9:52 PM
The formula is:
(foreground * alpha/0xff) + (background * (1 - alpha/0xff)
foreground = 0, white=0xff
So
(0 * 1) + (1 * (255 - 128) = 0 + 127
In other words, the max is 255 and not 256, and what's multiplied is the background with the inverse, so you get 127 and not 128.
(foreground * alpha/0xff) + (background * (1 - alpha/0xff)
foreground = 0, white=0xff
So
(0 * 1) + (1 * (255 - 128) = 0 + 127
In other words, the max is 255 and not 256, and what's multiplied is the background with the inverse, so you get 127 and not 128.
Old has always been the new new
February 3, 2026 at 2:55 PM
Old has always been the new new
Also, let's continue this party at github.com/w3c/csswg-dr... ?
[css-borders-4]: `border-shape` order of shapes in the two-shape variant · Issue #13308 · w3c/csswg-drafts
Currently the spec for border-shape: <<basic-shape>> <<basic-shape>> defines that the first shape is the outer one and the second shape is the inner one. I don't know if this was 100% thought out. ...
github.com
February 1, 2026 at 8:41 PM
Also, let's continue this party at github.com/w3c/csswg-dr... ?
Note that those have meaning beyond adding and removing - the inner shape is the overflow area for the contents and the outer shape interacts with shape-outside.
February 1, 2026 at 7:17 PM
Note that those have meaning beyond adding and removing - the inner shape is the overflow area for the contents and the outer shape interacts with shape-outside.
Yea we could bikeshed something better for this for sure.
February 1, 2026 at 7:11 PM
Yea we could bikeshed something better for this for sure.
I wonder if we should use keywords instead of rely on the order, e.g. `circle() enclosing square()` or `circle() clipped-by square()` or some variation.
February 1, 2026 at 10:19 AM
I wonder if we should use keywords instead of rely on the order, e.g. `circle() enclosing square()` or `circle() clipped-by square()` or some variation.
See related poll: bsky.app/profile/noms...
Which of these is circle() square()?
(Don't look at current spec, just think of what feels right)
Poll:
indieweb.social/@nomster/115...
or x.com/nomsternom/s...
(Don't look at current spec, just think of what feels right)
Poll:
indieweb.social/@nomster/115...
or x.com/nomsternom/s...
February 1, 2026 at 9:26 AM
See related poll: bsky.app/profile/noms...
... Latest canary with experimental features turned on ❤️
January 30, 2026 at 7:15 PM
... Latest canary with experimental features turned on ❤️
Oh sorry, joke went over my head!
January 29, 2026 at 1:15 PM
Oh sorry, joke went over my head!
Hmm no, this is about implementing what's under the hood of the new CSS corner-shape property, where the engine that's under the hood is blink (chromium).
January 29, 2026 at 8:47 AM
Hmm no, this is about implementing what's under the hood of the new CSS corner-shape property, where the engine that's under the hood is blink (chromium).
btw the hard bit in writing this is generating meaningful diagrams that show case some of the rendering and mathematical decisions, and LLMs are not very good at this yet. So I can't offload this writing task to Gemini unfortunately.
January 28, 2026 at 9:10 PM
btw the hard bit in writing this is generating meaningful diagrams that show case some of the rendering and mathematical decisions, and LLMs are not very good at this yet. So I can't offload this writing task to Gemini unfortunately.
Yea that is a more probable concurrent use case for this kind of compat...
January 28, 2026 at 3:00 PM
Yea that is a more probable concurrent use case for this kind of compat...
Mind filing a bug and I will CC the appropriate people?
January 23, 2026 at 7:59 AM
Mind filing a bug and I will CC the appropriate people?
You're preaching to the choir in my case.
January 22, 2026 at 11:37 AM
You're preaching to the choir in my case.
This is more of a pushback since it's already shipped in webkit and we agreed to it at some point in the past though never shipped it.
The justification was that attr() brings with it a lot of complexity and verbosity (fair) and that matching by ID by default is "obvious" (I beg to differ).
The justification was that attr() brings with it a lot of complexity and verbosity (fair) and that matching by ID by default is "obvious" (I beg to differ).
January 22, 2026 at 11:32 AM
This is more of a pushback since it's already shipped in webkit and we agreed to it at some point in the past though never shipped it.
The justification was that attr() brings with it a lot of complexity and verbosity (fair) and that matching by ID by default is "obvious" (I beg to differ).
The justification was that attr() brings with it a lot of complexity and verbosity (fair) and that matching by ID by default is "obvious" (I beg to differ).
I don't know if "the CSSWG" is pushing for it.
I preferred that name over "auto" because it's slightly less enticing but I still prefer to not have anything that has automagic attr-like capabilities.
I preferred that name over "auto" because it's slightly less enticing but I still prefer to not have anything that has automagic attr-like capabilities.
January 22, 2026 at 11:13 AM
I don't know if "the CSSWG" is pushing for it.
I preferred that name over "auto" because it's slightly less enticing but I still prefer to not have anything that has automagic attr-like capabilities.
I preferred that name over "auto" because it's slightly less enticing but I still prefer to not have anything that has automagic attr-like capabilities.