Terminals should generate the 256-color palette (gist.github.com)
452 points by tosh 16 hours ago
johncoltrane 15 hours ago
The good thing with the 256c palette is that colors in the 16-255 range are fixed, which gives us a very high level of confidence that 146 will be a muted violet and so on. This is very useful for colorscheme developers because it allows us to provide a pretty good and consistent experience across the widest range of terminal emulators.
If the 256c palette is generated from a -- potentially wild -- 16c palette then there is no guarantee anymore that 146 will indeed be the 146 I expect.
Turning 16-255 into the same kind of minefield as 0-15 seems very misguided to me.
hnlmorg 14 hours ago
I know 16 colours is limiting, but one of my biggest pet peeves is CLI / TUI developers creating their own custom themes using colours outside of that because odds are, they’re going to generate a colour scheme that is harder to read for a lot of people with visual impairments, people who prefer a white or coloured background for eye comfort, people are dyslexic and find non-black backgrounds easier to read, and others with visual difficulties, reading difficulties, or those who just like a different colour scheme per project or environment they’re working in so they can multitask more easily.
And the developers answer to this loss of control is to create multiple colour schemes and allow the user to configure the app. Which then means their users have to set up their terminal defaults and then configure every fscking app that ignores those terminal defaults(!!!) instead of defining their required colour schemes once in the terminal.
People use the terminal not because it’s pretty but because they find a text interface more efficient. If you want to make pretty things then build a web frontend. But please do not break my terminal because you feel the need to impose on me your own favourite colours of the hour.
mort96 13 hours ago
I agree. I always customize the blue color on my terminal because dark blue on black is completely unreadable to me (and I'm not even color blind!). For some reason, every single terminal emulator defaults to a blue that's unreadable on a black background (I think typically #00f).
If a tool overrides my color settings, it too usually picks a dark blue that's unreadable on my black background.
kevin_thibedeau 10 hours ago
hinkley 4 hours ago
jve 9 hours ago
ErroneousBosh 5 hours ago
zzo38computer an hour ago
If I make a terminal emulator, I would probably not support true colours in text but only in pictures (and it would be possible to disable pictures). I probably would not implement 256-colours either, because I agree with you; the user configures their own colours to use, and the terminal should use them. For text, only sixteen colours can be used.
Similarly, the user can also set their own fonts for the terminal, just as you can with colours and other functions. (However, some programs will have a reason to configure the fonts and palettes for the specific use, although most won't and even if they do the user might disable those features.)
A program can have an option (possibly by environment variable) to disable colours entirely; this might be necessary even if you can disable colours in the terminal emulator, because a program might want to use such things as reverse video, underlined text, etc to indicate some things when colours are disabled. (Disabled colours can also be useful if you do not have a colour display or if you want to print out a screenshot without a colour printer.)
ffaser5gxlsll 13 hours ago
> but one of my biggest pet peeves is CLI / TUI developers creating their own custom themes
An even bigger one is hardcoding black and white instead of using foreground/background and use reverse when needed.
Aardwolf 4 hours ago
gdb (the debugger) sometimes prints dark blue text on a black background, which is unreadable. I customize the dark blue color when possible in VTE's due to this, but it's not always possible everywhere
But I wonder what the developers of gdb were using that made them not notice this
AshamedCaptain 12 hours ago
Sadly, if users start customizing the 256 color palette, developers will simply switch to true color to continue this mess further...
heliumtera 10 hours ago
yes, this. the reason i use wolf (firefox fork) is because i can easily disallow styles overriding, font overriding, and use bitmap fonts. nothing else.
now that react devs finished destroying the web they have to impose their superior taste into terminal through 800mbs TUIs.
donatj 12 hours ago
> provide a [...] consistent experience
Please just don't. This is not the web.
Color usage in the terminal should be largely semantic, not stylistic.
Speaking for the group of people I know and work with, we don't want a "consistent experience" and hate TUIs that try to manhandle the color palette. Use color sparingly and with intention. Respect that different people have different settings.
johncoltrane 11 hours ago
First, I make third-party Vim colorschemes, not app. People install my colorschemes because they like the colors, not because I'm a monster with a gun pointed at their face. No one is harmed. No one is forced to do anything they don't want.
Outside of my text editor, where colors matter a lot to me for syntax highlighting, I'm definitely in the NO_COLORS camp (and in the NO_EMOJI camp, nowadays).
> Color usage in the terminal should be largely semantic, not stylistic.
I wholeheartedly agree but 0-15 sadly have zero inherent semantics, which is the single reason behind every terminal colors-related drama since forever: developer choses 9 to highlight an error message because it is generally a bright red by default --> user sets 9 to whatever makes sense to them --> error message is illegible.
ryandrake 9 hours ago
makapuf 4 hours ago
tambourine_man 9 hours ago
pjmlp 7 hours ago
I thought we had left TUIs behind in the 90's, instead it is rainbowns, unicorns, and whatever else going with those stylisic experiences.
ryandrake 9 hours ago
Yes, and giving developers control over colors, text size, typeface and so on has also been a usability and accessibility disaster on the web, too! The user should have this control.
tasuki 11 hours ago
> which gives us a very high level of confidence that 146 will be a muted violet
Is there anything you can do with that information though? This piece of information only becomes useful if you know what colour the background is. And you should also know the colour of text and everything else.
What if the background is muted violet? What if the background is white and the foreground is muted violet? I don't want you to ever use "muted violet" in my terminal, since you have no idea what colours there are in my terminal.
johncoltrane 9 hours ago
I make Vim colorschemes so I'm fully in control of all those aspects. If I decide I want a given token in muted violet it is because I know that it will works well with the background… that I have defined myself.
The exact values of your 0-15 don't matter to me and they don't matter to you either, because you chose to use my 256c colorscheme to begin with.
pkulak 5 hours ago
lucideer 7 hours ago
> This is very useful for colorscheme developers
I would posit then that this article simply doesn't apply to you at all. The feature being described is targetted at users who are effectively developing their own schemes (albeit in a rather simplified automated manner). If I'm taking something off the shelf, I'm using the appropriate recommended base16. I have no expectation that a wild base16 is going to align with any 3rd-party's curated scheme.
I do understand that this logic isn't always going to click with people given the differing contexts of a terminal-wide -vs- app-specific (i.e. vim) approach, but again: that disparity seems either a legacy issue (caused by things like this 16-256 mis-alignment) OR simply a philosophical difference (whereby people who customise at the term level shouldn't at the app level & vice-versa).
lloeki 14 hours ago
Terminals like iTerm2 have had a Minimum Contrast for a while that messes with (foreground) colours, sometimes very badly.
shimman an hour ago
For far too long I'm ashamed to admit, I would use vim with the adventure time theme on iterm2. Looking at it now I'm shocked my eyes didn't bleed more. I think the worst part was that visual mode was neon yellow bold text with neon pink bg. Relying on visual mode quite a bit while using vim I was self inoculating my subconscious with stills of a poor Jackson Pollock imitator while on multiple different amphetamines. Hopefully I find out my resistance soon.
Gormo 9 hours ago
> and consistent experience across the widest range of terminal emulators.
Instead of aiming to provide a "consistent experience", you should instead prioritize providing consistent functionality, while avoiding impeding users' control over their own particular experience.
johncoltrane 9 hours ago
No, because I don't develop programs. I make third-party Vim colorschemes that users install because they like the colors. There is no impending happening, here.
tambourine_man 9 hours ago
I think it's an interesting idea, but should be set as an option and default to off, for the reason you describe.
If the user sets a sensible 16 color palette though, many old utils could look great out of the box. I'm enticed by the idea.
ryan-c 14 hours ago
There are escape codes that can re-define palette entries. Usually including the 16-255 range.
johncoltrane 13 hours ago
Yes, I know that (see the venerable https://github.com/trapd00r/colorcoke, etc.) but those tricks are not used widely enough for them to be a concern. Using those tricks is also a deliberate choice so it is definitely on the user if my lovingly crafted 256c colorscheme is broken.
Having all terminal emulators run the equivalent of colorcoke without asking the user is not a very bright idea.
vova_hn2 12 hours ago
I'm sorry, but I find this mentality from app developers extremely annoying.
I personally prefer light themes everywhere, both in IDEs and in the terminal. I thought that just choosing my own color scheme for 0-15 would give me the color pallette that I prefer, but because app developers like you for some reason decided that you know better what colors do I prefer, this is actually not enough. I also have to configure each TUI application separately to have the color scheme that I like.
And I do not understand why people do it. Like, why would you deliberately break the universal customization system and force users to use your own, specific to your app?
Honesty, each time I encounter an app that uses 16-255 colors, I feel like someone just violated my personal space and intruded into my chosen color pallette with their own colors that don't fit.
johncoltrane 12 hours ago
I'm not an app developper. I make third-party colorschemes for Vim, which I assume are downloaded, installed, and used by people on their own volition, after they have looked at, and liked, the screenshots. Moreover, I take great care to make sure they are still usable in 16c, within reason.
Because all my work is based on 16-255, I can actually guarantee to my users that, given a properly configured terminal emulator, they will get the colors on the screenshots.
If I can't rely on 16-255 to be fixed anymore, then I won't be able to make any promise anymore. In practice, it just means adding a caveat in the README.md, but I'd prefer not to. Here's hoping this breaking change gets hidden behind a checkbox/flag.
gzread 6 hours ago
poly2it 11 hours ago
eddyg 9 hours ago
+1
We use an ansible task to ensure SYSTEMD_COLORS=16 is in /etc/environment on every system and it at least solves that problem...
jauntywundrkind 15 hours ago
This will be fascinating to see in practice, with ghostty for example shipping these changes! I expect that the concern you have here will largely be for naught, with some exception. What are some terminal apps you think might be affected, what are test cases?
I didn't read in fully, but what I was thinking in my head is not that we would just totally replace the rest of the colors with arbitrary palette. But that we would sub in better versions of the palette that also used user colors as the base. What was magenta is derived from what the user picked from blue and red.
There's always been such tension between design/creative and users. Apps & designers want their own brand identity, want creative control to make things just so. And are willing to throw user preference & desire on the pyre to get that exacting control. Personally that was always rubbed me extremely the wrong way; I would way rather allow some weirdness & funkiness in, if it leaves the user in control. But I understand the risk aversion, understand the Murphy's law corporatism that makes people and companies want to build strong laws that forbid anything but strictly approved systems, for fear that things go wrong. I understand. But I also things that's a dogshit world to live in.
johncoltrane 14 hours ago
I have a bunch of Vim colorschemes under my belt.
0-15 are, as I said, a minefield because they are user-customizable: there is no guarantee whatsoever that my user's 1 will be the same dark-ish red as mine… or that it will be dark-ish… or that it will even be vaguely red-ish. It is actually somewhat fun to design colorschemes within those crazy constraints but oh well.
On the other side of the spectrum, truecolors is a nice idea in principle but support is still spotty and inconsistent. In theory, this gives me, the designer, full control over the colors used in the UI, which is a good thing for us and for my users. In fine, if I want my colorscheme to be usable by most users, then I can't blindly rely on this.
Which leaves me with 16-255, which are more widely supported than truecolors and, more importantly, dependable. They have problems, as mentioned in the article, but their _fixed_ nature gives me confidence that the background color of the status-line, for example, will look exactly the same -- and exactly how I want it to look -- in all my user's environments. Which, again, is good for my users and for me. Losing that confidence is what worries me, here.
Like you said, maybe 146 will still be a muted violet —— just not exactly the same -- but I'm not sure about this and I think that, at the minimum, this "feature" should be put behind a checkbox/flag.
wredcoll 6 hours ago
kristopolous 9 hours ago
My streaming markdown renderer does something like that https://github.com/day50-dev/Streamdown
You define a baseline color in HSV and then everything else is a multiplier of that
For example
[style]
HSV = [0.7, 0.5, 0.5]
Dark = { H = 1.0, S = 1.2, V = 0.25 } # Make dark elements less saturated and darker
Symbol = { H = 1.0, S = 1.8, V = 1.8 } # Make symbols more vibrant
As a result you can simply move around the HSV to your preference in the config and things don't look like garbage where you have to hand tweak every color to get something legible.
For example, this simple loop https://github.com/day50-dev/Streamdown?tab=readme-ov-file#c...
It's effectively a swatch
tasuki 11 hours ago
I had exactly this thought!
Though, I have a problem with even just the basic 16 colours:
black red green yellow blue magenta cyan white bright black bright red bright green bright yellow bright blue bright magenta bright cyan bright white
~~~
Many themes take `black` to mean `black` and `white` to mean `white`. How is it supposed to work when one switches the theme between the dark and the light version?
What are `black`, `white`, `bright black`, and `bright white` supposed mean?
I take those as meaning (in order): `almost invisible on current terminal background`, `pretty contrasty`, `visible but not contrasty`, and `the contrastiest`.
I wish the colour names reflected that, instead of `black` and `white`: you usually care about the contrast, not about the precise colour.
Ferret7446 10 hours ago
The way it's supposed to work is you don't touch the terminal colors but instead configure the program to use different colors. So you don't invert black for a dark theme, instead the program uses white instead of black.
By the way, the terminal foreground and background colors are independent of the standard 16 colors, which complicates things.
tasuki 2 hours ago
> The way it's supposed to work is you don't touch the terminal colors but instead configure the program to use different colors.
That makes sense, but... seems infeasible in practice? "The program"? I tell you what program I need to configure: all of them! Ideally even the programs I'm not even using yet.
I think I'll continue doing it my wrong way then: redefining the meanings to mean less/more contrasty. That allows me to effortlessly switch the terminal between light[0] and dark[1] backgrounds. I've done worse things: I also hijacked a part of the 256-colors-cube for my own purposes...
[0]: https://github.com/tasuki/dotrc/blob/master/.config/kitty/so...
[1]: https://github.com/tasuki/dotrc/blob/master/.config/kitty/so...
skydhash 9 hours ago
Unless you provide a way to configure the colors, it’s better to simply rely on bold, reverse, and standout. If you using the 16p, don’t use black and white unless you know ehat the background is, and use the others sparingly and semantically (no rainbow soup). If you’re going for 256p, it should be in a theme engine that’s user configurable.
kccqzy an hour ago
The word “should” is way too strong. It’s ultimately just a personal preference. For me I would prefer a two-tier approach.
There are different programs inside the terminal that I would prefer to have different color treatment. A full-fledged editor like emacs should have full 24-bit color support because I have already configured it to have my preferred color schemes and I prefer to switch themes inside emacs. On the other hand, almost all other TUI programs should not be permitted to use more than the traditional 16 colors. I haven’t configured them myself so I don’t trust them to make good choices.
In other words different terminal capabilities for different programs.
rbanffy 2 hours ago
I’m concerned not everyone have their base colors set sensibly - as in “there’s no guarantee the base garish RGB green is green on this machine”. Maybe the right thing would be to put the color closest to a base color on the corresponding corner of the RGB cube, but that’s also not ideal - I have had terminal palettes that were all green or all orange/red, or the green/yellow of EL displays.
Maybe applying the saturation of the base set across all the generated palette would also work.
jimrandomh 16 hours ago
Yeah, when you point it out, this makes complete sense and every terminal should probably add this feature. I think I would generalize this to 24-bit color as well; 16 colors isn't enough to identify a unique tonemap, but if you fiddle with the parameters a bit I think it shouldn't be too hard to come up with something hacky that works.
Although, this should probably be optional (both as an option for terminals to have in their own settings, and via an escape sequence that opts out), because some users will have configured some programs with a color scheme that they don't want transformed. For example, if your terminal uses the Solarized color scheme, and your text editor _also_ uses the Solarized color scheme, then this could lead to double-applying a color transform and getting something odd.
Sardtok 14 hours ago
Interesting. You could build a LUT from the 16 color palette to map the 24 bit color space to something 24 bits or less. A bit like mapping 10 bit HDR to 24 bit sRGB.
Perhaps instead of the application overriding the setting, it could be done with an environment var, so the user can easily override it if the mapping messes with the look/usability of the program.
k3vinw 15 hours ago
It’s a messy situation for sure and what lead me to discover tinted theming: https://github.com/tinted-theming/base24/
It’s been a fairly decent stop gap measure. I use tinted shell to switch between color schemes.
epage 10 hours ago
Really interested in this for cargo/rustc. We run into issues where we need one or two more colors but all that is left after going for the basic semantic colors is magenta (odd choice), and black and white (not safe without inspecting the theme).
skydhash 9 hours ago
Other than red and green, for semantic reasons, a CLI tools shouldn’t bother with colors. And it should support text markers instead of colors for the occasional piping.
estebank 6 hours ago
That's what cargo/rustc do: all of the information is conveyed through text, but colors are used to make it easier to scan. It ends up using almost every usable base-16 color: red for the error marker and their primary spans, yellow for warnings and their primary spans, blue for secondary spans, the gutter separator and line numbers, bold text for the main message, cyan for notes and their primary spans, green for helps and their primary spans, green and red for structured suggestions, bright magenta for very specific hand-picked highlights (used to be bold/bright white, but that wasn't obvious enough in lots of terminals, used semantically equivalent to bold or italics in text, the case most people will encounter these in is in E0308 type errors where a specific type parameter didn't match) and white for anything else. Note that of the mentioned output the only thing that is differentiated in color only is the message highlights, and that's because it is generally done for either extra clarification ("Trait is not implemented for Foo, but it is implemented for Bar") or within code that we wouldn't want to pollute with additional non-code ASCII markers (people should be able to copy&paste those types without causing parse errors). Some people deride this variety as being too "christmas tree", but IMO it works quite well for scanning output. Color helps a lot when scanning longer lists and dense output. Colors might not help much if the tool is emitting a single line of output for each event, but rustc emits dozens of lines per error. And no, reducing the amount of output would not be the answer for rustc, you can already use `--message-format=short` if you want that, or try/contribute to one of the TUI projects consuming the json output for dynamic navigation of the output (to expand and contract sections). rustc and cargo default configuration have to cater to the most common case, while also keeping the number of non-essential features (like a built-in TUI) to a minimum.
xvilka 10 hours ago
Just use direct color mode (24bit, "true color")[1] and there will be no need for a palette.
> Fewer terminals support truecolor.
From what I know all modern terminal emulators in all operating systems support it now.
grimgrin 9 hours ago
they explicitly had an opinion about true color, right in the article we’re discussing
eqvinox 5 hours ago
> From what I know all modern terminal emulators in all operating systems support it now.
urxvt prominently doesn't (fully)
kristopolous 9 hours ago
Doesn't work. You don't have full control over every color all the time and you're going to make assumptions. That's kinda the point.
globular-toast 9 hours ago
The entire point is to not use true colour but rather a user-configurable theme, but with 256 colours.
yourapostasy 9 hours ago
We're probably close enough to a future with the technology and engineering able to implement it, to justify designing 48-bit perceptual bit depth standards. Optimistically presuming breakthroughs enabling in vivo biological upgrades to our eyes to match those of mantis shrip, we could design 160-bit standards knowing that is a "proven" biological technology capability. That gets within the same galactic supercluster of the known limits of physics limits.
At the currently known limits of physics where Heisenberg Uncertainty Principle, Abbe's Limit, Quantum Shot Noise and such become our sensing barriers, we suspect we need only about 6000 bits per pixel to represent a digital twin of the electromagnetic field of the sensor-covered volume of space. At 60 fps, that is 1.8 Zettabits per second. Scale out data volume accordingly when using using 18.5 sexdecillion fps (Planck time fps).
What surprises me is these "limits of the fabric of reality as we know it" mind experiments fairly concretely point the way on the many roads towards Kardashev Scale implementations, and is not that different from Archimedes' "Sand Reckoner" and Hindu cosmological Kalpa time scales. History doesn't quite repeat, but rhymes yet again.
igneo676 9 hours ago
Oddly enough, I'm colorblind and I have had the worst time with color schemes. Many are completely unreadable and lack sufficient contrast. Others I just don't like
So, I've started using AI models to generate color schemes for me.
Take an unreadable theme I like and generate one that is high contrast and still coherent. It's probably not good enough for a full spectrum vision person to use but wow has it improved my quality of life.
It wouldn't surprise me if this is exactly the type of problem that is solved in the same way for the rest of you
thayne 5 hours ago
> Many are completely unreadable and lack sufficient contrast.
I am not colorblind, and have the same experience. It's made worse by the fact that different apps use colors differently, so if I find a theme that works well for some clis and tuis, there will be others that will have unreadably low contrast with the default theme. So I have to tweak the colorscheme for each application as well.
eqvinox 5 hours ago
The problem with this change is that it breaks setups for people who have already adjusted their color schemes to work with the 256-color palette as it is. It's now essentially double-adjusted.
(…and people wonder why I use gvim rather than vim…)
vanderZwan 12 hours ago
Sounds like I should try this together with ametameric, which I've been using since I have protanomaly
urmane 7 hours ago
I see TFA mentions Solarized. I've been using Solarized light for longer than I can remember, even on Windows where I can (eg VSCode), and it goes a long way to making my eyes happy. I share the irritation with dark blue on black and can't stand dark mode (maybe I'm now old and no longer 31337 ... alas ...)
For my ncurses brothers, saw this but have not used yet: https://github.com/dankamongmen/notcurses
For my plan9 brothers: one day we will have a new acme, written in common lisp and displaying all the correct colors, and lo it will be glorious.
nubinetwork 12 hours ago
Kde has their own palette, and I hate it, the colours are all wrong, and the scheme is fugly... I usually set it to "Linux default" and hope its good enough to read
makapuf 9 hours ago
We should define a set of base colors for terminal apps that are used for themes so that we have a common set of colors for all term apps. Text, background, borders, hilight, muted then let the terminal set its theme.
Andrex 7 hours ago
Wild, potentially stupid thought: why couldn't a terminal let users supply a user.css like browsers? They'd only have to support the small subset of text styles.
skydhash 3 hours ago
Color 0 to 15 are user-customizable by all terminals I know of. Color 16 to 255 are standardized, and the same across all terminals. Most (good?) cli and TUI use the first sets, while any that uses the second set usually are used through a theme engine (per app customization though).
aragilar 15 hours ago
This definitely seems like a sensible starting option to generate 256 colours from a custom set of 8 (and then let the really pedantic users fiddle with the extended set). I would presume for "standard" themes these values would be pregenerated and adjusted slightly if needed.
jmbwell 9 hours ago
Color schemes voluntarily added by the user to an app like vim, great.
All the more reason for developers to keep the app itself responsive to the user’s environment by default.
Don’t bake in elaborate visual choices. It’s a usability thing first and a style thing somewhere much farther down the list.
Keep it simple from the factory. Don’t get in the way of customization. Let the user’s environment do the work of adapting it to the user.
cyanbane 3 hours ago
Should call it Super Terminal Entertainment System.
stuaxo 9 hours ago
This new palette should be behind a new control code, that should sort out the issue of opting in.
DiabloD3 6 hours ago
And this is why we have the Tc extension that terminals must implement now: I just use the 24 bit value I want directly.
thayne 6 hours ago
The OP specifically says why that isn't ideal:
The mainstream solution is to use truecolor and gain access to 16 million colors. But there are drawbacks:
Each truecolor program needs its own theme configuration.
Changing your color scheme means editing multiple config files.
Light/dark switching requires explicit support from program maintainers.
Truecolor escape codes are longer and slower to parse.
Fewer terminals support truecolor.DiabloD3 5 hours ago
> Each truecolor program needs its own theme configuration.
Yes, but programs that need more than 16 colors are already often doing this.
> Changing your color scheme means editing multiple config files.
Already has been implemented at least several times over.
> Light/dark switching requires explicit support from program maintainers.
Yes, we have an escape sequence to ask the terminal to ask the OS for this. Most programs that need to care about this already use it.
> Truecolor escape codes are longer and slower to parse.
Not in a way that matters. It only shows up on torture test benchmarks on especially slow parsers. It wouldn't effect normal usage even on a terminal running on some tiny SBC.
> Fewer terminals support truecolor.
XTerm itself, rxvt, iterm2, Alacritty, Wez, Kitty, Ghostty, wt/modern conhost, mintty, st, everything that uses libvte, everything that uses libvterm (such as neovim's built in term), emacs's term, asciinema; tmux, screen, dvtm safely handle them; the linux console snaps it to the nearest 256 color; iterm.app, the original garbage osx one, is the only major term I can think of that does not handle it.
linhns 6 hours ago
Nice to see Ghostty implemented it already. Feature-packed with sane defaults.
leni536 14 hours ago
What I would like is HDR colors, just to access more saturated light colors. I don't want less saturated blue to make it lighter, just crank up the blue channel to 11. I still don't want brighter colors than #fff though.
hackrmn 13 hours ago
I think your wish is self-contradicting. `#fff` is so-called _device_ colour -- a device like a LED-based display uses it directly to drive the LEDs, where `#fff` means that the red, the green and the blue channel are already "cranked to 11". The `f` here is equivalent to 11. HDR uses a different color format, I think -- exactly because `#fff` is otherwise either ambigous, or has to map to a different colour gamut -- where, for instance, `#fff` actually means the whitest white cranked up to 11, at however many nits (say 1500) the monitor may emit, which would make your "standard" or "SDR" white (per sRGB, say) that's usually has the emitted strength of around 100 nits, be somewhere at `#888` (I haven't taken into account the _curve_ implied here, at any rate I don't think it's going to be a linear relationship between nits and device primary N-bit colour numbers).
Also, `#fff` is ambigous -- if you mean device colour, then there's no brightness (nits) specified at all, it may be 200 or 2000 or 10,000. If sRGB is implied, as in `#fff in sRGB colour space` then the standard specifies 80 nits, so when you say you don't want brighter than that, then you can't have much of HDR since sRGB precludes HDR by definition (can't go brighter than 80 nits for the "whitepoint" aka white).
I think if you want HDR you need a different colour space entirely, which either has a different peak brightness, or one where the brightness is specified additionally to e.g. R, G and B primaries. But here my HDR knowledge is weak -- perhaps someone else may chime in. I just find colour science fascinating, sorry to go on a tangent here.
account42 10 hours ago
It's pretty obvious that gp is asking for the brightest blue to be as bright as the brightest white but no brighter.
And no, #fff is not a "device color". The syntax originates from the web where sRGB is implied ever since we had displays brighter than that.
hackrmn 6 hours ago
leni536 9 hours ago
I meant #fff as the nominal white point, which is not the brightest white the device can produce. It's how #fff is displayed on Firefox on an otherwise HDR screen. Assuming the screen is not set to max brightness and otherwise HDR capable, it means that it's not the max brightness an individual pixel/subpixel can produce, so desaturating blue color just to make it brighter can be an unnecessary compromise only dictated by color representation in the software stack.
hackrmn 6 hours ago
foldr 3 hours ago
> If you've spent much time in the terminal, you've probably set a custom base16 theme.
Hmm, I suspect that almost everyone who works in the terminal has never done this. I don’t really care what the colors look like, beyond choosing between whatever built in themes my terminal has. Is this really the minority experience?
kccqzy 37 minutes ago
I'm pretty sure the word "custom" refers to the act of choosing among builtin themes in the terminal. Some terminals like ghostty have hundreds builtin.
layer8 12 hours ago
I’d also add a quantization slider that would quantize the in-between colors to less than 256 different colors; in the extreme position to just the 16 colors.
ku1ik 13 minutes ago
That would be a super-niche setting, but an interesting idea nonetheless.
therealmarv 10 hours ago
seems this is implemented in latest iTerm2 commit https://github.com/gnachman/iTerm2/commit/39bafa8d6651865951...
dwb 15 hours ago
Agree, and I love how concise, yet persuasive and practical this proposal is.
evolve2k 9 hours ago
A key critique
The article is well argued and well written, as far as I’m concerned.
What’s needed if you really want adoption is to define a term; something like 256-ex for extended. Or whatever. But folks and apps need to be able to say; we’ve implemented or we support 256-ex. Without this label is hard to drive adoption.
The heartbleed thing from a few years back best taught me this.
Good luck I hope to see broad adoption it’s a great idea.
Give it a name, better yet move the copy onto a website with the same name also and a little icon folks can add to their website if they implement this k to their terminal app.
King-Aaron 16 hours ago
> Complex and color-heavy programs struggle with such a small palette.
Damn if only there was some other system that could be operating with that in mind
worthless-trash 14 hours ago
I feel like you're saying one operating system does it better, but I fail to think of one.
gzread 14 hours ago
Windows allows you to set any pixel to any 24-bit colour.
OJFord 14 hours ago
anthk 8 hours ago
Harsh truth for Unix:
- A shell is not a terminal.
- Rc it's simpler than sh.
- You can totally put shells under graphical windows as in 9front
- You can do a better Unix than Unix itself while ditching out for good the VT220 interface
- Serial terminals aren't a thing under rio(9)
wang_li 6 hours ago
That's hardly a harsh truth. We use the terminal model because it works good enough and there are decades of software that rely on it. There are certainly other models that could work, but so far they haven't been adopted. I'd love Mathematica style notebooks for my shell work.
anthk 4 hours ago
Again, don't confuse the shell (command line interpreter) with a terminal with escape codes and the like. Check how 9font does it, you get the rc shell and a few more languages with a REPL (lua ports and the like) inside a graphical window, not by emulating a VT220 terminal running the shell inside. You can freely resize the 9front's rio window manager's windows running rc (or anything else, even games, graphical browsers, images, video players...), copy, paste, cut the text, save the history, grep the contents of a window's text itself and tons more.
wang_li 24 minutes ago
delaminator 14 hours ago
Terminals should not exist, emulating a serial teletype emulating a Hollerith machine
anthk 11 hours ago
That's 9front/Plan9. No actual terminals, but windows with shells.
delaminator 7 hours ago
As you can see from the down votes, people love emulating Hollerith machines. They've probably never tried writing a terminal emulator.
xvilka 6 hours ago
stackghost 15 hours ago
It's perennially baffling to me why we're still clinging to VT220/xterm compatible terminals. I even see people claiming they prefer working in the terminal, though it's not clear to me what type of work those people are doing.
Give me a proper graphical application any day, but I recognize that it's historically been a lot more work to produce a GUI in the pre-LLM era.
But golly gee whizz if we're going to keep the command line around, can we move on from 1983?
vanviegen 14 hours ago
GUI apps are good for discoverability. They generally are not optimized for rapid use by power users though. That's of course not an inherent limitation of GUI apps, it's just that dominant paradigms hardly seem to consider power users.
I'm still annoyed and baffled by the fact that Ubuntu had searchable application menus 10 years ago, which were awesome and worked for just about any program, and then dropped them when they abandoned Unity. And neither KDE not Gnome thought to bring them back. In stead, many apps have since dropped application menus entirely, in favour of... some mishmash of icons and ad hoc menu buttons?
Also, even in the post-LLM era, building a decent GUI app is a lot more work than building a decent terminal app.
robinsonb5 13 hours ago
Another factor is the lack of a good cross-platform GUI toolkit (by which I mean one that (a) doesn't feel out-of-place and jarring on any platform and (b) doesn't turn a 50K app into a 1GB download.)
Between that and the level of complexity of modern GUI toolkits - and the size of their dependency trees - distribution of a GUI app is a much bigger headache than distributing a terminal app.
iberator 12 hours ago
skydhash 8 hours ago
tezza 13 hours ago
Terminals are text. Text adds features missing from gui namely:
* Ad Hoc
requirements change and terminal gives ultimate empty workbench flexibility. awesome for tasks you never new you had until that moment.
* Precision
run precisely what you want, when you want it. you are not constrained by gui UX limits.
* Pipeline
cat file.txt | perl/awk/sed/jq | tee output.result
* Equal Status
everything is text so you can combine clipboard, files, netcat output, curl output and then you can transform (above) and save. whatever you like in whatever form you like, named whatever you like.
anthk 11 hours ago
Unix is not about a terminal, that the obsolete medium, kinda like preaching about DOS PC with 386's... and CGA video cards.
9front copes with actual text throwing the terminal as a different tool to run legacy stuff in such as SSH or some small ANSI C tools either with the APE compat layer or NPE binding ANSI C code to Plan9 native functins.
You can resize windows with shells under 9front freely, no more SIGWINCH. No more broken cuts and pastes while running a terminal multiplexer. No control code risks executing potential malware.
wredcoll 6 hours ago
consp 15 hours ago
Why? There is huge compatibility layer build on top of this and changing even these minor things will break stuff in places you do not expect. Want a fancy terminal? Install another one. By default most allow changing to many terminal formats. Break things to move forward is fun when it's not your problem to solve down the line.
eigenspace 12 hours ago
Personally, I find a REPL-based programming interface to be the most flexible and powerful way to interact with code.
My typical interface is that I have an editor open where I write package code, and then I have a julia REPL open beside it where I load the package (or includet the script or testset). Any changes I make with my editor to functions or structs in the package / script / testset are automatically reflected in my running julia session via Revise.jl.
I then execute the functions interactively from the REPL, introspect into data or generated code, debug, run benchmarks etc all from the REPL.
GUIs are great for many things, but they lack the flexibility I need for my day to day work.
hnlmorg 14 hours ago
Graphical applications are great if your workflow mirrors the workflow that the GUI was designed for. However if your workflow differs then you’ll often be fighting the GUI.
Think of it like different types of programming languages. Some are more suited on different types of problems than others.
For me, my workflow is about composing and gluing different processes together rather than using a standard defined tool repeatedly in the same way as that tools UI was originally designed. So the CLI is ideally suited for me in all the areas where a GUI would be high friction.
mr_toad 13 hours ago
> it's not clear to me what type of work those people are doing.
As little as possible. Anything that can be done in a terminal can be scripted and automated, put under version control, and deployed using modern CI/CD and devops tools.
ale42 12 hours ago
> Anything that can be done in a terminal can be scripted and automated
Sure, next time I'll need to quickly edit a configuration file I'll setup a complete CI/CD pipeline to run vim (or nano, or whatever) inside it.
pferde 9 hours ago
amelius 13 hours ago
Terminals should be able to show images, so you could run Jupyter notebooks in them.
Maxious 13 hours ago
Terminals can show images https://sw.kovidgoyal.net/kitty/graphics-protocol/
amelius 13 hours ago
Ok, maybe I should have said "framebuffers" instead of images.
Because that Kitty protocol seems limited in that interaction was not one of their goals.
ux266478 8 hours ago
poly2it 11 hours ago
aragilar 11 hours ago
Sixel and ReGIS have been around for decades.
flomo 14 hours ago
If any end users actually cared about terminal/TUI apps, there would be modern APIs for whatever they want.
Since this is really just a legacy system operator monk / retrocool interface, they spend years debating ancient DEC theology.
gjvc 13 hours ago
It's a shame to see you get downvoted (presumably because you don't cite any evidence for your assertions). As a counter-point, I will give the oft-quoted-by-me 1990s promotional video of the use of IBM CallPath on an AS/400 which should get you all misty-eyed https://youtu.be/5pY6Xxptp9A?t=2058
Enjoy.
pjmlp 13 hours ago
This is a limitation of UNIX terminals, in other platforms not tied to a no longer existing tty interface, this isn't an issue.
Unfortunely, given that we are stuck with UNIX derived OSes, this is indeed a possible issue.
However I would argue, for fancy stuff there is the GUI right there.
comboy 13 hours ago
I don't know any GUI that beats tmux work portability.
Of course everything depends on exact nature of the work, but personally I would gladly leave text based interfaces behind, it's just that there's nothing better.
pjmlp 11 hours ago
I know UNIX since being introduced to it via Xenix in 1992 thereabouts, and never found a use for tmux.
More so, I use the terminal as strictly necessary and nothing more.
wredcoll 6 hours ago
anthk 11 hours ago
9front/plan9 leaves tmux and the like as toys. The moment you can use system tools, 9p and files on everything (even the text of the editor itself) you wont be back to these unusable teletype emulators, be XTerm clones or terminal multiplexors.
anthk 11 hours ago
Unix downvoters should have a look on 9front (the most featureful/supported plan9 fork out there) in order to realize that you can follow the Unix philosophy (even more than OpenBSD itself) tossing the VT100 emulators to the legacy 'vt' emulator if you want somehow use ancient software (or SSH) while setting actual graphical windows with the shell running inside with no terminals at all. No more damn control codes messing up everything. No more linebreaks. You can cut down your text with the mouse from day one. Even under Sam, which is something like a graphical Ed crossed with Notepad and much more powerful expressions than Ex/Vi ones.
cbm-vic-20 6 hours ago
The 9front web site is certainly.. something.
networked 6 hours ago
(I didn't downvote you.) Plan 9 made some really cool design decisions, and I keep looking for places to use 9P. I also believe that Plan 9, as an artifact of its era, bet too heavily on the mouse and pixel-mapped graphics. Text, character-cell terminal emulation, and SSH won for good reasons. "GUI > terminal" is a separate matter from "Plan 9 > Unix".
In some important ways, even text with hacky escape codes is more useful and more robust than a fundamentally graphical interface. Text scales up and down with display size and pixel density. Text works over high-latency links and enables simple latency compensation like in Mosh. Text gives you universal copy-paste. Text is more accessible for humans and for machines.
I use a virtual terminal daily on my phone. If we literally used rio GUIs designed around the three-button mouse and desktop pixel density, it would be a lot less ergonomic. I'd like to see a successor to the VT220 lineage, but it's easier to imagine it built on text.