Skip to contentSkip to conversation

tty[1-6]: colors a negative accessibility/usability trend

DISCUSSION Felix Miata 10 participants 25 messages Feb 12, 2015 – Mar 2, 2015 Util-Linux package

lore
Share & tools
lore
#0 lore raw

Red, blue and green in particular produce poor contrast on the black background of a typical framebuffer screen. As these screens are configured as functional twins of those modes we find outselves in at rescue time, it amounts to an unfortunate and unnecessary accessibility/usability obstacle that is IMO is a subset of a larger general problem found under the moniker A11Y[1].

A very good, very recent article discusses the mindset that leads to default settings that thwart use by the disadvantaged[2], however slight that impediment may be.

With util-linux >= 2.25 we can turn colors off via

	# touch /etc/terminal-colors.d/disable [3]

for some commands, but far too few. It should apply to all, including many outside the purview of util-linux, and more importantly, do so by default.

It should be up to those who wish a legibility reduction to discover how to and apply the reduction, not the other way around as it is now. It is much more difficult for those who cannot handle low contrast to improve it than for those who find it too high to reduce it, a variation on the chicken/egg paradigm.

If we can get all components of util-linux to adhere to maximizing legibility, it would have the potential serve as an important springboard for other projects to do the same, and maybe even spill onto the web.

[1] http://a11yproject.com/
[2] http://alistapart.com/article/reframing-accessibility-for-the-web
[3] http://karelzak.blogspot.de/2014/04/terminal-colorsd.html

Signature · 6 lines
"The wise are known for their understanding, and pleasant
words are persuasive." Proverbs 16:21 (New Living Translation)

 Team OS/2 ** Reg. Linux User #211409 ** a11y rocks!

Felix Miata  ***  http://fm.no-ip.com/
#1 lore raw
Quoted 2 lines · Felix Miata

Red, blue and green in particular produce poor contrast on the black background of a typical framebuffer screen.

Agreed, though the bold variants are better.
Better again are 256 color variants.
The Linux console lagged xterms in support, but does support it now.

Quoted 9 lines

As these screens are configured
as functional twins of those modes we find ourselves in at rescue time, it amounts to an unfortunate and unnecessary accessibility/usability obstacle that is IMO is a subset of a larger general problem found under the moniker A11Y[1].

A very good, very recent article discusses the mindset that leads to default settings that thwart use by the disadvantaged[2], however slight that impediment may be.

I agree completely. color is useful but one has to be very careful in how it's used. ls --color for example is too aggressive IMHO, and I adjust to highlight rather than color with:
http://www.pixelbeat.org/scripts/l
I've been pusing back upstream against adding new color combinations.

Quoted 6 lines

With util-linux >= 2.25 we can turn colors off via

	# touch /etc/terminal-colors.d/disable [3]

for some commands, but far too few. It should apply to all, including many outside the purview of util-linux, and more importantly, do so by default.

coreutils will keep this scheme in mind too.

thanks,
Pádraig.

#2 lore raw
Quoted 5 lines · <[email protected]> writes:

I agree completely. color is useful but one has to be very careful in how it's used. ls --color for example is too aggressive IMHO, and I adjust to highlight rather than color with:
http://www.pixelbeat.org/scripts/l
I've been pusing back upstream against adding new color combinations.

I don't particularly like colorization. But technically, it seems to me that what is needed is a systematic way for the user to indicate his colorization preferences to *all* utilities. And a corresponding way for the system to provide defaults for those user preferences.

Only when there is a systematic framework for colorization will all the programs allow colorizing to be configured.

Dale

maintainer
#3 lore raw
Quoted 8 lines · Dale R. Worley

I don't particularly like colorization. But technically, it seems to me that what is needed is a systematic way for the user to indicate his colorization preferences to *all* utilities. And a corresponding way for the system to provide defaults for those user preferences.

Only when there is a systematic framework for colorization will all the programs allow colorizing to be configured.

It would be possible to create a shared library from our lib/colors.c to support terminal-colors.d/... :-)

Karel

Signature · 2 lines
 Karel Zak  <[email protected]>
 http://karelzak.blogspot.com
#4 lore raw
Quoted 7 lines · Karel Zak
Quoted 6 lines · Dale R. Worley

I don't particularly like colorization. But technically, it seems to me that what is needed is a systematic way for the user to indicate his colorization preferences to *all* utilities. And a corresponding way for the system to provide defaults for those user preferences.

Quoted 5 lines
Quoted 2 lines

Only when there is a systematic framework for colorization will all the programs allow colorizing to be configured.

It would be possible to create a shared library from our lib/colors.c to support terminal-colors.d/... :-)

Or add the needed to ncurses. Isn't that better than adding a new library?

Signature · 2 lines
Sami Kerola
http://www.iki.fi/kerolasa/
maintainer
#5 lore raw
Quoted 8 lines · Sami Kerola
Quoted 7 lines · Karel Zak
Quoted 6 lines · Dale R. Worley

I don't particularly like colorization. But technically, it seems to me that what is needed is a systematic way for the user to indicate his colorization preferences to *all* utilities. And a corresponding way for the system to provide defaults for those user preferences.

Quoted 8 lines
Quoted 5 lines
Quoted 2 lines

Only when there is a systematic framework for colorization will all the programs allow colorizing to be configured.

It would be possible to create a shared library from our lib/colors.c to support terminal-colors.d/... :-)

Or add the needed to ncurses. Isn't that better than adding a new library?

you want to link programs like dmesg, gcc or ls with ncurses monster?

The another story is that ncurses provides completely abstract layer for colors and I didn't found a way how to use it together with color escape sequences. This is reason why for example cfdisk supports only enable/disable terminal-colors.d feature, but no schemes to specify colors.

Karel

Signature · 2 lines
 Karel Zak  <[email protected]>
 http://karelzak.blogspot.com
#6 lore raw
Quoted 18 lines · Karel Zak
Quoted 8 lines · Sami Kerola
Quoted 7 lines · Karel Zak
Quoted 6 lines · Dale R. Worley

I don't particularly like colorization. But technically, it seems to me that what is needed is a systematic way for the user to indicate his colorization preferences to *all* utilities. And a corresponding way for the system to provide defaults for those user preferences.

Quoted 8 lines
Quoted 5 lines
Quoted 2 lines

Only when there is a systematic framework for colorization will all the programs allow colorizing to be configured.

It would be possible to create a shared library from our lib/colors.c to support terminal-colors.d/... :-)

Or add the needed to ncurses. Isn't that better than adding a new library?

Quoted 7 lines

you want to link programs like dmesg, gcc or ls with ncurses monster?

The another story is that ncurses provides completely abstract layer for colors and I didn't found a way how to use it together with color escape sequences. This is reason why for example cfdisk supports only enable/disable terminal-colors.d feature, but no schemes to specify colors.

this is why ncurses has the ability to split its lib into a smaller libtinfo. you get all the logic for detecting terminal capabilities without all the rest of the ncurses layers. for people who do want to kill colors across the board, they can use a different TERM that does not support them. forcing people to duplicate that in a new database seems wrong to me.
-mike

maintainer
#7 lore raw
Quoted 27 lines · Mike Frysinger
Quoted 18 lines · Karel Zak
Quoted 8 lines · Sami Kerola
Quoted 7 lines · Karel Zak
Quoted 6 lines · Dale R. Worley

I don't particularly like colorization. But technically, it seems to me that what is needed is a systematic way for the user to indicate his colorization preferences to *all* utilities. And a corresponding way for the system to provide defaults for those user preferences.

Quoted 8 lines
Quoted 5 lines
Quoted 2 lines

Only when there is a systematic framework for colorization will all the programs allow colorizing to be configured.

It would be possible to create a shared library from our lib/colors.c to support terminal-colors.d/... :-)

Or add the needed to ncurses. Isn't that better than adding a new library?

Quoted 7 lines

you want to link programs like dmesg, gcc or ls with ncurses monster?

The another story is that ncurses provides completely abstract layer for colors and I didn't found a way how to use it together with color escape sequences. This is reason why for example cfdisk supports only enable/disable terminal-colors.d feature, but no schemes to specify colors.

Quoted 5 lines

this is why ncurses has the ability to split its lib into a smaller libtinfo. you get all the logic for detecting terminal capabilities without all the rest of the ncurses layers. for people who do want to kill colors across the board, they can use a different TERM that does not support them. forcing people to duplicate that in a new database seems wrong to me.

What we want to duplicate? What exactly in lib/colors.c is duplicate? The code evaluates filenames and parses files with "name colorcode".

Anyway, the core of the terminal-colors.d/ is enable/disable feature, for this purpose is unnecessary any library.

Karel

Quoted 1 line

-mike

Signature · 2 lines
 Karel Zak  <[email protected]>
 http://karelzak.blogspot.com
#8 lore raw
Quoted 28 lines · Karel Zak
Quoted 27 lines · Mike Frysinger
Quoted 25 lines · Karel Zak
Quoted 8 lines · Sami Kerola
Quoted 7 lines · Karel Zak
Quoted 6 lines · Dale R. Worley

I don't particularly like colorization. But technically, it seems to me that what is needed is a systematic way for the user to indicate his colorization preferences to *all* utilities. And a corresponding way for the system to provide defaults for those user preferences.

Quoted 8 lines
Quoted 5 lines
Quoted 2 lines

Only when there is a systematic framework for colorization will all the programs allow colorizing to be configured.

It would be possible to create a shared library from our lib/colors.c to support terminal-colors.d/... :-)

Or add the needed to ncurses. Isn't that better than adding a new library?

you want to link programs like dmesg, gcc or ls with ncurses monster?

The another story is that ncurses provides completely abstract layer for colors and I didn't found a way how to use it together with color escape sequences. This is reason why for example cfdisk supports only enable/disable terminal-colors.d feature, but no schemes to specify colors.

Quoted 8 lines
Quoted 5 lines

this is why ncurses has the ability to split its lib into a smaller libtinfo. you get all the logic for detecting terminal capabilities without all the rest of the ncurses layers. for people who do want to kill colors across the board, they can use a different TERM that does not support them. forcing people to duplicate that in a new database seems wrong to me.

What we want to duplicate? What exactly in lib/colors.c is duplicate? The code evaluates filenames and parses files with "name colorcode".

whether the terminal even supports color in the first place. if i picked a terminal that doesn't support color so i didn't have to worry about it, it's a bit crazy i also have to go to multiple config files and also tell them i don't want color otherwise i get corrupted output.
-mike

maintainer
#9 lore raw
Quoted 8 lines · Mike Frysinger
Quoted 2 lines · Karel Zak

What we want to duplicate? What exactly in lib/colors.c is duplicate? The code evaluates filenames and parses files with "name colorcode".

whether the terminal even supports color in the first place. if i picked a terminal that doesn't support color so i didn't have to worry about it, it's a bit crazy i also have to go to multiple config files and also tell them i don't want color otherwise i get corrupted output.

For now it checks isatty() and nothing else, it would be possible to check number of supported colors in terminfo too, or is there any other way?

Anyway, we don't want to create (duplicate) any database, all it provides are knobs for customization and enable/disable.

Karel

Signature · 2 lines
 Karel Zak  <[email protected]>
 http://karelzak.blogspot.com
#10 lore raw
Quoted 5 lines · Karel Zak, le Mon 16 Feb 2015 10:47:47 +0100, a écrit :
Quoted 4 lines · Mike Frysinger
Quoted 2 lines · Karel Zak

What we want to duplicate? What exactly in lib/colors.c is duplicate? The code evaluates filenames and parses files with "name colorcode".

Quoted 8 lines
Quoted 4 lines

whether the terminal even supports color in the first place. if i picked a terminal that doesn't support color so i didn't have to worry about it, it's a bit crazy i also have to go to multiple config files and also tell them i don't want color otherwise i get corrupted output.

For now it checks isatty() and nothing else, it would be possible to check number of supported colors in terminfo too, or is there any other way?

It should also check how to change the colors. Not all terminals will support the ANSI way.

Quoted 2 lines

Anyway, we don't want to create (duplicate) any database, all it provides are knobs for customization and enable/disable.

It provides only that because the source code currently hardcodes the \033[%sm sequence, but it really should not, and use terminfo instead.

Samuel

maintainer
#11 lore raw
Quoted 8 lines · Samuel Thibault

It should also check how to change the colors. Not all terminals will support the ANSI way.

Quoted 2 lines

Anyway, we don't want to create (duplicate) any database, all it provides are knobs for customization and enable/disable.

It provides only that because the source code currently hardcodes the \033[%sm sequence, but it really should not, and use terminfo instead.

Good point, I'm ready to accept patches to implement this feature :-)

Karel

Signature · 2 lines
 Karel Zak  <[email protected]>
 http://karelzak.blogspot.com
maintainer
#12 lore raw
In reply to Samuel Thibault
Quoted 11 lines · Samuel Thibault
Quoted 10 lines · Karel Zak, le Mon 16 Feb 2015 10:47:47 +0100, a écrit :
Quoted 8 lines · Mike Frysinger
Quoted 2 lines · Karel Zak

What we want to duplicate? What exactly in lib/colors.c is duplicate? The code evaluates filenames and parses files with "name colorcode".

whether the terminal even supports color in the first place. if i picked a terminal that doesn't support color so i didn't have to worry about it, it's a bit crazy i also have to go to multiple config files and also tell them i don't want color otherwise i get corrupted output.

Quoted 6 lines
Quoted 3 lines

For now it checks isatty() and nothing else, it would be possible to check number of supported colors in terminfo too, or is there any other way?

It should also check how to change the colors. Not all terminals will support the ANSI way.

OK, linked with libtinfo, now it checks if the current terminal supports colors ("colors" from tinfo) and all the stuff is disabled for terminals like vt100.

Quoted 5 lines
Quoted 2 lines

Anyway, we don't want to create (duplicate) any database, all it provides are knobs for customization and enable/disable.

It provides only that because the source code currently hardcodes the \033[%sm sequence, but it really should not, and use terminfo instead.

This is not implemented yet, I (or any volunteer?) will try to add to code an extra layer to avoid the hardcoded sequences and read the colors from terminfo.

Karel

Signature · 2 lines
 Karel Zak  <[email protected]>
 http://karelzak.blogspot.com
#13 lore raw
Quoted 23 lines · Karel Zak
Quoted 11 lines · Samuel Thibault
Quoted 10 lines · Karel Zak, le Mon 16 Feb 2015 10:47:47 +0100, a écrit :
Quoted 8 lines · Mike Frysinger
Quoted 2 lines · Karel Zak

What we want to duplicate? What exactly in lib/colors.c is duplicate? The code evaluates filenames and parses files with "name colorcode".

whether the terminal even supports color in the first place. if i picked a terminal that doesn't support color so i didn't have to worry about it, it's a bit crazy i also have to go to multiple config files and also tell them i don't want color otherwise i get corrupted output.

Quoted 6 lines
Quoted 3 lines

For now it checks isatty() and nothing else, it would be possible to check number of supported colors in terminfo too, or is there any other way?

It should also check how to change the colors. Not all terminals will support the ANSI way.

OK, linked with libtinfo, now it checks if the current terminal supports colors ("colors" from tinfo) and all the stuff is disabled for terminals like vt100.

Quoted 8 lines
Quoted 5 lines
Quoted 2 lines

Anyway, we don't want to create (duplicate) any database, all it provides are knobs for customization and enable/disable.

It provides only that because the source code currently hardcodes the \033[%sm sequence, but it really should not, and use terminfo instead.

This is not implemented yet, I (or any volunteer?) will try to add to code an extra layer to avoid the hardcoded sequences and read the colors from terminfo.

I'd be careful to not over engineer that. coreutils currently hardcodes ansi sequences, and I've not heard specific complaints about it. What terminals are not catered for here?

#14 lore raw
Quoted 8 lines · Pádraig Brady
Quoted 2 lines · Karel Zak

This is not implemented yet, I (or any volunteer?) will try to add to code an extra layer to avoid the hardcoded sequences and read the colors from terminfo.

I'd be careful to not over engineer that. coreutils currently hardcodes ansi sequences, and I've not heard specific complaints about it. What terminals are not catered for here?

At this point, it's hard to imagine anyone writing a new terminal
emulator with incompatible color escape sequences.

Anyone on really rare hardware terminals or crufty software that is
different for some reason can always disable colors for software that hard-codes sequences. They're nice but non-essential.

The only scenario I'm coming up with where it would be really good to
use a library to find what escape sequences to use would be future-proofing against the day when somebody extends the ANSI color escape sequences to support more colors, or 3D for stereoscopic-vision terminals. And even then, only if for some reason they choose to extend the escape sequences in a non-backwards-compat way.

I'd hold off on spending time on this until some new idea
for color escape codes comes along. I think there's enough inertia behind the current ANSI escape sequences that terminal emulators will need to support current-style escape sequences for decades after the appearance of a new way of doing things. I think decades is long enough for util-linux to make this proposed change and have it percolate into long-term-support releases before any terminal emulators think they can get away with dropping support for current-style color escape sequences.

Anyway, that's my 2 cents. I'd maybe change my mind if anyone has a
real-life case where the hardcoded ANSI codes don't work for them, on a terminal they actually want to use, but terminfo would. (I'm not talking about ancient cruft you can find that's incompat, but rather stuff people actually do use regularly.)

Signature · 6 lines
#define X(x,y) x##y
Peter Cordes ;  e-mail: X(peter@cor , des.ca)

"The gods confound the man who first found out how to distinguish the hours!
 Confound him, too, who in this place set up a sundial, to cut and hack
 my day so wretchedly into small pieces!" -- Plautus, 200 BC
maintainer
#15 lore raw
In reply to Pádraig Brady
Quoted 27 lines · Pádraig Brady
Quoted 26 lines · Karel Zak
Quoted 11 lines · Samuel Thibault
Quoted 10 lines · Karel Zak, le Mon 16 Feb 2015 10:47:47 +0100, a écrit :
Quoted 8 lines · Mike Frysinger
Quoted 2 lines · Karel Zak

What we want to duplicate? What exactly in lib/colors.c is duplicate? The code evaluates filenames and parses files with "name colorcode".

whether the terminal even supports color in the first place. if i picked a terminal that doesn't support color so i didn't have to worry about it, it's a bit crazy i also have to go to multiple config files and also tell them i don't want color otherwise i get corrupted output.

Quoted 6 lines
Quoted 3 lines

For now it checks isatty() and nothing else, it would be possible to check number of supported colors in terminfo too, or is there any other way?

It should also check how to change the colors. Not all terminals will support the ANSI way.

OK, linked with libtinfo, now it checks if the current terminal supports colors ("colors" from tinfo) and all the stuff is disabled for terminals like vt100.

Quoted 3 lines
Quoted 2 lines

Anyway, we don't want to create (duplicate) any database, all it provides are knobs for customization and enable/disable.

Quoted 8 lines
Quoted 5 lines
Quoted 2 lines

It provides only that because the source code currently hardcodes the \033[%sm sequence, but it really should not, and use terminfo instead.

This is not implemented yet, I (or any volunteer?) will try to add to code an extra layer to avoid the hardcoded sequences and read the colors from terminfo.

I'd be careful to not over engineer that.
coreutils currently hardcodes ansi sequences,

Yes, I know about it and coreutils has been my specimen for terminal-colors :-)

Quoted 2 lines

and I've not heard specific complaints about it.
What terminals are not catered for here?

I don't know, my plan is to do some research to better understand what Samuel is talking about.

Karel

Signature · 2 lines
 Karel Zak  <[email protected]>
 http://karelzak.blogspot.com
#16 lore raw
Quoted 6 lines · Karel Zak, le Mon 02 Mar 2015 09:59:29 +0100, a écrit :
Quoted 2 lines · Pádraig Brady

and I've not heard specific complaints about it.
What terminals are not catered for here?

I don't know, my plan is to do some research to better understand what Samuel is talking about.

Mmm, I thought it was well-known, but apparently not :/

I mean using this (see terminfo(5) for the details):

#include <term.h>
#include <stdio.h>
#include <unistd.h>

int main(void) {
int my_bg_color = 0;
int my_fg_color = 4;

setupterm(NULL, STDOUT_FILENO, NULL); tputs(enter_bold_mode, 1, putchar); tputs(enter_italics_mode, 1, putchar); tputs(enter_underline_mode, 1, putchar); tputs(enter_blink_mode, 1, putchar); tputs(enter_reverse_mode, 1, putchar); tputs(tparm(set_a_foreground, my_fg_color), 1, putchar); tputs(tparm(set_a_background, my_bg_color), 1, putchar); printf("Hello, world!\n"); tputs(exit_attribute_mode, 1, putchar);
return 0;
}

On terminals which support it, you can even get rgb colors by using initialize_color or initialize_pair. If you want a more complete example, see lstopo-text.c inside the hwloc package, we include ncurses, but we don't actually need it for this kind of code. Try with

lstopo -.txt

Doing it really properly as exampled above doesn't look terribly overengineered to me.

Samuel

maintainer
#17 lore raw
Quoted 33 lines · Samuel Thibault
Quoted 6 lines · Karel Zak, le Mon 02 Mar 2015 09:59:29 +0100, a écrit :
Quoted 2 lines · Pádraig Brady

and I've not heard specific complaints about it.
What terminals are not catered for here?

I don't know, my plan is to do some research to better understand what Samuel is talking about.

Mmm, I thought it was well-known, but apparently not :/

I mean using this (see terminfo(5) for the details):

#include <term.h>
#include <stdio.h>
#include <unistd.h>

int main(void) {
int my_bg_color = 0;
int my_fg_color = 4;

setupterm(NULL, STDOUT_FILENO, NULL); tputs(enter_bold_mode, 1, putchar); tputs(enter_italics_mode, 1, putchar); tputs(enter_underline_mode, 1, putchar); tputs(enter_blink_mode, 1, putchar); tputs(enter_reverse_mode, 1, putchar); tputs(tparm(set_a_foreground, my_fg_color), 1, putchar); tputs(tparm(set_a_background, my_bg_color), 1, putchar); printf("Hello, world!\n"); tputs(exit_attribute_mode, 1, putchar);
return 0;
}

Quoted 4 lines

On terminals which support it, you can even get rgb colors by using initialize_color or initialize_pair. If you want a more complete example, see lstopo-text.c inside the hwloc package, we include ncurses, but we don't actually need it for this kind of code.

This is misunderstanding, I know about tputs :)

The question is how often the harcoded sequences are problem, which terminals? how often people use it? Would be enough to detect these terminals and use things without colors there?

Karel

Signature · 2 lines
 Karel Zak  <[email protected]>
 http://karelzak.blogspot.com
#18 lore raw
In reply to Sami Kerola
Quoted 2 lines · Sami Kerola

Or add the needed to ncurses. Isn't that better than adding a new library?

That would work, but only for utilities that use ncurses. OTOH, it would be good if ncurses understood and provided information from the general color configuration to programs that use it.

Quoted 2 lines · Bruce Dubbs

Why not use the LS_COLORS environment variable used by coreutils' dircolors? Does util-linux need to create something different?

Using an environment variable is the standard Unix way of configuring the user-oriented behavior of programs. But the "definition" of LS_COLORS seems to be very specific to "ls".

The /etc/terminal-colors.d system seems a bit oversized (all those files!), but it seems unlikely that anything simpler would work for a large set of utilities, each of which the user may want to customize.

The main deficiency of terminal-colors.d is that it's system-wide. But I see this in a blog:

$HOME/.config/terminal-colors.d overrides the global setting

so that is taken care of.

You also want to define an environment variable to override *that*, e.g., files in $UTIL_LINUX_TERMINAL_COLORS override the ones in $HOME/.config/terminal-colors.d -- that allows you to set up custom environments for particular trees of processes by creating a suitable colors directory and setting UTIL_LINUX_TERMINAL_COLORS.

Dale

maintainer
#19 lore raw
Quoted 3 lines · Dale R. Worley
Quoted 2 lines · Bruce Dubbs

Why not use the LS_COLORS environment variable used by coreutils' dircolors? Does util-linux need to create something different?

because it's absolutely inelegant solution

echo $LS_COLORS | wc --bytes
1710

Quoted 4 lines

The main deficiency of terminal-colors.d is that it's system-wide. But I see this in a blog:

$HOME/.config/terminal-colors.d overrides the global setting

it's already implemented.

http://man7.org/linux/man-pages/man5/terminal-colors.d.5.html

Quoted 5 lines

You also want to define an environment variable to override *that*, e.g., files in $UTIL_LINUX_TERMINAL_COLORS override the ones in $HOME/.config/terminal-colors.d -- that allows you to set up custom environments for particular trees of processes by creating a suitable colors directory and setting UTIL_LINUX_TERMINAL_COLORS.

for now it follows

$XDG_CONFIG_HOME/terminal-colors.d
$HOME/.config/terminal-colors.d
/etc/terminal-colors.d

Karel

Signature · 2 lines
 Karel Zak  <[email protected]>
 http://karelzak.blogspot.com
#20 lore raw
In reply to Karel Zak
Quoted 7 lines · Karel Zak
Quoted 6 lines · Dale R. Worley

I don't particularly like colorization. But technically, it seems to me that what is needed is a systematic way for the user to indicate his colorization preferences to *all* utilities. And a corresponding way for the system to provide defaults for those user preferences.

Quoted 5 lines
Quoted 2 lines

Only when there is a systematic framework for colorization will all the programs allow colorizing to be configured.

It would be possible to create a shared library from our lib/colors.c to support terminal-colors.d/... :-)

Why not use the LS_COLORS environment variable used by coreutils' dircolors? Does util-linux need to create something different?

-- Bruce

maintainer
#21 lore raw
In reply to Felix Miata
Quoted 6 lines · Felix Miata

With util-linux >= 2.25 we can turn colors off via

	# touch /etc/terminal-colors.d/disable [3]

for some commands, but far too few. It should apply to all, including many outside the purview of util-linux, and more importantly, do so by default.

This is never ending story and there is no setting which works for everyone. All I want to provide tools to make it possible to support various use-cases. It's downstream (distors) job to provide knobs to switch between the use-cases (profiles).

For example I can imagine packages
fedora-terminal-colors-{high-contrast,disable,...}.rpm
to setup coreutils, util-linux, gcc, ... whatever.

Quoted 5 lines

It should be up to those who wish a legibility reduction to discover how to and apply the reduction, not the other way around as it is now. It is much more difficult for those who cannot handle low contrast to improve it than for those who find it too high to reduce it, a variation on the chicken/egg paradigm.

We can also add --{disable,enable}-colors-by-default to move the decision to downstream and use default according the current terminal-colors.d setting. Your default will be disable, my enabled.

For example for fedora I will keep the colors enabled, because it's current distribution policy.

Karel

Signature · 2 lines
 Karel Zak  <[email protected]>
 http://karelzak.blogspot.com
#22 lore raw
Quoted 2 lines · Karel Zak, le Thu 12 Feb 2015 14:21:10 +0100, a écrit :

For example for fedora I will keep the colors enabled, because it's current distribution policy.

But all distributions need to be accessible. It is a per-user preference, not a per-task or per-distribution preference. So upstream at least needs to provide ways to configure it per user, and downstream provide a uniform way to configure the various shipped tools.

Samuel

maintainer
#23 lore raw
Quoted 5 lines · Samuel Thibault
Quoted 2 lines · Karel Zak, le Thu 12 Feb 2015 14:21:10 +0100, a écrit :

For example for fedora I will keep the colors enabled, because it's current distribution policy.

But all distributions need to be accessible.

I'll be happy to provide support for both ways (colors and non-colors) as downstream maintainer, but for example Fedora does not care... (for now).

Quoted 4 lines

It is a per-user
preference, not a per-task or per-distribution preference. So upstream at least needs to provide ways to configure it per user, and downstream provide a uniform way to configure the various shipped tools.

Yeah, util-linux upstream already provides per-user (or per-machine) configuration for colors (man terminal-colors.d). It would be nice to support this in another packages too. Then we can improve things on downstream level.

We have tons of crazy themes for desktops UI, mp3 playes, web browsers, etc. It's time to support something like this for terminals too ;-)

Karel

Signature · 2 lines
 Karel Zak  <[email protected]>
 http://karelzak.blogspot.com
#24 lore raw
In reply to Samuel Thibault

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!
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