Gtk-redshift is great - a very interesting application. One problem I noticed was that if internet isn't connected when starting, it doesn't work and doesn't seem to give a warning. I'm wondering why it uses geolocation instead of taking a picture from webcam or light sensor, to check if you're in bright sunlight, in bright indoor lights, or in a dark room.
One thing to take into consideration is that not all users have a camera or light sensor.
Concerning Unternet connection, you can set your location manual, for example -l 50:-130
for 50掳 north, 130掳 west. Or with ~/.config/redshift.conf:
[redshift]
location-provider=manual
[manual]
lat=50
lon=-130
I haven't done too much research in this area but think a light sensor could be useful. My questions would be, how many computers come with light sensors these days? Is there a reliable API to access these sensors?
On the other hand, I think the webcam data would be hard to work with. AFAIK there is a lot of preprocessing going on in the device/driver so the application never gets to see the raw captured values. Maybe someone can come up with an algorithm that can reasonably predict the lightness in the environment. My initial concern is that it may be quite CPU/power intensive.
I have started working on analysing input from webcam and output from X and framebuffer. I have come too far with webcam support yet, but I think it can be done without being too intensive.
Concerning light sensor, one could buy cheap light sensor and connect it to a Raspberry Pi or an Arduino, or other similar platform.
@maandree Great. I'm more concerned about a unified API to the light sensor, otherwise we might have to support dozens of different devices. I don't have a linux computer with a light sensor available at this time so I'm afraid I can't really help much with that part of the research.
What I was thinking about on light sensor is that you would by a something like https://www.sparkfun.com/products/8688 and you specify to which GPIO-ports
you have connected it. Then everything would work without a problem since
the program would read analog voltage readings from a GPIO-port. Nothing
else on the computer is aware of the sensor, so there are no API issues.
Concerning light sensor, one could buy cheap light sensor and connect it to a Raspberry Pi or an Arduino, or other similar platform.
Actually, redshift doesn't work on Raspberry Pi currently.
https://github.com/jonls/redshift/issues/257
Any news on this one?
IMO, redshift shouldn't even attempt at accessing a webcam or something like this, but rather just read some sensor API and another piece of software should handle requests under that API.
This way (unix way, by the way) you won't have to add too too much stuff to redshift and it would (should) be easy for people with custom sensors to write a small program to just handle some API requests.
analyze manual brightness changes history?
@Goury if a library like this was available we could consider using it but I haven't heard of or seen anything like this. We can reconsider this issue if something like this becomes available.
@jonls
TF are you closing this issue?
Clightd is out there for almost a decade.
99% of people've never heard of GNU GPL v3, does this mean Redshift doesn't exist?
Most helpful comment
I haven't done too much research in this area but think a light sensor could be useful. My questions would be, how many computers come with light sensors these days? Is there a reliable API to access these sensors?
On the other hand, I think the webcam data would be hard to work with. AFAIK there is a lot of preprocessing going on in the device/driver so the application never gets to see the raw captured values. Maybe someone can come up with an algorithm that can reasonably predict the lightness in the environment. My initial concern is that it may be quite CPU/power intensive.