On Oct 18, 08 14:37:12 +0200, Christian König wrote:
Hopefully i don't have missed something important, if i did just leave me a note.
Thanks a lot for your work.
Now to my conclusions. First of all we really need to change something, the number of replies to this thread shows that people are unhappy with the current situation.
While I agree that we have to change something, not all answers showed that people are unhappy. Just for the record.
Second: What we really need isn't a change to the naming scheme radeonhd uses, but a standard xrandr naming scheme all drivers use, just imaging how usefully an hotkey in mplayer/xine/vlc would be to switch your current output to TV-OUT/HDMI (something i do at the moment with a shell script before mplayer starts), or some option in OpenOffice Impress to switch from PANEL to an external output when the presentation starts....
I don't see drivers agreeing on a general naming scheme. Also, IMHO using the name for meta information is the wrong approach, this has been used innumerable APIs, and it always failed horribly one way or another. That said, the upcoming RandR 1.3 properties (yes, I'm working on it again) might make this issue mood, because that will have standardized names for signal types etc.
Additionally: Currently xrandr distinguish between outputs and connectors (AFAIK), but only have names for the outputs.
Connectors have not been evaluated, when RandR 1.2 was designed.
So for the case of DVI we currently end up with something like: DVI-[ADI]_[0-9]+/(analog|digital)
When we add the options of split TMDS lines we would end up with: DVI-[ADI]_[0-9]+/(analog|single digital [12]|dual digita)
Which is in my opinion just a big mess.
In the split TMDS case you just *need* two outputs, because only this way you could drive them with different signals. What you typically want to.
From the POV of an programmer which has to get get his presentation on the wall i just want to connect the beamer and hit start in OpenOffice, wich is the equivalent off running "xrandr --output DVI --auto --above PANEL" and getting OpenOffice to output full screen on DVI. I don't really care if this connector is a single link, dual link, analog or what ever, it should just work because of auto detection whats connected.
No, you actually don't want to explicitly specify DVI. Remember: most laptops still have VGA... As you have to verify the connection state anyway, it doesn't matter whether there is one or two DVI outputs for analog and digital.
From the POV of the multimedia guy wanting to watch the latest block buster on my TV i want to start the movie, select the language, hit a button and get comfortable, because video (and audio, but that's another question) gets routed to my TV.
In multimedia this typically gets complicated, because the setups are provably ambiguous. Some setups are clear, but not all.
From the POV of the hardware freak wanting to get multiple outputs over the same connector to work i need to be able to specify exactly which part of the connector should be driven by which output.
All your points show that this is really an application issue. So far the application had issues finding out what the internal and external display or what the TV is, but that point should be moot with the standard properties. However, I'm not claiming that the application logic will be easy. It will probably be a big heuristic mess.
Am Donnerstag, den 16.10.2008, 12:20 +0200 schrieb Egbert Eich:
Sounds like an interesting device :) More comprehensive description would be "annoying". The really problem with (E)EDID and DVI/HDMI is that the standard doesn't say a word about what should happen with (E)EDID information in the case of a (de)multiplexer, splitter or pass through device.
I would assume because this is not a valid use case at all. Still, if you can support it we probably should.
Sometimes i even ended up with wiring an EEPROM to the DDC bus, because we didn't got an M$ windows driver to do what a want them to do. Making modelines fully configurable was a really god idea when designing X.
:-)
Matthias
--
Matthias Hopf