I haven't really gone through X11 before, but the API really is terrible. How have we managed to hang onto this monstrosity for so long? I get people have a fondness for something that has held up for so long, but the huge complexity and difficulty of hardware acceleration means it's time to throw it out.
Head over there and drag the kitten around to see why. Now your average Linux desktop builds a lot on X11 to get it to perform how it does now, but all this effort is a huge waste. Just compare it to something like Apple's UIKit or Android's layouts - it's well past the sell by date.
X has had a feature, called backing store, to get around that since the 80s. Individual windows can request backing store, and there's an option you can set in xorg.conf to enable backing store for all windows.
X11 actually anticipated a future of composited windows, each with their own separate buffer to draw into, well before its rivals -- due in no small part to being intended to run on much more advanced hardware than Windows or Mac OS.
The issue is the added complexity of having to enable backing stores and the multiple code paths. It's not about "how much we can glue inside X11 until it sort of works sometimes".
You can throw out 95% of X11 and most developers would be much happier, bugs would go down, and everything would still work fine.
Without that, even in a slow JavaScript implementation, you can barely see the white background.
As I explain in the article, this is entirely because of memory pressure limits of the 80s. This is what Windows 7 does too. And in a modern environment (compositing manager), windows do get their own backing pixel memory.
Then use a different one, like xcb. Or even abandon it altogether and use Wayland.
> How have we managed to hang onto this monstrosity for so long?
Because it (largely) works.
And 99% of people don't ever have to come into contact with it.
> Now your average Linux desktop builds a lot on X11 to get it to perform how it does now
If you mean "builds a lot on" you mean "ignores the vast majority of outdated features of", then yes.
> but all this effort is a huge waste
Not really. A lot of the stuff replacing X11 cruft had to be implemented anyway, and the weird X11 hacks, well, they were just what was needed at the time to get work done. No use in crying over spilt milk.
> Just compare it to something like Apple's UIKit or Android's layouts
What would be the point in that? They serve very different purposes.
>> but the API really is terrible
>
> Then use a different one, like xcb.
The problem is that it is difficult to use xcb with GLX. I just say glXChooseFBConfig. And I consider using Xlib for these API functions - as suggested on http://xcb.freedesktop.org/opengl/ as an even larger mess.
http://magcius.github.io/xplain/article/x-basics.html
Head over there and drag the kitten around to see why. Now your average Linux desktop builds a lot on X11 to get it to perform how it does now, but all this effort is a huge waste. Just compare it to something like Apple's UIKit or Android's layouts - it's well past the sell by date.