I think this post is one of those "technically correct" but functionally wrong rants.
My experience with multiple monitors at different dpis hits nearly every case of failure he points out.
It's a lot more work to configure.
Apps fail to account for it.
Spanning two monitors results in terrible scaling problems.
Apps that do account for it at start up won't account for it during reposition, so they look fine if they open on the right monitor and terrible if moved to the other.
Getting solid workflows for flexible positioning requires hacks like mentioned at the end for xrandr.
Etc...
So sure, you can do it and it sucks.
That's not really the win I think the author seems to think it is.
"Technically correct" about fundamental limitations of reality but you still hate the truth and it's X11's fault?
I'm not even trying to defend X11, I don't have much love for it. I've done some Xlib programming in the past and I've hated it. I've never used Wayland, and I'm mostly on Windows these days.
But, I don't see how one could make a point that X11 is bad because of poor DPI support.
Some of the problems are indeed due to the sheer difficulty to implement proper high DPI/mixed DPI support, but some mentioned in the GP are definitely inflicted by the X11 protocol:
>Spanning two monitors results in terrible scaling problems.
>Apps that do account for it at start up won't account for it during reposition, so they look fine if they open on the right monitor and terrible if moved to the other.
Especially the second point. Applications on Wayland simply get told what scale they should draw on. No need to determine whether they are on the right monitor or the other monitor, yada yada. For those problems, how can you say X11 is not bad, when clearly other protocols have shown the problem is solvable?
I've been running a mixed DPI setup for a long time now, with a 4K display center, and a vertical 1080p display on either side (so 3 displays total).
My conclusion from running this and similar mixed DPI setups over many years is that mixed DPI is extremely poorly tested, if at all, by all vendors. KDE Plasma on Wayland finally pretty much gets it right 98% of the time. X11 wasn't a great experience, regardless of what's technically possible.
Windows has so many annoying issues. Apps are often blurry on the low DPI displays. The mouse cursor has no concept of screen DPI, and treats the entire working area as having the same DPI, so you have to hit the "exits" of the high DPI display just right to land on the low DPI ones. Positioning a window across different displays only scales it correctly on one. There's probably more, but I've been 100% Windows free for a couple of years now.
I wonder how OS X handles it. I don't like that OS, but it sounds like the kind of thing Apple would care about getting right.
macOS (and iOS/iPadOS) is one of the least "technically correct" when it comes to DPI (it uses only the render-at-integer-multiple-and-scale approach) but it does support per display settings pretty well. Everything is also built with the assumptions displays will be high DPI so low DPI screens have crap fonts and whatnot. For windows across multiple displays I'd have to check to see the behavior of but I wouldn't hold up hope it's particularly great. Like you say, Microsoft Windows has one of the most ideal technical implementations (fractional per monitor DPI + it at least attempts to display the window across monitors in a way doesn't bitmap scale) and reached this state many years ago but it means nothing if all the apps aren't also updated to support it (much like this article).
The cursor thing isn't really to do with DPI, it's a general thing with mixed reported display sizes regardless if they have the same DPI. I wish more systems had the option to cross borders at the relative position between monitors rather than the absolute but neither is necessarily more correct and I'm sure many prefer the absolute method.
> macOS doesn't handle it at all, a window is never visible on multiple screens at the same time.
It absolutely is during dragging. macOS is perfectly capable of drawing a window on two screens at the same time, but it doesn't let a window cross two screens while resting which I think is a really good user experience choice.
> Positioning a window across different displays only scales it correctly on one.
Apart from automatic OS-level scaling applied as a post-processing step, which is almost guaranteed to look bad, this one is basically impossible to fix (from a technical standpoint). If you need to move "smoothly" between monitors, get identical monitors.
I'm not sure where all these "X11 is actually great" posts recently are common from but X11 does not have DPI solved. I encourage you to read this comment [1] to better understand what the actual situation is like.
Well, I guess I live in some other universe, because X11 really does "just work" for everything I want it to do. I don't really care about X11 vs. Wayland as such, but I have a bunch of WM-related tools that only work on X11 (and is not the sort of thing that can work in xwayland), so...
Aside: I don't know what it is about the Asahi Linux people that they always need to have the most aggressive takes, assuming hostile malicious intent from anyone who doesn't agree with them or doesn't share the same priorities. These people are just absolutely toxic.
"If you think this idea is a bit stupid, shed a tear for the future of the display servers: this same mechanism is essentially how Wayland compositors —Wayland being the purported future replacement for X— cope with mixed-DPI setups."
A bit disingenuous to quote an outdated fact. Wayland has supported "proper" DPI (i.e. one without downscaling) for quite a while now, and it's been widely implemented across the ecosystem (on the server side: KDE, GNOME, wlroots-based compositors, Smithay-based compositors, on the client side: GTK, Qt, SDL).
aw I thought one of the advantages of Wayland is better DPI scaling
now where did I hear that, many years ago..
sad that corporate interests (or wetf) are blocking X11 updates, though I don't trust the quality of Xlibre given the regressions the dev introduced previously
was planning to go back to Sway once I finally get a 4K monitor..
I think the fundamental design flaw with the scale all screens approach that wayland(and mac and X11 when set that way) employs is that your application is unable to have better dpi aware rendering. however that may actually be an advantage because it means your application does not need a better dpi aware rendering engine, that is things just work everywhere, it may look theoretically worse in it's scaled context, but at least the application writer does not have to do anything so everything just works.
The fundamental issue that it's a hard problem (no matter if on X11/Wayland/Windows/Mac), which can only be solved by the apps and toolkits themselves.
It's only a hard problem if you want to make it one. Mac and Wayland have a pretty elegant and simple solution that does not complicate things. There is nothing wrong at rendering at higher DPI and scaling down. But that's not the problem with X11, the problems there are deeper as explained in the comment I referenced.
There's nothing wrong with OS level down-scaling? Can you show examples screenshots and videos of this? I will assume that
- it will require a lot more memory for cached textures etc. since these are created in a virtualized pixel space
- it will look more blurry and wrong, most importantly the text will be less than crisp.
Apart from that, in many situations you have to scale _up_. Try any older random Windows application (that hasn't declared "DpiAwareness"), even older Windows system dialogs, on a HiDpi monitor (say 27" 4K which is about ~163dpi). They get up-scaled by the system and they look _bad_.
I think this post is one of those "technically correct" but functionally wrong rants.
My experience with multiple monitors at different dpis hits nearly every case of failure he points out.
It's a lot more work to configure.
Apps fail to account for it.
Spanning two monitors results in terrible scaling problems.
Apps that do account for it at start up won't account for it during reposition, so they look fine if they open on the right monitor and terrible if moved to the other.
Getting solid workflows for flexible positioning requires hacks like mentioned at the end for xrandr.
Etc...
So sure, you can do it and it sucks.
That's not really the win I think the author seems to think it is.
"Technically correct" about fundamental limitations of reality but you still hate the truth and it's X11's fault?
I'm not even trying to defend X11, I don't have much love for it. I've done some Xlib programming in the past and I've hated it. I've never used Wayland, and I'm mostly on Windows these days.
But, I don't see how one could make a point that X11 is bad because of poor DPI support.
Some of the problems are indeed due to the sheer difficulty to implement proper high DPI/mixed DPI support, but some mentioned in the GP are definitely inflicted by the X11 protocol:
>Spanning two monitors results in terrible scaling problems.
>Apps that do account for it at start up won't account for it during reposition, so they look fine if they open on the right monitor and terrible if moved to the other.
Especially the second point. Applications on Wayland simply get told what scale they should draw on. No need to determine whether they are on the right monitor or the other monitor, yada yada. For those problems, how can you say X11 is not bad, when clearly other protocols have shown the problem is solvable?
> Applications on Wayland simply get told what scale they should draw on.
xrandr ?
TIL the major reason mixed DPI is subpar on X11 is because Gnome is blocking support on Gtk.
I've been running a mixed DPI setup for a long time now, with a 4K display center, and a vertical 1080p display on either side (so 3 displays total).
My conclusion from running this and similar mixed DPI setups over many years is that mixed DPI is extremely poorly tested, if at all, by all vendors. KDE Plasma on Wayland finally pretty much gets it right 98% of the time. X11 wasn't a great experience, regardless of what's technically possible.
Windows has so many annoying issues. Apps are often blurry on the low DPI displays. The mouse cursor has no concept of screen DPI, and treats the entire working area as having the same DPI, so you have to hit the "exits" of the high DPI display just right to land on the low DPI ones. Positioning a window across different displays only scales it correctly on one. There's probably more, but I've been 100% Windows free for a couple of years now.
I wonder how OS X handles it. I don't like that OS, but it sounds like the kind of thing Apple would care about getting right.
macOS (and iOS/iPadOS) is one of the least "technically correct" when it comes to DPI (it uses only the render-at-integer-multiple-and-scale approach) but it does support per display settings pretty well. Everything is also built with the assumptions displays will be high DPI so low DPI screens have crap fonts and whatnot. For windows across multiple displays I'd have to check to see the behavior of but I wouldn't hold up hope it's particularly great. Like you say, Microsoft Windows has one of the most ideal technical implementations (fractional per monitor DPI + it at least attempts to display the window across monitors in a way doesn't bitmap scale) and reached this state many years ago but it means nothing if all the apps aren't also updated to support it (much like this article).
The cursor thing isn't really to do with DPI, it's a general thing with mixed reported display sizes regardless if they have the same DPI. I wish more systems had the option to cross borders at the relative position between monitors rather than the absolute but neither is necessarily more correct and I'm sure many prefer the absolute method.
macOS doesn't handle it at all, a window is never visible on multiple screens at the same time.
Low DPI screens are near unusable anyway on macOS without subpixel rendering
> macOS doesn't handle it at all, a window is never visible on multiple screens at the same time.
It absolutely is during dragging. macOS is perfectly capable of drawing a window on two screens at the same time, but it doesn't let a window cross two screens while resting which I think is a really good user experience choice.
Ha, so that's just one issue instead of dozens!
Take that, Apple haters.
> Positioning a window across different displays only scales it correctly on one.
Apart from automatic OS-level scaling applied as a post-processing step, which is almost guaranteed to look bad, this one is basically impossible to fix (from a technical standpoint). If you need to move "smoothly" between monitors, get identical monitors.
> X11 wasn't a great experience, regardless of what's technically possible.
I use FreeBSD as my daily driver and as well I use four screens, 2x4k and 2x1080p.
There are glitches and it's no near perfect but I would highly praise Xorg/X11. I've had no issues in a long time; maybe it's your distro.
I'm not sure where all these "X11 is actually great" posts recently are common from but X11 does not have DPI solved. I encourage you to read this comment [1] to better understand what the actual situation is like.
[1]: https://lobste.rs/s/ceylzx/forbidden_secrets_ancient_x11_sca...
Well, I guess I live in some other universe, because X11 really does "just work" for everything I want it to do. I don't really care about X11 vs. Wayland as such, but I have a bunch of WM-related tools that only work on X11 (and is not the sort of thing that can work in xwayland), so...
Aside: I don't know what it is about the Asahi Linux people that they always need to have the most aggressive takes, assuming hostile malicious intent from anyone who doesn't agree with them or doesn't share the same priorities. These people are just absolutely toxic.
The major fundamental issues are there for Wayland too.
Most of the comment is a rant about "X11 proponents". Can we please at least try to keep identity politics out of display servers?
> The major fundamental issues are there for Wayland too.
Which fundamental issues does Wayland have with DPI? I'm not aware of there being any fundamental issues with the DPI handling there.
A quote from the article.
"If you think this idea is a bit stupid, shed a tear for the future of the display servers: this same mechanism is essentially how Wayland compositors —Wayland being the purported future replacement for X— cope with mixed-DPI setups."
A bit disingenuous to quote an outdated fact. Wayland has supported "proper" DPI (i.e. one without downscaling) for quite a while now, and it's been widely implemented across the ecosystem (on the server side: KDE, GNOME, wlroots-based compositors, Smithay-based compositors, on the client side: GTK, Qt, SDL).
aw I thought one of the advantages of Wayland is better DPI scaling
now where did I hear that, many years ago..
sad that corporate interests (or wetf) are blocking X11 updates, though I don't trust the quality of Xlibre given the regressions the dev introduced previously
was planning to go back to Sway once I finally get a 4K monitor..
I think the fundamental design flaw with the scale all screens approach that wayland(and mac and X11 when set that way) employs is that your application is unable to have better dpi aware rendering. however that may actually be an advantage because it means your application does not need a better dpi aware rendering engine, that is things just work everywhere, it may look theoretically worse in it's scaled context, but at least the application writer does not have to do anything so everything just works.
The fundamental issue that it's a hard problem (no matter if on X11/Wayland/Windows/Mac), which can only be solved by the apps and toolkits themselves.
It's only a hard problem if you want to make it one. Mac and Wayland have a pretty elegant and simple solution that does not complicate things. There is nothing wrong at rendering at higher DPI and scaling down. But that's not the problem with X11, the problems there are deeper as explained in the comment I referenced.
There's nothing wrong with OS level down-scaling? Can you show examples screenshots and videos of this? I will assume that
- it will require a lot more memory for cached textures etc. since these are created in a virtualized pixel space - it will look more blurry and wrong, most importantly the text will be less than crisp.
Apart from that, in many situations you have to scale _up_. Try any older random Windows application (that hasn't declared "DpiAwareness"), even older Windows system dialogs, on a HiDpi monitor (say 27" 4K which is about ~163dpi). They get up-scaled by the system and they look _bad_.