- cross-posted to:
- [email protected]
- cross-posted to:
- [email protected]
Source: https://linux-hardware.org/?view=os_display_server
Reporting is done by users who voluntarily upload their system specs via
# hw-probe -all -upload
Source: https://linux-hardware.org/?view=os_display_server
Reporting is done by users who voluntarily upload their system specs via
# hw-probe -all -upload
So not skewed at all…
Do you have a better way of measuring it?
In what direction would voluntary self-reporting of all system specs skew the display server statistic (and why)?
No better way of measuring doesn’t mean this is a good way of measuring.
What way do you imagine would be more precise?
A method that attempts to collect data from a randomized or representative population rather than relying on self-report.
The fact that you need consent to get this data would make a randomized approach impossible.
Yes. It just may be possible that accurate poll data on such things isn’t possible.
Steam hardware survey but that will skew towards gamers. That said, it would be a good indicator on how compatible Wayland is.
The Steam hardware survey will skew towards whatever it is the Steamdeck uses in the surveyed categories.
Could always go for opt-out instead opt-in metrics. Fedora had some recent controversy with it.
canonical has been doing this for years too, and a significant portion of linux users are on ubuntu. i’m not sure if a good portion of users enable it though.
Yeah, this is pretty textbook selection bias.
Ok. How? Describing it is easy. What’s your implementation?
I don’t have one. WTF? You expect that I should be able to propose a way of gathering accurate data just because I can see obvious flaws in their polling methodology?
Unavoidable analytics, apparently. Yay?
Well do you want useful stats or not /s
But seriously, a lot of opt-in (that never get opted in to) data is insanely useful for developers, but it has such a bad stigma that we never get anywhere close to the amount of usefulness a larger dataset could provide.
Tbf a lot of that stigma has to do with trust violation.
I like the way kde does it. On first install it gives a slider with how much analytics you want to send. I just do all of it because I trust KDE, but it’s nice that it asks you. They probably have some pretty good data.
This is the important point IMHO. This kind of feedback is exactly something I’d love to do, but I don’t think I had any idea about it before this post. Just a little popup on a new install/upgrade would be a much broader net.
I imagine people who care about this sort of thing are more likely to report it. And people who care about this sort of thing are also more likely to be early adopters and go through the effort of switching to Wayland.
The way to get a more random sample is not something I want (built-in, automatic telemetry by default). So I’m fine with having skewed data for something like this.
Its a pretty good survey and has a good sample size. Statistics is hard. I won’t take the criticism too seriously.
err, why? actually it can be skewed against wayland(wayland users tend to be more security aware), and why the suprise, KDE, GNOME are wayland from the get go, steam deck too, hyprland and sway etc
It can skew either way equally. We’re just left to do armchair psychology about the type of people who would submit data to this site. So the numbers are effectively useless.
You’re discounting the trend here. Assuming the methodology is consistent, over a short time we’re seeing a noticeable change, bias or not.
I’m not actually. Does anybody doubt that wayland use is increasing? Distros have increasingly been making it the default. I’d be surprised if use weren’t increasing. In fact it might be under-represented in this data depending on whether all distros are being accurately represented or not.
But the change in the numbers is not useless since the psychology of the Wayland users vs. x11 didn’t change
That seems probable but was there any doubt that Wayland use is increasing? Wayland has been changing to the default distro by distro. The only reason this is “news” is because somebody has claimed that “Wayland usage has overtaken X11”.
fair
Which is probably close to every Linux user who knows about it…
That will definitely be part of it. It’s going to be some cross-section of “people who know about it” and “people who are motivated to have their data recorded” which is going to skew the data in ways we can’t reliably understand. Maybe “newbies” are more likely to report than grizzled neckbeards? Maybe desktops are over-represented vs. servers? Maybe one distro lets its users know about it and so its defaults are over-represented? We can’t know.
I mean, it’s not like it’s sensitive information.
hw-probeexcludes such information.Definitely not, most newbies don’t even know about it. It’s really useful for gathering general hardware system info for bug reports, and I often end up having to tell newbies about it.
Absolutely true, if you look at the database it’s vast majority desktop and laptop systems.
I know of no such distro, I’ve hopped between many popular and unpopular distros over the years, haven’t found a single one that does it.
Maybe an abscure distro, but it’s impact would be questionable.
Also, there’s a point to be made in the other direction. The command you need to run is :
sudo -E hw-probe -all -uploadWithout -E the environment isn’t preserved and it’ll think you’re on X11, despite being on Wayland.
I know about the tool. I’ve used it. I don’t report my system data. I’m not “motivated” to have it recorded. I couldn’t care less about their data gathering.
Since you’re relying on people actively reporting their data they need to be motivated to actually do it. That doesn’t mean they’re afraid of what is being gathered (though have you seen the Linux community?) just that they haven’t, for whatever reason, taken the time to do so.
For the rest of it - I was just giving sample potential sources of bias. I wasn’t proposing any of those as actual flaws. Just that their polling methodology couldn’t account for any of them or whatever actual biases may exist in their data. It’s just a list of self-report crap.
Wait Steam Deck now runs Desktop mode in Wayland?
plasma do, unless valve changed that
On launch Steam Deck had it’s desktop/Plasma session set to X11, hence my question
yep, plasma was still x11 from default when steam deck launched, plasma 6 switched to wayland as default, now i don’t know if steam deck was updated to plasma 6
I just did that, why not, but it misreported my DE anyway, so I’d take the OP post with quite a grain of salt.
Why would it be skewed? What would be the cause for a subset of linux users, that upload hardware probes with extraneous information about their display server, to skew the extraneous data?
Anti Commercial-AI license
Because a huge portion of the people willing to do this are already on Wayland, but I believe there exists an even larger percentage on X that are not submitting any data.
And another commenter said:
What is the basis for that assumption?
So because one cannot know which type of people submit data to the site it should be disregarded? That’s basically saying any poll or questionnaire with anonymous yet unique answers are invalid. That’s a pretty bad argument.
Anti Commercial-AI license
This is basically a survey or poll. You want people to provide you with data about what they’re running. To get an accurate view of the entire population you need a representative and randomized sample. If you’re relying entirely on self-reported data you’re not going to be getting a reliably randomized subset of people. You’ll get people who are motivated to report their usage to a third party. That can lead to persistent biases in the data.
It may be that Wayland use is being under represented because the people reporting want to show that “X11 is still king!” Or it could be that this website is shared frequently with certain user groups (e.g. in some arch (btw) forum or something) and so you’re getting a skew towards that population and away from the whole.
We don’t know who these users are and we can’t “offset” for those factors. And the data isn’t reliably randomized so it’s subject to those biases whether we know about them or not.
Though as another person pointed out the trend itself may be of some interest if the population being polled is consistent. Though I doubt anybody suspected that Wayland use is NOT increasing?
Anonymous polls are indeed useless for several reasons.
Man I spent 4 paragraphs saying what you just said in one sentence. 😅
by default, your content is all rights reserved, the most restrictive license possible. AI trains on “all rights reserved” content all the time. You really think adding a CC-BY-NC is gonna do anything?