(Historic wallpapers)

My primary request is to have one consistent visual theme between releases, as in GNOME or Ubuntu. I don't like how our wallpaper currently changes unpredictably between releases.

My secondary request would be for the style to be abstract (e.g. F18-F23 except F21, or F32, or F35) rather than artistic (e.g. F33, F34, F36), but this is a less important request because it's possible for the artistic style to look really good (I would be happy if every wallpaper used the watercolor style from F34). There is so much room for personal preference here that I doubt we'll ever come to a consensus on which style is best.

• f*-backgrounds (the default for each release)

I think these packages are OK, but I'd like to move away from installing them by default. Instead, leave them around for users to install only if desired. Our default wallpaper should come from a new package that updates every release, so if you upgrade from F35 -> F36 and had previously selected the default wallpaper, you get the new default wallpaper automatically instead of keeping the F35 wallpaper after upgrade. Also, accumulating old release wallpapers based on the number of times you've previously upgraded is weird. (I can tell that my last reinstall was F32 because that's the oldest release wallpaper I have available....)

• fedora-workstation-backgrounds

Not sure why we need this. GNOME provides extra backgrounds too, which we remove by subpackaging them into the gnome-backgrounds-extra subpackage that we don't install. Having a different set of extra backgrounds from upstream doesn't seem super useful since none of these have any relation to any GNOME or Fedora visual style. We could just contribute our extra backgrounds to upstream if desired?

Regardless, I don't see the extra backgrounds as particularly important compared to our default.

• gnome-backgrounds

These are nice to have.

We discussed this ticket at today's WG meeting. Main takeaways so far:

• There's some desire on the group for the default wallpaper to be more identifably "Fedora", perhaps by using a consistent style from release to release
• We want to ensure that the wallpapers meet our technical requirements, in terms of appropriateness for different display sizes, good fit with standard desktop UI elements, etc
• I think we do want the wallpaper selection to be actively maintained

I have some notes on technical wallpaper requirements, that I've promised to add to the WG docs.

I've added some background creation guidelines to the GNOME HIG. So far they are mostly technical and practical, as opposed to artistic.

I'll share the new guidelines with the Fedora design team and see if they have any feedback.

There are two outstanding questions which it would be good to have help from the WG on:

1. What should the maximum size of individual background images be? We have some interest in making the images bigger, in order to better support portrait displays, but there's a concern not to use too much GPU memory, but we're uncertain what the limit should be.
2. What should the maximum disk size be for our wallpaper set? I think we might be interested in expanding the range of wallpapers that we make available, but we obviously don't want to inflate the size of the install media too much.

Seems like the backgrounds are around 80 MB in F35 and 54 MB in F36 (the size decrease is probably due to some of the wallpapers now being provided as SVGs).

Second question first: honestly, I think we shouldn't worry too much about disk size. I mean, no sense in going crazy, but I think it's reasonable for us to go beyond the current limit (for Workstation). We're still under 2GB, and I just checked and Ubuntu's desktop download is around 3GB.

We shouldn't be gratuitous, but the wallpaper set is removable for cases where disk space is important, so I think it shouldn't be our primary concern.

I have very different opinions about this for Cloud deliverables, but in this context.....

For the first question: I think GPU memory consumption is a valid concern (https://gitlab.gnome.org/GNOME/mutter/-/issues/1271 -- seems like texture compression would be nice but is complicated!). Can we use the different-aspect-ratio option to address monitors in portrait orientation? (A lot of the wallpapers could use a different layout or crop for those cases anyway.) That's more space on disk (and on the network), but at least that is compressed.

I think we shouldn't worry too much about disk size. I mean, no sense in going crazy, but I think it's reasonable for us to go beyond the current limit

The reason I'm asking about disk size is that we could decide that we want to dramatically expand the number of wallpapers. It's not hard to imagine a situation where we have, say, 60 wallpapers, comprised of three or four different sets.

I'm less concerned with the total image size than I am with the proportion of the image used by the wallpapers. I don't think people would be too impressed if 50% of their download was wallpapers. :smile:

Can we use the different-aspect-ratio option to address monitors in portrait orientation?

One question I have here - can that approach can be used when display orientation is dynamic? Like, is the wallpaper image only picked when the display is configured, or will it be selected on the fly when the orientation changes?

is the wallpaper image only picked when the display is configured, or will it be selected on the fly when the orientation changes?

According to @fmuellner , it should be updated on the fly. So having multiple wallpaper images for different aspect ratios does seem viable.

Some updates on this issue:

• We're going to try using webp for the bitmap wallpaper files, both in GNOME and Fedora, in order to keep the file size down
• We're also going to try and use 4096x4096 resolution for the wallpaper images. If this turns out to use too much memory, we'll generate a portrait and landscape version of each wallpaper, to use instead of the square one.
  • The anticipated challenge here is photographs, particularly of landscapes. Many of these won't be 4096 tall. We might have to just get as close as we can.
• Mairin's putting a new initiative together to refresh the supplemental Fedora wallpapers. See: https://gitlab.com/fedora/design/extra-default-wallpapers
  • This will experiment with a combination of abstract images and photography, to see what can be achieved in terms of meeting our goals for image size as well as supporting dark mode. Only supporting dark mode with the abstract images is an option, depending on how the photos come out.

According to @fmuellner , it should be updated on the fly. So having multiple wallpaper images for different aspect ratios does seem viable.

That's based on glancing at gnome-desktop, gnome-shell and mutter code, not on actual testing.

That said, if it doesn't work I'd consider that a bug that needs fixing.

I'd like to propose not using the Fedora 37 default wallpaper in Fedora Workstation. It's inconsistent with our brand identity goals. Instead, either design something altogether different for Fedora Workstation, or, alternatively, just use the default wallpaper from GNOME 43.

I do not see us as having a separate brand identity goals from the rest of Fedora. Consequently, I firmly disagree on the idea of using a different wallpaper from the rest of the Fedora desktop deliverables.

Also, this is a tremendous amount of disrespect for the work that the Fedora Design Team does every release cycle to produce something that allows Fedora to stand out from the rest of the distributions. Especially given the incredible restrictions that GNOME gives us for branding, I think it's more than fair that we use the Fedora release wallpaper for our releases.

F37: https://raw.githubusercontent.com/fedoradesign/backgrounds/main/default/f37-01-day.png

F36: https://raw.githubusercontent.com/fedoradesign/backgrounds/f36-backgrounds/default/f36-01-day.png

Also, this is a tremendous amount of disrespect for the work that the Fedora Design Team does every release cycle to produce something that allows Fedora to stand out from the rest of the distributions.

It's pretty artwork, to be sure, but it's too far afield from what a professional computer operating system should be aiming for. You wouldn't see Ubuntu doing stuff like this.

F37: https://raw.githubusercontent.com/fedoradesign/backgrounds/main/default/f37-01-day.png

F36: https://raw.githubusercontent.com/fedoradesign/backgrounds/f36-backgrounds/default/f36-01-day.png

Also, this is a tremendous amount of disrespect for the work that the Fedora Design Team does every release cycle to produce something that allows Fedora to stand out from the rest of the distributions.

It's pretty artwork, to be sure, but it's too far afield from what a professional computer operating system should be aiming for. You wouldn't see Ubuntu doing stuff like this.

And yet, Fedora artwork regularly gets noticed while the Ubuntu one doesn't.

The design team uses an open, community process which includes a lot of room for feedback along the way. Please work with the design team if you feel like something isn't in line with the edition goals. Obviously, Workstation is a key stakeholder -- but that goes both ways!

Action item: Allan to discuss establishment of a more consistent style/theme Fedora wallpapers with relevant community contributors

Action item: Allan to discuss establishment of a more consistent style/theme Fedora wallpapers with relevant community contributors

Any updates?

Discussed again today. Action item is the same: Allan to discuss establishment of a more consistent style/theme Fedora wallpapers with relevant community contributors

I said that I'd provide a status update on this issue, so here's what I know:

• Regarding the default wallpaper:
  • The WG has previously expressed a desire to have a thematic approach so the wallpapers have a common style from release to release. Talking that idea over with the design team is still outstanding (and is on my to do list once again).
  • In previous releases there was a community process around the default wallpaper. That didn't happen for F39 and we had a last-minute scramble to put a default wallpaper together. It would be good to find out what exactly happened in F39 and what the plan is for F40.
• Regarding the additional optional wallpapers that we include:
  • For many years we included the same wallpaper set which wasn't updated, and had some quality issues associated with it. After talking to Mo about this there were some improvements for F37 with the introduction of some new wallpapers.
  • Since the F37 improvements there hasn't been any sign of further improvements and the previous extra-default-wallpapers initiative has seen activity in a year. I also notice that these wallpapers are still being shipped as .webp - something that GNOME stopped doing due to performance issues.

Any objections to getting rid of gnome-backgrounds-extras and moving those wallpapers into the main gnome-backgrounds package? Seems silly to put so much effort into designing the wallpapers and then not actually include them.

So a couple days ago there was a last-minute attempt to change the F40 default wallpaper to this. It has been deferred until F41, but frankly, I think we've reached the point where we just need to stop using the wallpapers developed by Fedora design. We should address this sooner rather than later, so there's plenty of time to develop a contingency plan.

So a couple days ago there was a last-minute attempt to change the F40 default wallpaper to this. It has been deferred until F41, but frankly, I think we've reached the point where we just need to stop using the wallpapers developed by Fedora design. We should address this sooner rather than later, so there's plenty of time to develop a contingency plan.

These are an incredibly important part of Fedora brand identity. I'd really encourage anyone in Workstation who is concerned about them to get involved in the Design Team community process by which they are created.

That said, obviously there was a process problem this time around. We're working on that — the goal is to actually be a step ahead, with the new wallpaper landing (ideally) at the branch, so when Rawhide is F41, it will have the planned-for-F41 wallpaper (e.g., now) and the F42 wallpaper should land sometime this August, giving _plenty of time for any issues to be worked out by next April.

These are an incredibly important part of Fedora brand identity.

No they're not; the wallpapers are currently something completely different each release, so there's no possibility to build a brand around them. This is in contrast to, say, GNOME or Ubuntu. The Working Group has previously expressed interest in changing this.

I'd really encourage anyone in Workstation who is concerned about them to get involved in the Design Team community process by which they are created.

That's one possible outcome here, but Workstation WG would probably need a volunteer to do this.

Any objections to getting rid of gnome-backgrounds-extras and moving those wallpapers into the main gnome-backgrounds package? Seems silly to put so much effort into designing the wallpapers and then not actually include them.

If anybody doesn't like this, please complain. The impact would be more non-default wallpapers installed by default.

Why were they split?

Guess: to save space on the live image?

Discussed today:

• Agreed: maintain the gnome-backgrounds-extras package (i.e. do not ship the rest of the GNOME default wallpapers) due to performance problems with the wallpapers panel in gnome-control-center; revisit if performance problems are resolved
• Further discussion required to decide what to do about Fedora default wallpapers

Discussed today:

• Agreed: maintain the gnome-backgrounds-extras package (i.e. do not ship the rest of the GNOME default wallpapers) due to performance problems with the wallpapers panel in gnome-control-center; revisit if performance problems are resolved

Are there upstream issues about this in particular? There has been a recent rework of the Background/Wallpaper loading with a focus on performance, this might no longer be an issue in gnome-46 (fedora 40).

It would be good to at least confirm that the performance issues are an issue with the number of wallpapers we are interested in shipping.

Also note that there's some upstream design work that would address this issue: https://gitlab.gnome.org/Teams/Design/settings-mockups/-/issues/62

Would be interested in hearing about the gnome-backgrounds specific performance issues. The performance improvements in Settings is discussed here -- https://gitlab.gnome.org/GNOME/gnome-control-center/-/merge_requests/2057

It would be good to at least confirm that the performance issues are an issue with the number of wallpapers we are interested in shipping.

Performance has improved by order of magnitude, both for panel resizing and opening in Settings, see the MR linked in the previous comment. That's out since GNOME 45.2.

So I really don't think that's still an issue that should hold anyone back from adding more wallpapers.

I was Jens who was noticing performance issues. I've asked him to report a bug. (I believe he was already using GNOME 46.)

From my perspective, Fedora shipping the GNOME branded default and skipping on the rest of gnome-backgrounds is a big misstep.

The default is tied heavily with the GNOME brand, while the others that would help in providing a fresh set of mixed/varied wallpapers is left to vanilla shipping distros like Debian or VanillaOS. I would do the exact opposite to the current status quo, ie ship all the supplemental wallpapers from gnome-backgrounds and perhaps skip the gnome default.

I was Jens who was noticing performance issues. I've asked him to report a bug. (I believe he was already using GNOME 46.)

Jens, did you get a chance to report a bug for this?

The default is tied heavily with the GNOME brand, while the others that would help in providing a fresh set of mixed/varied wallpapers is left to vanilla shipping distros like Debian or VanillaOS. I would do the exact opposite to the current status quo, ie ship all the supplemental wallpapers from gnome-backgrounds and perhaps skip the gnome default.

I don't think shipping the GNOME default hurts anything. We can probably just ship everything, after the performance issues are fixed.

Agreed: Fedora Workstation wants future default wallpapers to reflect either a Fedora or a Fedora Workstation brand identity. We are no longer interested in shipping generic wallpapers that change dramatically release to release as we have historically done.

• "Brand identity" does not mean "use Fedora or Fedora Workstation logo in the wallpaper" but rather "maintain some sort of consistent style from release to release." This doesn't mean all wallpapers need to look extremely similar, like Ubuntu (although it could).
• We do not yet have consensus on whether to use a Fedora identity (one wallpaper for all spins/editions) or a Fedora Workstation identity (where we no longer use the same wallpaper as other Fedora spins/editions).

We also certainly do not want to see major changes to the default wallpaper after the wallpaper deadline (which for Fedora 41 is July 23).

Also agreed: we will include all the supplemental wallpapers from gnome-backgrounds-extras after performance concerns with the wallpaper settings panel are resolved. Jens to report an upstream issue.

I did some testing today in a VM: it seems to me the very first time the backgrounds are loaded (in F40) gnome-control-center takes a while as they populate slowly. After that they seem to appear quickly, but I need to do a little more testing.

OT but actually a worse problem to me seems to be gnome-software grabbing a CPU for a while as soon as gnome-shell starts, which is quite annoying: makes the whole desktop start up quite laggy in a vm, or small machine I imagine.

I tested more in a fresh install of F40-WORK-x86_64-LIVE-20240520.iso in a VM with 2 cpus and 4GB.
In the default Workstation, after clicking on "Change Background..." it takes a couple of seconds for the wallpaper thumbnails to populate.

But after installing gnome-backgrounds-extras, it is more stuttery, and takes some seconds for all the wallpapers in the list to fully appear. I guess it is not the end of world - might well be faster on a beefier machine.
The second time one presses "Change Background..." it takes less than half a second to show them all.

So this was basically with GNOME 46.1, before upgrading to latest 46.2.
Also note I made sure to wait for the instance to be idle (with top) before starting the "Change Background".

As an update from the performance side of things; thanks to detailed profiling by Jeff Fortin and with help from Christian Hergert, we managed to greatly cut down on the initial loading of the wallpapers in GNOME Settings. The main thread is now not blocked anymore. This mostly affected initial loading, when thumbnails don't exist yet, but also loading from a "hot" cache is now smoother and faster.

See https://gitlab.gnome.org/GNOME/gnome-control-center/-/merge_requests/2807

The bottleneck is now mainly file IO inside libgnome-desktop, so I don't see easy ways to improve further.

Wallpapers for F41 are available now, and they're the same ones from the end of the F40 release cycle: https://gitlab.com/groups/fedora/design/-/epics/34#note_2114915766

My suggestion is we decline to use these as they are inconsistent with the desired branding goals for Fedora Workstation.

It's too late now, as these have now shipped on the Beta ISOs.

We cannot time travel and change it for beta, but we can certainly change it after beta.

• Agreed: maintain the gnome-backgrounds-extras package (i.e. do not ship the rest of the GNOME default wallpapers) due to performance problems with the wallpapers panel in gnome-control-center; revisit if performance problems are resolved

Hi @petersen could you confirm whether the performance issues you noticed are resolved?

I think my earlier conclusion was that it is probably okay to add gnome-backgrounds-extras:
I think the performance is not a showstopper and it is only a off-thing after install/upgrade.
So we could go ahead and try and get wider feedback?

But I will try to retest with F41 anyway.

Okay I may need to test in a fresh vm perhaps, but currently with gnome-backgrounds-extras installed in my 4GB f41 VM, gnome-settings-daemon seems to oomd for me before displaying its wallpapers.

We discussed this again today and reaffirmed our previous decision. Allan will communicate this to the Fedora wallpaper developers.

Okay I may need to test in a fresh vm perhaps, but currently with gnome-backgrounds-extras installed in my 4GB f41 VM, gnome-settings-daemon seems to oomd for me before displaying its wallpapers.

Yes, that's an unfortunate side effect of the recent async improvements. It's being tracked at https://gitlab.gnome.org/GNOME/gnome-control-center/-/issues/3198. Seems to be a problem with 10 concurrent threads somehow maxing out low memory devices. I suspect the sandboxed thumbnailing is taking a lot of memory.

I had a call with Madeline Peck last week, where we discussed the wallpapers in general.

• I said that we'd provide recommendations for the file formats to be used in the wallpapers. (GNOME uses jxl and Fedora is shipping a mix of webp and png; previously we've been warned off webp at least).
• We discussed the best way for the working group to be aware of the default wallpaper process. These are tracked as epics here. Madeline kindly offered to ping us in our Matrix channel when the wallpaper is ready each cycle, but I feel that a more organic solution would be better. More discussion is needed here.
• We discussed ambitions for the supplemental wallpapers.
  • Upstream we are hoping to have a more capable wallpaper selection UI which will help with this, and we'd like to have expanded number of wallpapers from both GNOME and Fedora.
  • Currently the Fedora supplementals aren't actively maintained, and it's a bit unclear what the official upstream project is (I could only find fedora-workstation-backgrounds in Pagure, when Fedora design is primarily in Gitlab nowdays). Madeline said that she'd look into this.
• We also discussed the current process for the default wallpaper. No specific outcomes there, but I have a better understanding of how things are operating currently at least.

I know that this doesn't cover every topic that the WG wanted discussing, but think that it's a good start to establishing a working relationship.