GNOME 3 drops fallback mode in 3.8 | Hacker News Reader

GNOME 3 drops fallback mode in 3.8

mail.gnome.org

29 pointsEdiX59 comments

This is insane. I dont know who takes decisions in Gnome but has to be fired immediately. Gnome 2 was a nice and fast DE, it ran anywhere, it was great. They made so many bad decisions with Gnome 3, stopped caring for the user long ago and now they code only from themselves.

Comment 13 on http://igurublog.wordpress.com/2012/11/05/gnome-et-al-rottin... suggests Red Hat is doing Embrace, Extend and Extinguish ( https://en.wikipedia.org/wiki/Embrace,_extend_and_extinguish ).

There are only 2 people of Red Hat on the GNOME release team (team is 9 at the moment). But yeah, due to the sheer amount of manpower that Red Hat spends on maintaining various software projects, their influence of course is pretty great.

But they still have to convince the other GNOME developer community. Decisions are pretty often made in consensus, so they don't override things.

Note that the blog you are linking to gets some things wrong. You have to see some things in context. E.g. a designer is of course going to have a bit of limited view.

The Red Hat people on the release team sometimes "block" a decision made by other Red Hat people; seems like normal commity behaviour to me. Though of course if they all work together, their idea of the best way forward is going to be similar.

I was part of the decision. I'm not sure if you get the concept of volunteers. My influence and decision abilities is not based on "company ideas" as you seem to think.

Meaning: I wasn't hired, I cannot be fired.

If I may ask, what is it that drives you to volunteer to work on a project that has so much negativity surrounding it these days?

I've found GNOME 3 to be a very interesting case study. Here we have what was once a very successful project that has gone through a very rough patch lately.

Many existing users aren't happy with what it has become, and although changes are happening, they only seem to be provoking more and more negativity. I can't recall any GNOME 3-related news where the sentiment was positive. At best, it's a sense of "well, things aren't getting any worse!" And while I don't have any numbers, I gather that GNOME 3 isn't really attracting all that many new users, either.

The future doesn't exactly look bright. Based on the comments here regarding this decision, for example, it clearly isn't what the user community wants. There seems to be much more excitement and positivity around projects like MATE and Cinnamon, which basically go against what the GNOME team is doing.

I don't mean to be insulting, but I have a very difficult time understanding why anyone would still want to be publicly associated with the GNOME project these days.

Er, on the other hand, many people actually like gnome 3 (myself included), but it's typically the people who don't like something that are the most noticeable, because they complain a lot.

I.e. "I saw lots of complaints on HN" is not a particularly good measure of anything...

But there are a lot of complaints in a lot of different places.

I'm sure complainers are probably over-represented in public fora, but not to the extent you seem to think.

I don't doubt that some people like GNOME 3. I'm just not convinced that, given all of the feedback that I've seen, that this is a very large group of people. I would not be at all surprised if the number of people who like GNOME 3 is much smaller than the number of people who have, silently or otherwise, switched to Xfce, KDE, or some other environment.

Like Nursie mentioned in another comment, the negative sentiment toward GNOME 3 isn't just present here. It arises essentially wherever GNOME 3 is discussed. Having followed many open source projects for many years, I've never seen anywhere near this level of outright hatred on such a scale. Yes, disagreement is common, but when it comes to GNOME 3 it's far beyond that.

I have followed GNOME since almost the start. Helped out since 2005.

I help out out because it is fun. I know a very large group of people and everything I do is appreciated.

User community for me is Mageia, where I help out as well. The things I do over there are also very much appreciated.

On various sites there is a lot of negativity. E.g. on Phoronix some person ("BOSS" or similar nick) often says GNOME developers should due.

I rather focus on making GNOME better and having fun than deal with the death threats and so on, bad reporting ("Debian defaults to XFCE"), etc.

Being public about helping out with GNOME: Why not? I don't mind explaining why things are done the way they are. Often the assumptions are just that, assumptions. I also go to FOSDEM, totally different atmosphere than displayed on forums.

> Based on the comments here regarding this decision, for example, it clearly isn't what the user community wants.

I like Gnome 3 very much, but I am not constantly writing blog postings and comments about how much I like it.

Just to ensure that complaints are not the only opinions developers like you hear, let me tell you that I use Gnome Shell everyday, and I love it. I migrated a lot of the people in our office (a small Actuarial firm w. ~12 people) to it and everyone likes it.

Same, I'm using it right now on my work laptop and I wouldn't dream of using another desktop environment. It's fast, fluid, and getting more functional with each release.

Drop the fallback mode if it means better quality of the rest. Seriously, Gnome Shell runs fine on my Intel integrated graphics, i suppose it runs fine on nvidia and amd as well, so why fallback mode anyway.

How about communicating to distributions to fallback to gnome2 or xfce or something similar as a recommendation?

It may run fine on many, but it runs horrible on so many others. Running both Nvidia and ATI binary drivers often result in absolutely horrid performance with changes to the UI often not being rendered until after the animations are over.

Apart from that, GDM itself performs horribly. I can almost always see a lag. Shouldn't happen on core i 2nd Gen atleast!

Obviously that the performance is sometimes absolutely terrible is known. I'd wish it was not like that. Llvmpipe is not a good solution, nor are bad drivers.

But IMO bad drivers are unacceptable. Why do we accept such sorry state? We had fallback mode for various years, as well as the need of GNOME 3 and Unity. Yet various drivers still are terrible.

IMO I'd rather take the "my way or highway" approach: a driver is just plain not good enough. The hardware can do way more. Time to get your driver into the year 2005+.

"How about communicating to distributions to fallback to gnome2"

Because Gnome 2 is officially deprecated and abandoned. The G3 team pretty much deliberately set out to destroy it and make the new stuff as incompatible as possible AFAICT. MATE is becoming a possibility but debian don't seem to want to take it on.

Xfce, yes, that works fine.

We did no such thing.

Speaking on behalf of GNOME (and yes, I can! :P)

We actually asked for feedback to distributions various times. We would've made a decision earlier but e.g. OpenBSD would have problems. Also this time there has NOT been any official announcement yet. What we do is mostly in the open, so things are announced beforehand.

Now, with fallback I thought it should be announced in the 3.6 release notes that we were thinking about it. Unfortunately it was taken out, you can still see what I wrote in the git module though.

For distributions, the suggestion is llvmpipe, though that is terrible. What we are more after is better drivers, with llvmpipe as a workaround.

One thing that is overlooked is that not only gnome-shell relies on good drivers. Various applications have started relying on this as well. So the value of fallback mode was already decreasing.

> What we are more after is better drivers

Gnome 2 didn't need any driver to work, it was perfectly fine with a frame-buffer.

You know what else do not need any driver to work? Windows.

Llvmpipe DO NOT WORK except in latest-generation, very expensive hardware. You are alienating a huge portion of your users, it's specially painful because they are the poor-ones.

Please provide reference to that huge portion.

The background of this is that fallback mode did not work very well. Various applications have been requiring the same hardware support.

Obviously at the moment this will cause pain. But I don't understand why there is not more anger towards the sorry state of graphics drivers on Linux. My GPU of 6-7+ years old support GNOME 3 fine. At that time I bought a silent GPU, it was NOT the quickest GPU. Why are we accepting that we have such poor drivers?

The anger should be focussed towards the drivers, not something which tries to use 7+ old technology.

I'm sure you know Gnome ran on a lot of things that are not Linux. BSDs for example. The implementation of gnome-fallback in OpenBSD works beautifully. I have a small ARM board that consumes about 1 watt, and you could install gnome on it and you couldn't tell it from my ubuntu desktop, it was that good.

I expect games to need driver support, not a simple DE. If you take in account that not even OS-X need a graphic card to work, you know you are doing it wrong by requiring one. (Don't start with llvm-pipe, I dont need the DE consuming 100% of the CPU)

I just wanted to be another person to chime and say I enjoy gnome3. Sometimes a good product is best when you build it for yourself instead of trying to please others.

Thanks!

> Meaning: I wasn't hired, I cannot be fired.

Oh I see the problem then. Dude please reconsider it. Do you ever tried to run that stupid LLVM-pipe stuff in a VM? it's impossible, and I have a nice machine. ARM? MIPS? Sparc? forget it.

Show some respect and you'll get respect back.

Yes, I've used it in a VM. Yes, I know about arm and so on. Suggest to focus more on being nicer towards others, less about questioning things where the answer can easily be found. If you're not courteous, why would anyone spoon feed you answers?

Yes I was kind of an asshole. Sorry about that. I understand that you are working for free, I really appreciate the work that you people do, and gnome 2 was great. Actually from all the new DEs, unity, KDE, etc. I still prefer the look of the gnome shell. The problem is, I can't run it! too slow, need too much kernel support, etc. I don't need any answers, I just need the damn thing to work in my OpenBSD and my beagleboard the way gnome 2 did.

Mate Desktop is a fork/rename/continuation of Gnome 2.

If you're the sort of person that used fallback mode and hated Gnome Shell, you'd have served yourself well by jumping ship a long time ago.

All this bodes well for Xfce, which is already an outstanding effort. Nice to see it included in future releases of Debian as default, for some spotlight.

This is wrong, see http://www.debian.org/devel/debian-installer/News/2012/20121....

  - Most of GNOME fits onto CD#1 again (network-less installation).
  - LXDE and Xfce now live on separate images.

GNOME was removed from CD1 due to space reasons, but this was only temporarily (they've meanwhile fixed the space issues). That GNOME didn't fit onto CD1 anymore and that Xfce would be the future was big in the news. That this change was reverted some weeks later, and no Xfce isn't even on the first CD, didn't make the news though. Guess that's not as interesting as GNOME-hate.

Well, if you had submitted it, I'm sure it would get frontpaged here.

I just really wasn't aware of the revert and the git commit message sounded promising. [1]

[1] http://anonscm.debian.org/gitweb/?p=tasksel/tasksel.git;a=co...

All Xfce needs now, is a compositing manager with sync to vblank support to fix those nasty Nvidia tearing issues. Unfortunately the Xfce developers seem to be ignoring this.

You can use Compiz in XFCE, which has reasonably good sync to vblank support.

I know there are quite a few compmgrs that have sync to blank e.g, Compiz, Kwin or Mutter, but they all come with a redicoulus amount of dependencies, which is why I'm avoiding those. I also tried cairo-compmgr but it's really buggy.

That's a shame. I was opting to use fallback mode so I could continue using gnome without gnome-shell.

When 3.8 hits debian unstable I guess I'll move to a different window manager.

Same here. Fallback mode was the only thing that made GNOME 3 usable. The Linux desktop has really died today.

That's the thing. The GNOME project understands itself more as an OS than just as a "window manager".

Most (not all, but most) of the hate against GNOME 3 is not against GNOME, but against gnome-shell and gnome-control-center. These are just a few (though essential) pieces of the GNOME desktop. If you don't like the shell you can still use GNOME's apps from another WM (Xmonad, sawfish, KWin) or even from another DE (like Xfce).

Ditto.

Makes sense, since no one was interested in maintaining it.

I guess on Linux (or any POSIX OS) it's not really necessary, since everything should still be fixable from the terminal. However it reinforces the notion that someone who can't use the terminal (= 99% of users probably) should stick to Windows or Mac - or maybe Ubuntu.

Well, I think part of the idea is that the fairly recent LLVMPipe software driver [1] should be installed by default in distros and is well supported/performant enough so you can install the drivers in the full composited desktop...

1. http://www.mesa3d.org/llvmpipe.html

When installing GNOME3 I had a little chicken-and-egg-problem. Without installing additional drivers GNOME3 wouldn't work, so the fallback mode was handy to search the internet for the correct drivers for my chipset and install them.

But as long as they have some solution for this problem (and I think it is pretty common, because debian always installs with generic drivers for your hardware), I think it is no problem to remove the fallback mode.

Debian Wheezy will default to XFCE (I think), so you can use that instead of fallback mode for this use case.

You may find you just like using it....

IIRC the reasoning was that GNOME Shell is too big to fit on the default install CD, rather than any idealogical thing.

This is wrong, see http://www.debian.org/devel/debian-installer/News/2012/20121...

  - Most of GNOME fits onto CD#1 again (network-less installation).
  - LXDE and Xfce now live on separate images.

Quite a recent change then!

Shame, XFCE would be a good default.

I hope that this decision causes them to re-think their position.

I hate GNOME 3, but I agree with this move. If you like classic desktops, and you have been using Gnome-Fallback you are doing it wrong: GNOME 3 is designed to be used with the Gnome-Shell.

XFCE and Mate are both excellent Desktop Environments which follow a "classic" approach. (Or you can try KDE too).

So what will people running GNOME in VMware see? GNOME failed to launch?

llvmpipe in case VMware doesn't do 3D accelleration.

They'll still see Gnome Shell. They added LLVM support for software-rendering about a year ago: http://www.phoronix.com/scan.php?page=news_item&px=MTAxM...

Fallback Mode was only still in use by people who preferred Gnome 2 and forced it on, everybody else with Gnome 3 saw Gnome Shell.

Per the link, even back then they discussed dropping Fallback Mode after full LLVM had been added:

> "As said by Red Hat's Adam Williamson on the Fedora mailing list, 'That's really a policy decision for the GNOME / Fedora desktop teams, not for ajax. But based on what they've said in the past, I expect that once most hardware that previously needed the fallback mode is covered, fallback mode will die. AIUI, fallback mode isn't meant to be a GNOME 2-by-stealth for Shell refuseniks, it's purely an attempt to accommodate hardware which doesn't support Shell.'"

I had to look this up, and it appears in this context, "fallback mode" refers to "sort of a failsafe mode, which should kick in on machines with unsupported graphics cards".

I guess it turns of compositing and such modernities.

http://library.gnome.org/users/gnome-help/3.3/fallback-mode....

There's a follow up to the linked post that explain the cases where Gnome 3 it's not an option and the fallback mode is OK:

https://mail.gnome.org/archives/release-team/2012-November/m...

Basically: some GPU drivers are not good enough to run Gnome (it was way worse about 2 years ago), some old machines can't deal with the software rendering (that is not available for all the platforms, it's only for x86 and ARM -I think-), and there is some performance impact when running 3D apps on Gnome 3 (I've noticed that myself, my laptop fans go berserk!).

Besides the fallback mode was used by some users that were looking for a Gnome 2 alike experience, but apparently that was not the reason of fallback mode to exist.

The idea is that by dropping fallback, people can change gnome-panel and metacity as they wish. Meaning: making it work (and look) more like GNOME 2.

I'm not sure if this is entirely feasible in practice. Actually fallback mode consist of various changes in loads of projects. I'm guessing it'll be broken pretty quickly in GNOME 3.7.x. The intention was still that you could still get a more GNOME 2 experience by not needing gnome-panel+metacity for fallback mode... but after all the investigations, not sure if that can still be achieved.

For a brief period of time, I used Fedora and Gnome3. Having a radeon hd46XX, gnome refused to display anything properly, except in fallback-mode. I think this should be its use: not a UX choice, but a mode for when graphic acceleration fails.

Xmonad withought a de, thats my cup of tea nowadays. I am content.

i wonder if xmonad's Config.Gnome will get ported away from the fallback mode (if this is possible).

Unfortunately, gnome-shell desktop components are closely tied to mutter. For example, the fallback gnome-panel [1] and the gnome-shell gnome-panel [2] are two entirely different beasts (same goes for cinnamon, where they've never had a fallback panel to begin with). There's just no way to run mutter and xmonad at the same time.

It is becoming ever so difficult to be an xmonad user who enjoys the occasional niceties of modern desktop environments (panels, widgets etc). The state of component reuse in modern linux DE is just miserable.

Kudos to the KDE people for keeping their desktop 'dissectible' after all this time. I've had great success in using plasma-desktop with xmonad in the past. Unfortunately, KDE looks/feels wrong to me in ways that no theme has ever been able to correct (see [3] [4]). I wish they adopted 'flatter' look & feel guidelines - right now, there is just too much bling, roundedness and qt widget animations for my taste (relying on SVGs for theming plasma is overkill too, imho). I have a lot of gripes (still can't get over Akonadi needing mysql) with KDE, but I really do respect the work they do.

We can't have nice things ...

[1]: http://git.gnome.org/browse/gnome-panel/tree/gnome-panel

[2]: http://git.gnome.org/browse/gnome-shell/tree/js/ui/panel.js

[3]: http://upload.wikimedia.org/wikipedia/commons/5/54/KDE_4.png

[4]: http://cinnamon.linuxmint.com/wp-content/uploads/2012/01/1.2...

Are you still using plasma-desktop with xmonad? Do you have some pointers on how to do that? As an xmonad user, I'd like to try it out.

You could either run plasma-desktop through .xinitrc or set KDEWM to xmonad in .xsession. You will have to send plasma-desktop to the float layer. It's best if you make xmonad act like an EWMH compatible WM by using the EwmhDesktops [1] log hook and I think the ManageDocks [2] manage hook also played a role. The relevant xmonad.hs changes are:

    import XMonad.Hooks.ManageDocks
    import XMonad.Hooks.EwmhDesktops

    managehook = composeAll
                 [ className =? "Plasma-desktop" --> doFloat]
    
    defaults = defaultConfig {
             , manageHook = managehook <+> manageDocks
             }

    main = xmonad $ ewmh defaults

There are a lot of dialogs that pop up in KDE applications that you'll probably want to add to the float layer. I haven't tried it, but there is this plasmoid [3] that can offer you better integration with xmonad's way of working with workspaces (xmobar/dzen2 style).

Good luck!

[1]: http://xmonad.org/xmonad-docs/xmonad-contrib/XMonad-Hooks-Ew...

[2]: http://xmonad.org/xmonad-docs/xmonad-contrib/XMonad-Hooks-Ma...

[3]: http://gitorious.org/xmonad-log-plasmoid

Reply on news.ycombinator.com