Confused [LWN.net]

Confused

Posted Nov 4, 2011 14:37 UTC (Fri) by kragilkragil2 (guest, #76172)
Parent article: Rawhide gets GNOME Shell for all display types

Is Adam talking about Gnome or just Fedora.

I am pretty sure I saw a blog post from some Gnome guy recently that said that Gnome3 was designed with fallback mode in mind and that it will stay for a long time.(Sorry, but Google(I) can't find the post atm)

So no fallback in Fedora, or no fallback at all?

to post comments

Confused

Posted Nov 4, 2011 15:55 UTC (Fri) by ovitters (guest, #27950) [Link] (22 responses)

Adam is talking about Fedora of course.

For GNOME, we haven't decided yet. Seems most are switching to other desktop environments if they don't want GNOME shell, so would ease maintenance to not have fallback mode.

Confused

Posted Nov 4, 2011 16:16 UTC (Fri) by hadess (subscriber, #24252) [Link] (17 responses)

> Adam is talking about Fedora of course.
>
> For GNOME, we haven't decided yet. Seems most are switching to other
> desktop environments if they don't want GNOME shell, so would ease
> maintenance to not have fallback mode.

Huh, we've already decided that we don't want to have to reimplement everything twice when we add a new feature to GNOME Shell, or a feature that uses GNOME Shell. For example, there's about 5 reimplementations of system status icons in various modules I maintain, and no fallback support for Zoom, or On-Screen Keyboard a11y features.

It was always going to go away, the only thing we haven't decided is when.

Confused

Posted Nov 4, 2011 17:23 UTC (Fri) by kragilkragil2 (guest, #76172) [Link] (15 responses)

You guys need to get your story straight. Some say it will stay, some are undecided and some say it was always supposed to go away.

So if one likes Gnome programms and some of the Gnome3 tech(notifications etc), but also wants a task bar and simple menu and not Gnome-Shell(and you don't have the time to maintain FBM yourself) one is screwed, right?

Confused

Posted Nov 5, 2011 14:56 UTC (Sat) by ovitters (guest, #27950) [Link] (14 responses)

It is called GNOME fallback for a reason. Always said it would not stay forever. But the initial assumption was that it would stay for entire 3.x, to ease transition.

Seems that people just switch to something else instead. Which is fine as then fallback can be dropped earlier without impact.

Confused

Posted Nov 5, 2011 21:09 UTC (Sat) by kragilkragil2 (guest, #76172) [Link] (13 responses)

>Seems that people just switch to something else instead. Which is fine as then fallback can be dropped earlier without impact.

It _seems_ that way, at least if you look at internet comments. No way to be sure or is there? So for a hunch you might alienate another part of your user base.

I am really not sure this KDE4/Gnome3/Unity thing where you introduce lots of regressions and try to live with the fallout is really the right way of doing things.
FOSS marketing is basically the people who use it, alienating large parts of those people does not seem like the way to success. If every internet post of your new product sees only lots of bitching on the internet it is hard to get the buzz going. KDE hasn't recovered from KDE4 and it has been nearly 4 years. Every KDE4 post sees major whining, just look below.
Slow gradual change where every version builds on the basis of its predessor seems like the much better way of doing things. Linus&Miguel seem to know that. Desktop designers .. not so much.

Confused

Posted Nov 6, 2011 15:28 UTC (Sun) by ovitters (guest, #27950) [Link] (12 responses)

I chose my words charefully. I said seem with a reason. I am aware of everything you think I am not.

Anyway, bringing up a kernel hacker and "lots of regressions" is not going to impress me much. Instead I rather look at GNOME 3.4 and see what we can improve. Just like GNOME 2.x.

We called gnome-panel v3 fallback mode for a reason. I know at least one distribution renamed it to "GNOME classic". That is wrong. You can get the same experience (was intended to make possible), but that was never going to stay forever. This all was communicated before GNOME 3.0 was released.

Having concrete data on how much fallback mode is used would be helpful. Fact is that it is really difficult to obtain that. I find it interesting in this light that you say "KDE hasn't recovered from KDE4". Where is the data to support this claim?

Standing at the sidelines and second guessing is cool, but bit of trust would be nice. I've been around GNOME since around the 1.4 days, same for many others involved with GNOME. Various developers returned with the plan to do GNOME 3.

Confused

Posted Nov 6, 2011 19:16 UTC (Sun) by kragilkragil2 (guest, #76172) [Link] (10 responses)

>Anyway, bringing up a kernel hacker and "lots of regressions" is not going to impress me much. Instead I rather look at GNOME 3.4 and see what we can improve. Just like GNOME 2.x.

That will probably happen, but my point is that resetting major parts of what your users come to exspect from a desktop is not going to win you lots of roses.

>This all was communicated before GNOME 3.0 was released.

I doubt that. See above.

>I find it interesting in this light that you say "KDE hasn't recovered from KDE4". Where is the data to support this claim?

Nearly every KDE4 post on any website that allows comments since 2007.
I was talking about marketing. FOSS marketing is done on the internet.

>Standing at the sidelines and second guessing is cool, but bit of trust would be nice.

Oh I trust that by the time Gnome 3.10 is released most complains will be gone, but I wish user facing components would see gentler migrations, even if that would mean maintaining older components a lot longer and not making progress in other parts that fast. The kernel is a good example IMO. Maybe doing it that way would encourage more people/companies to participate and therefore reduce the burden on devs.
Gnome could have learned a lot from KDE4, but didn't IMO.
But to be totally honest I wish Qt would have been LGPL from the start, maybe we would live in a very different world right now.

Confused

Posted Nov 6, 2011 19:53 UTC (Sun) by ovitters (guest, #27950) [Link] (2 responses)

Just because not everyone noticed, doesn't negate that it was communicated. Doubting if that occurred while I am from the team who decided on this and communicated this: oh well. That people will miss such communication is a different topic.

I don't get what you mean with marketing and hasn't recovered. I disagree that FOSS marketing is only done on the internet. Ubuntu is very well known, and it is not due to just internet marketing. You still haven't provided data on KDE marketing. I don't think it is possible to get such data, so we make the best decision. But keeping fallback around forever was never the plan, nor in line with the choices we've made over the last years. Our decisions are pretty predictable.

Comparing GNOME3, to KDE4 is inaccurate. We've been giving pretty much the same answers and promises for a long time. GNOME 3.0 was really stable, we didn't say "oh just a development release" just before releasing. This is mostly what I surprised me about KDE4 (bit buggy + change of message just before release). We have delayed GNOME3 various times to increase the quality.

Think getting lots of criticism on big changes is close to impossible to avoid. The more users something has, the more bugs are discovered. What is stable for 1000 people might be really buggy once 100.000 people use it. But you have to release at one point, because 6 additional months with the same amount of people (1000) will have the same benefit as just having 100x the amount of people using it.
The focus within 3.x is iterative design. Designers are trying things out and changing things which do not work. Meaning: overview mode works slightly different in 3.2, and might change a bit again in 3.4 or 3.6. Same for Contacts and Documents.

Regarding your clarification on what you meant with regressions: I agree that changing user expectations is going to cause frustrations. It is annoying to suddenly have things work in a totally different way. I find most of that criticism totally logical and expected it. I saw the same during 2.0 though, except people seemed to be a bit more patient at that time.

IMO people have really high expectations of what a project can do and that bar raises each time you try and reach it.

Confused

Posted Nov 22, 2011 2:59 UTC (Tue) by Zizzle (guest, #67739) [Link] (1 responses)

Haha, GNOME 3 was only "stable" because it didn't do anything.

Change the font size? No.
Taskbar? No.
Single click application lanuch? No.
Applets? No.

It was like a "high quality" Hello World in my eyes.

I think most people complained about KDE4 for the same reasons. The functionality they were used to was gone. Not because it was buggy.

When something is buggy at least you can be reasonably sure it will eventually stabilize. It's even pretty easy to fix crashers and bugs as someone unfamiliar with the project. Major features, not so much.

When functionality is missing and the devs are saying, "you don't need that, you were doing it wrong!", that is what really annoys people.

KDE4 was quite different

Posted Nov 22, 2011 7:36 UTC (Tue) by khim (subscriber, #9252) [Link]

I think most people complained about KDE4 for the same reasons. The functionality they were used to was gone. Not because it was buggy.

Nope. KDE 4.0 was as buggy as the come. It leaked memory, it crashed all the time. Limited functionality is was just an icing on the cake. GNOME3 is paradise in comparison: sure, some functions are missing, but at least the ones that are there work reliably.

This is chicken and egg problem

Posted Nov 6, 2011 19:57 UTC (Sun) by khim (subscriber, #9252) [Link] (6 responses)

Oh I trust that by the time Gnome 3.10 is released most complains will be gone, but I wish user facing components would see gentler migrations, even if that would mean maintaining older components a lot longer and not making progress in other parts that fast.

"Gentler migration" is expensive and userspace is much larger then kernel. This is chicken-and-egg problem: there are very few customers which are ready to pay for desktop thus it's hard to do things like "gentler migrations" and because of regular "rough migrations" it's hard to attract users who will pay for desktop. It's not clear how this circle can broken.

Gnome could have learned a lot from KDE4, but didn't IMO.

They tried - but looks like "fallback mode" was just a waste of resources: looks like people either use GNOME Shell or they drop GNOME completely.

This is chicken and egg problem

Posted Nov 7, 2011 7:43 UTC (Mon) by hein.zelle (guest, #33324) [Link] (4 responses)

(Apologies if these comments are unity-specific and don't apply to Gnome 3)

I haven't switched away, and since ubuntu 11.04 I've definitely started using the "classic" interface. I've tried unity for a week, and couldn't find half of the things I wanted to get working, not even with a lot of searching. I think if you want real data, a software-based poll or at least a widely announced poll on the web is required. Guessing either way is probably not adequate.

Personally, what I think is missing (for unity, I haven't tried gnome3) is documentation on how to use it. One of the major strong points of ubuntu was a well organized menu. In unity, I can't find anything. A "how to use unity effectively" tutorial would be a welcome addition, and could resolve a lot of the issues people are having. Having to search on the internet for hotkeys is not the right way, if you ask me.
Examples of things I could not find:

- how do I change the look/feel/behaviour? (answer: install a separate, non-default package. Found via google, not the menu or help)
- how do I make the icons smaller (1680 wide screen)?
- focus-follows-mouse and lazy-lock (can't do?)
- access to the original menu?
- access to the original system configuration tools?

I'll try again in a little while with ubuntu 11.10 to see if things have improved, but that list of problems is a bit too long for my taste. Add to that a couple of software bugs (image viewer won't go full screen, etc), and the fun is complete.

This is chicken and egg problem

Posted Nov 7, 2011 10:16 UTC (Mon) by jku (subscriber, #42379) [Link] (3 responses)

Getting real data is much, much more difficult than setting up a poll on the web. Think about it: Any amount of advertizement and announcements cannot remove the bias inherent in that method. A hundred people responding, 10000 people responding... They'll still have the same self-selection bias that cannot be fixed by increasing sample size.

This is chicken and egg problem

Posted Nov 8, 2011 9:00 UTC (Tue) by hein.zelle (guest, #33324) [Link] (2 responses)

I don't quite understand your point - I probably misunderstood. If you're trying to find out how many people are actually using the fallback mode of gnome, wouldn't an automatic software solution (like debian popcon) be much more objective than a web poll? Of course a web poll would be easier to implement.

Anyway, I agree that getting objective data is hard no matter which method you choose. It may be enough to get a feeling for how intensively the fallback mode is used, though.

This is chicken and egg problem

Posted Nov 8, 2011 16:15 UTC (Tue) by ovitters (guest, #27950) [Link]

Privacy issues results in inreliable data. Popcon is not used by everyone, and I don't think it measures this. I think we would violate privacy laws if we would add code in GNOME to send such data.

That said, if we'd have something used by many, it would be interesting to see.

This is chicken and egg problem

Posted Nov 9, 2011 8:27 UTC (Wed) by jku (subscriber, #42379) [Link]

I was just pointing out that one of the solutions you suggested, a web poll, is useless for the task at hand -- it just can not answer questions like "how many people use X" or even "what percentage of people use Y".

This is chicken and egg problem

Posted Nov 8, 2011 18:53 UTC (Tue) by vonbrand (guest, #4458) [Link]

I had Gnome on my netbook, and found it easier just to switch to XFCE while Gnome shell didn't work there. The fallback mode was too different from real Gnome 2 for my liking (or perhaps too similar, with just the right annoying minor differences).

Confused

Posted Nov 9, 2011 1:48 UTC (Wed) by brouhaha (guest, #1698) [Link]

I am aware of everything you think I am not.

One wonders how you have accomplished such a stupendous feat.

Confused

Posted Nov 5, 2011 14:53 UTC (Sat) by ovitters (guest, #27950) [Link]

I am talking about the when. Initially it was said that it would stay during the whole 3.x. Atm, might be more logical to do it earlier. But earlier was not decided, just suggested.

Confused

Posted Nov 4, 2011 16:19 UTC (Fri) by csigler (subscriber, #1224) [Link] (3 responses)

Is it common for software projects to respond to a loss of end users in this way, that is, by eliminating those features less well implemented which have caused users to switch to other software platforms, rather than improving them? Given that the users of GNOME aren't customers in the strictest sense -- there is no direct revenue stream -- this may be more common than I'm aware of.

Clemmitt

Confused

Posted Nov 4, 2011 16:26 UTC (Fri) by drag (guest, #31333) [Link] (1 responses)

If a project does not want to waste their time supporting features that are never going to make anybody happy, then why do it? As long as the software rendering works well enough for composition then there is now no technical need for fallback mode.

If fallback mode is never going make people happy that hate Gnome shell, then what is the point? Just waste man hours supporting yet another half-assed 'alternative'?

Confused

Posted Nov 16, 2011 21:46 UTC (Wed) by cas (guest, #52554) [Link]

If fallback mode is never going make people happy that hate Gnome shell, then what is the point? Just waste man hours supporting yet another half-assed 'alternative'?

The potential for fallback-mode to improve to the point where it *IS* an adequate replacement for gnome2 panel is the only reason I persist with Gnome 3 on my desktop machine at work (on my home desktops, I've stayed with Gnome 2, and was planning on doing so until fallback mode became good enough).

If the story has now changed, and fallback mode will vanish completely at some point, then there's no reason for me to persist. I may as well switch to XFCE or something, and do it sooner rather than later. Even though i don't want to.

Confused

Posted Nov 4, 2011 23:28 UTC (Fri) by daniels (subscriber, #16193) [Link]

It depends on whether or not their aim is to be everything to everyone or not.

Confused

Posted Nov 4, 2011 16:00 UTC (Fri) by sgordon (guest, #70042) [Link]

The precursor to that statement that was removed in the article was:

"That's really a policy decision for the GNOME / Fedora desktop teams,
not for ajax."

I wouldn't take it as a definitive statement of anything.

Confused

Posted Nov 4, 2011 16:11 UTC (Fri) by drag (guest, #31333) [Link] (14 responses)

There is no technical need for fallback mode anymore because they have eliminated the dependency for hardware acceleration.

The Linux/Mesa software OpenGL stack has improved to the point were it can provide the necessary APIs at (hopefully) decent performance on composited desktops.

Confused

Posted Nov 4, 2011 19:29 UTC (Fri) by clump (subscriber, #27801) [Link] (13 responses)

Fallback mode supports features the 'normal' mode does not. Static workspaces, for example. This is just one technical difference, and unfortunately the one most important to me.

Confused

Posted Nov 4, 2011 21:15 UTC (Fri) by luya (subscriber, #50741) [Link] (12 responses)

You can modify that behaviour through Javascripts Gnome 3 uses.

Confused

Posted Nov 4, 2011 21:27 UTC (Fri) by clump (subscriber, #27801) [Link] (11 responses)

I mainly use fallback-mode, but otherwise I already modify some configs to disable the hot corner. While I'm glad I can make that particular modification, it could get reverted by the package manager.

Confused

Posted Nov 4, 2011 21:43 UTC (Fri) by drago01 (subscriber, #50715) [Link] (10 responses)

Not if you implement it as an extension ... there is even one already doing that.

Confused

Posted Nov 4, 2011 21:59 UTC (Fri) by drag (guest, #31333) [Link] (9 responses)

Yes. Extensions are extensions, like firefox extensions. It's a scriptable window environment.

I expect that some scripts will need to be updated between major releases, but hopefully Gnome developers will be smart enough to keep required changes to a minimum.

Confused

Posted Nov 5, 2011 13:53 UTC (Sat) by clump (subscriber, #27801) [Link] (8 responses)

Extensions can do some of the things I like about fallback mode, though to my knowledge not all of the extensions I'd like exist or are properly maintained. I like fallback mode because: I can disable animations, static workspaces, no hot corners, alt+right-click modifies panels, etc.

Perhaps a decent comprimise would be to drop fallback mode, but allow more customisation without needing extensions?

Confused

Posted Nov 5, 2011 15:00 UTC (Sat) by ovitters (guest, #27950) [Link] (7 responses)

There is work ongoing to make an extension infrastructure.

Before various extensions were all imported to GNOME git and maintenance was taken over by one person. Giving a bus factor of one. The new infrastructure around it has as goal to ensure the extensions really can be used in practice. But then you need not only allow multiple extension developers, but also scale the review of them.. extensions can do anything. E.g. abuse the latest kernel root bug.

Confused

Posted Nov 5, 2011 16:53 UTC (Sat) by clump (subscriber, #27801) [Link] (6 responses)

I certainly respect the work being done with extensions, so I don't wish to diminish it. That said, I don't need extensions when running fallback mode.

Extensions are beautiful for shaping software far beyond its original design. Rather than shape the software, I simply wish to have more control. It would be reasonable to control certain defaults of GNOME Shell, if even only by command line tools, rather than require extensions.

Currently GNOME Shell and its fallback mode aren't equal because I can control more with the latter. Why keep fallback mode around if GNOME Shell can offer the same degree of control?

Does this idea sound reasonable?

Confused

Posted Nov 6, 2011 20:08 UTC (Sun) by ovitters (guest, #27950) [Link] (1 responses)

GNOME shell will never run applets. So technically it'll be different. You could probably do the same within an extension, but nobody within GNOME is focused on providing such an experience. We want to provide the ability to make this possible though.

So there is a reason for keeping panel+applets.

But I think we both agree. Seems logical to 'remove' fallback and extensions + gnome-tweak-tool/dconf is good for people who want more control. Problem is not so much with giving control, but should be clear it is a tweak. Once it is clear that we consider it a tweak, it avoids giving the impression that it is guaranteed always work. A tweak might of course break sometimes, eventhough there is no intention to have bugs in them.

Cannot think of a good example, but maybe networkmanager/pulseaudio. We rely on it, if you do not you're on your own (sort of, this is not the best example).

Note: remove fallback in practice that it just means it is still available, but not taken into account during development. So could become more buggy (e.g. on screen keyboard might only work in gnome-shell, not in gnome-panel). Not sure though how quickly it'll become buggy or unusable. I know that system settings sometimes does things differently (hacks) to ensure it works in both.

Confused

Posted Nov 10, 2011 22:00 UTC (Thu) by jmorris42 (guest, #2203) [Link]

> GNOME shell will never run applets.

Why? If XFCE's panel can import an applet why can't GNOME Shell? Is there some technical limitation, is it purely political or simply a lack of round 'tuits?

Confused

Posted Nov 7, 2011 0:29 UTC (Mon) by gmaxwell (guest, #30048) [Link] (3 responses)

> if even only by command line tools

If thats the way it works then the commands will be undocumented, buggy, and frequently broken by changes. You'd be better off without them in gnome best to convince your Linux distribution that gnome doesn't meet your needs and that you'd rather they not so tightly integrate it so that you could use something else.

Confused

Posted Nov 7, 2011 18:09 UTC (Mon) by tuna (guest, #44480) [Link] (2 responses)

I have seen no such proof that tweeks (such as week numbers) behave in a more buggy way than the rest of Gnome 3. Do you have any examples?

Confused

Posted Nov 8, 2011 16:17 UTC (Tue) by ovitters (guest, #27950) [Link] (1 responses)

Focus follows mouse is buggy. But that is since 3.0 and it was improved in 3.2. :)

Confused

Posted Nov 8, 2011 19:28 UTC (Tue) by tuna (guest, #44480) [Link]

Much is buggy in 3.0. Try suspending with two users logged in for example. Or wonder why screen gets locked when you suspend.

Lots of bugs are fixed in 3.2.