User: Password:
|
|
Subscribe / Log in / New account

Seeking Enlightenment (The H)

Seeking Enlightenment (The H)

Posted Nov 2, 2012 3:02 UTC (Fri) by bojan (subscriber, #14302)
In reply to: Seeking Enlightenment (The H) by ovitters
Parent article: Seeking Enlightenment (The H)

> the graphics card I bought 4+ years ago

Forget hardware. Try running mutter/llvmpipe in VNC. The CPU usage spikes are still beyond anything that one would find comfortable.

In fact, the whole Gnome Shell paradigm of "let's change every pixel on the screen every time something happens" is pretty bad for remote work.


(Log in to post comments)

Seeking Enlightenment (The H)

Posted Nov 2, 2012 3:57 UTC (Fri) by geofft (subscriber, #59789) [Link]

VNC seems to be not the right protocol there. I gather that SPICE has optimization for this exact use case, though I haven't tried it out myself yet.

Seeking Enlightenment (The H)

Posted Nov 2, 2012 4:54 UTC (Fri) by bojan (subscriber, #14302) [Link]

VNC is not my choice either. It's just that this is what works now and what infrastructure where I work supports. I cannot do my work 5 years from now.

Seeking Enlightenment (The H)

Posted Nov 2, 2012 8:45 UTC (Fri) by dsommers (subscriber, #55274) [Link]

Even though VNC performance isn't optimal, compared to SPICE, making SPICE "clever" to not to complete screen refreshes when not needed, isn't solving the issue on the right place.

Rather go to the root of the problem, fixing gnome-shell or whatever else responsible for these massive screen refreshes. This will make the overall performance better in the long run. "Fixing" SPICE (or even VNC for that matter) is just an ugly hack.

Seeking Enlightenment (The H)

Posted Nov 3, 2012 4:25 UTC (Sat) by bojan (subscriber, #14302) [Link]

> fixing gnome-shell or whatever else responsible for these massive screen refreshes

It's the overview. If you want to start an app, for instance, almost every single pixel on the screen will be changed because of the overview. And then, it will be changed back again to what it was before. What a waste of bandwidth (the higher the resolution and depth, the bigger the waste). And all for no good reason whatsoever.

Seeking Enlightenment (The H)

Posted Nov 3, 2012 4:34 UTC (Sat) by mjg59 (subscriber, #23239) [Link]

Define "good reason". There are great UI reasons for the behaviour of the activities menu, just as there are great resource usage reasons for not doing so. You've applied weightings to one and assumed that they're universal.

Seeking Enlightenment (The H)

Posted Nov 3, 2012 5:24 UTC (Sat) by bojan (subscriber, #14302) [Link]

Gnome design documentation talks about distraction avoidance nonsense. As if autohide panel didn't already exist...

Apart from that, what are other reasons? To throw a kitchen sink at users when they want to start an app? To reduce desktop visibility? To annoy with useless animations?

Seeking Enlightenment (The H)

Posted Nov 3, 2012 16:24 UTC (Sat) by lsl (subscriber, #86508) [Link]

When I want to start some GUI program with the Gnome Shell I press the Mod4 key (or Super/Windows/whatever key) and start typing its name. I think most users appreciate it that they are shown in launching what program pressing the return key would now result. You certainly can argue about whether the displaying of results has to be full screen but an "autohide panel" is definitely not a solution to that problem.

Seeking Enlightenment (The H)

Posted Nov 3, 2012 18:07 UTC (Sat) by robbe (subscriber, #16131) [Link]

The panel can function the same: popup when you press Mod4, offer search, etc. Just not take over the whole screen. But that's how Windows does it, so it must be bad, I guess.

Seeking Enlightenment (The H)

Posted Nov 4, 2012 12:20 UTC (Sun) by sorpigal (subscriber, #36106) [Link]

But we *must* consume the whole screen! The user might be using a tablet, or a phone, and there's simply no way we could detect that and do something different on regular monitors!

Seeking Enlightenment (The H)

Posted Nov 4, 2012 14:48 UTC (Sun) by Arker (guest, #14205) [Link]

The biggest issue on a phone isnt even that the screen is smaller, it's that typing is not possible. A phone user is hunting and pecking whether with a physical or on-screen keyboard, so they will be entering characters slowly and hoping to find at match without entering the entire string. SOME users on a PC will be doing the same, and the same response is appropriate in any case of a slow typer - use all that idle time to search and display possible matches.

But when you have fast input coming in you should NEVER be interrupting that to search and display possible hits - that is counterproductive on any platform. (Even on a phone if someone can somehow send the input fast enough the same thing would apply.) The computer in that case should simply keep listening to the input. If the user hits mod4-firefox-<enter> at a decent pace without pause, there doesnt need to be any sort of GUI update for it at all, the only response needed is to launch firefox.

Seeking Enlightenment (The H)

Posted Nov 4, 2012 15:00 UTC (Sun) by mjg59 (subscriber, #23239) [Link]

So I start typing something, pause, type another few characters, realise that what I was searching for is already on the screen, go to tap it and then you update the results underneath me and I launch the wrong application?

Seeking Enlightenment (The H)

Posted Nov 4, 2012 16:29 UTC (Sun) by mathstuf (subscriber, #69389) [Link]

I would say that once the UI has been shown it should be kept up-to-date with the typing at any speed.

Seeking Enlightenment (The H)

Posted Nov 4, 2012 20:31 UTC (Sun) by mjg59 (subscriber, #23239) [Link]

The context was in terms of avoiding unnecessary repaints because of the bandwidth consumed. If it's an unacceptable performance hit it's an unacceptable performance hit independent of whether the user pauses or not. It's fine, though - it turns out that Gnome doesn't wait for any drawing to be carried out before responding to keypresses. You're free to interact with it while it's redrawing things.

Seeking Enlightenment (The H)

Posted Nov 5, 2012 21:39 UTC (Mon) by Arker (guest, #14205) [Link]

The thing is, if you want to come up with some absolute standard of when a performance hit is too much, you have to make way too many assumptions to do so. A slow user on a fast system will need all the help he can get - while a fast user on a slow system is going to want to throttle you if the computer seizes up trying to do searches and display suggestions when he knew what he wanted. There are going to be a whole continuum of use cases and explicitly tuning for one spot on the spectrum detunes all the others.

Simply waiting for a break in the input before triggering (and thus not triggering at all when input is fast and complete) gives you a kind of autotune for all kinds of cases. I am not talking about a long pause, just a momentary hesitation. Even on a new computer a fast typist can sometimes get ahead of the display, and on a slow machine it isnt hard at all to do that. Once it's all caught up and hits an idle loop, you can worry about whether it might make sense to make a suggestion.

Seeking Enlightenment (The H)

Posted Nov 6, 2012 3:34 UTC (Tue) by mjg59 (subscriber, #23239) [Link]

What are you defining as a slow system? I haven't found any hardware that runs Gnome 3 that is unable to reasonably keep up with my typing.

Seeking Enlightenment (The H)

Posted Nov 6, 2012 18:44 UTC (Tue) by tuna (guest, #44480) [Link]

I like Gnome 3, but my crappy netbook (Acer Aspire 522, all AMD parts) is pretty slow. If you have a couple of applications running it can take more than five seconds after you write "firef" in the overview and the Firefox icon shows up.

My iMac was much faster, but that can't run recent Fedoras (https://bugzilla.redhat.com/show_bug.cgi?id=865031)

Seeking Enlightenment (The H)

Posted Nov 4, 2012 20:06 UTC (Sun) by Arker (guest, #14205) [Link]

Personally I am a curmudgeon and won't be truly happy unless I can flip a switch somewhere to make it actually wait for a TAB before searching or updating searches. That to my mind is truly the best answer.

But mathstuffs answer would probably please a lot more people, and it will even make me happy in comparison to the alternative.

Seeking Enlightenment (The H)

Posted Nov 6, 2012 7:52 UTC (Tue) by mgedmin (subscriber, #34497) [Link]

Use Alt-F2. Tab-completion the way to always worked in bash. (No double-Tab to show all matches, though.)

Seeking Enlightenment (The H)

Posted Nov 14, 2012 14:52 UTC (Wed) by Lennie (guest, #49641) [Link]

You are kidding, right ?

You can't make a system which works slightly different on a desktop than on a phone ?

Seeking Enlightenment (The H)

Posted Nov 14, 2012 14:58 UTC (Wed) by mpr22 (subscriber, #60784) [Link]

And here we see proof that no matter how blatantly obvious a piece of sarcasm is, there will always be someone who botches their Detect Sarcasm roll :)

Seeking Enlightenment (The H)

Posted Nov 14, 2012 16:45 UTC (Wed) by Lennie (guest, #49641) [Link]

That is what slight sleep deprivation can do to you at times.

Seeking Enlightenment (The H)

Posted Nov 4, 2012 2:27 UTC (Sun) by Arker (guest, #14205) [Link]

I disagree. This is something that Apple figured out years ago (and subsequently forgot, the company has definitely reached senility at this point, but I digress.) You should definitely not be getting visual feedback if you hit a control button and start typing - at that point the interface should be concentrating on accepting your input not on talking back. WHEN and IF you pause, hesitate for a moment, THEN it's appropriate to think about bringing up a visual hint. Bringing it up every time is NOT helpful it's annoying as all fsck, the result is a very annoying system that effectively is always talking instead of listening. Bad computer!

Seeking Enlightenment (The H)

Posted Nov 3, 2012 8:41 UTC (Sat) by epa (subscriber, #39769) [Link]

Going to the root of the problem would involve running the desktop manager locally and only the particular applications that need remoting will be sent over the wire.

Seeking Enlightenment (The H)

Posted Nov 4, 2012 22:38 UTC (Sun) by ovitters (subscriber, #27950) [Link]

Rather go to the root of the problem, fixing gnome-shell or whatever else responsible for these massive screen refreshes.

I like the animation; it is pretty minimal. Obviously when stuff is remote you have to ensure that it is efficient. But I wouldn't want all the limitations and trade-offs that you need for something like that applied to when I have a local connection

When it is remote, it needs to be responsive / quick. You immediately suggest not doing screen refreshes. But that is a pretty easy view because your focus is that animation. What if you have some application that is also graphically intense?

It seems better to determine on a case by case basis what the best solution is. For the animation, maybe making gnome-shell efficient when used during a remote desktop viewing (ideally anything, VNC, SPICE, remote X, ...). But making e.g. SPICE more efficient would also help for some applications.

I like things that automatically do the right thing, eventhough that might take a while to get right. During which a visible 'knob' might be seen as a better solution. I still love that you don't need an xorg.conf, etc.

Seeking Enlightenment (The H)

Posted Nov 5, 2012 22:47 UTC (Mon) by bojan (subscriber, #14302) [Link]

> You immediately suggest not doing screen refreshes.

Unnecessary screen refreshes should not be done. Nobody needs a dash, a workspace switcher and expose when they just want to start an app.

Seeking Enlightenment (The H)

Posted Nov 5, 2012 22:52 UTC (Mon) by ovitters (subscriber, #27950) [Link]

I sometimes use overview mode, sometimes ALT-F2.

The different screen is to avoid any distractions. It focusses you on what you're currently doing (current application). I understand that for you the reasoning behind the way GNOME works is not what you like. But that is why there are multiple desktops and also why there are extensions in GNOME.

Seeking Enlightenment (The H)

Posted Nov 2, 2012 6:31 UTC (Fri) by idupree (guest, #71169) [Link]

I experienced this locally while I (my user) accidentally didn't have video-card privileges. Using llvmpipe, which I eventually realized, gnome-shell would frequently use more than half of my quad core Sandy Bridge i7 CPUs. For reference, my screen is 1920x1080 (about twice as many pixels as 1366x768). That was enough CPU to make things a bit laggy and make me debug the problem. I'm kind of impressed that Linux was silently so resilient to that permissions problem that I didn't even know anything was wrong at first. How do systems with weaker CPUs manage?

Seeking Enlightenment (The H)

Posted Nov 2, 2012 16:48 UTC (Fri) by drago01 (subscriber, #50715) [Link]

> In fact, the whole Gnome Shell paradigm of "let's change every pixel on the screen every time something happens" is pretty bad for remote work.

No idea where you got this from but this is just wrongs. It only changes the pixels that actually changed (i.e tracks damage events) and copies the changed region from the back to front buffer.

Once we have GLX_EXT_buffer_age (hopefully soon for 3.8) it will be able to do that without copying anything.

Seeking Enlightenment (The H)

Posted Nov 2, 2012 17:06 UTC (Fri) by paulj (subscriber, #341) [Link]

It's all well and good tracking and updating only the regions that change, but when common operations like opening a menu are implemented with effects that change the entire the screen, then... (which I think was bojan's point ;)).

Seeking Enlightenment (The H)

Posted Nov 3, 2012 4:17 UTC (Sat) by bojan (subscriber, #14302) [Link]

Exactly. That is the problem of the overview - it is unnecessary. Just because phones do it, doesn't mean my desktop should too.

Seeking Enlightenment (The H)

Posted Nov 5, 2012 4:00 UTC (Mon) by liam (subscriber, #84133) [Link]

Overview could be very useful. For instance, grouping windows into task groups a la Firefox's Panorama, or search results which let you explore your file resources in quick, rich ways (say, typing in "background-repeat" gives you a list of its properties and clicking on it would then open up devhelp, or whatever tool can display the docs). Mobile devices, to my knowledge, don't do these types of things b/c they really aren't designed for serious multitasking.
If you're happy with IceWM, or the like, then GS isn't for you.

Seeking Enlightenment (The H)

Posted Nov 5, 2012 22:46 UTC (Mon) by bojan (subscriber, #14302) [Link]

Overview is a kitchen sink approach. It is entered for _everything_, including trivial things like starting an app. As I said before, unnecessary.

Seeking Enlightenment (The H)

Posted Nov 5, 2012 22:48 UTC (Mon) by ovitters (subscriber, #27950) [Link]

Your dislike of GNOME is understood.

Seeking Enlightenment (The H)

Posted Nov 5, 2012 23:17 UTC (Mon) by bojan (subscriber, #14302) [Link]

Now you are trivialising. It is not a dislike (which would suggest something subjective).

Objectively, when a user wants to start an app, dash, expose, workspace switcher (placed too far to be useful) and animations are not necessary. Also, changing every single pixel on the screen (twice) is not necessary either.

Seeking Enlightenment (The H)

Posted Nov 6, 2012 0:15 UTC (Tue) by ovitters (subscriber, #27950) [Link]

An interface is subjective, not objective. To you the animation is unneeded. But in fact the animations give usability hints. E.g. if you put your mouse in the corner it highlights via the animations that the mouse triggered a hot corner.

Saying such an animation is objectively not needed: it was done as a usability hint. Basically to guide the user. I don't see how you can state that such a hint is not at least in part subjective.

Technically you are right: animations are not needed. But a user interface is more than implementing technology.

Seeking Enlightenment (The H)

Posted Nov 6, 2012 0:29 UTC (Tue) by dlang (subscriber, #313) [Link]

He was not saying that animations are bad, that is subjective

he was saying that animations are not needed if the user already knows what to do without them.

That's a very different statement.

If the user needs the hint, they can be good. If the user doesn't need the hint, it's hard to see what value they provide at all.

The suggestion that you hold off on the animation until the user gives some indication that they need help (by not doing anything for a short time) is a pretty good one.

Seeking Enlightenment (The H)

Posted Nov 6, 2012 2:30 UTC (Tue) by bojan (subscriber, #14302) [Link]

> An interface is subjective, not objective.

So, I take it this is all subjective nonsense as well, right?

http://developer.gnome.org/hig-book/

Seriously for a second, take my mobile phone (Samsung Galaxy S2), for instance. If you wanted to display a list of applications in its own dialog or menu over the running app, you could probably do that, but each of the buttons/icons would probably be too small to press. So, this is done in full screen on that device. This is a purely functional decision and is not subjective at all.

A different example. Fast forward to Galaxy Note 10.1 and you'll find multi window mode. Two apps, one next to the other. Enough space to do both, especially when using the pen. Something like this would probably not work on my SGS2.

And then there is the desktop, which uses mouse/glide point as a pointing device (very precise) and has a much, much larger screen. And yet, Gnome insists to treat this as a tablet or a phone by overlaying kitchen sink over my work every time I decide to do even the most trivial "something else". It changes my view (twice) in the process, forcing VNC to repaint large areas. It moves my windows around (expose), although I never wanted to do this. These are nonsensical usability regressions for a _desktop_ UI.

PS. If you wanted to see how a pretty good UI for one device is terrible on another, download Profimail for Android. I used Profimail on Nokia's Symbian for a long time and it was pretty much the best IMAP capable mail client on that platform. On an Android phone with no keyboard, it is pretty clumsy. The UI is all wrong.

Seeking Enlightenment (The H)

Posted Nov 6, 2012 14:47 UTC (Tue) by pboddie (guest, #50784) [Link]

The one area where I can understand people wanting to employ large areas of the screen is precisely that of the applications or "start" menu, mostly because when you navigate inside that you aren't usually wanting to look in there at the same time as look at something else, although there are obviously exceptions. For example, you might want to drag something out of such a menu onto the desktop to make a shortcut to an application. That would be difficult if the menu were full-screen. (It also appears to be difficult when Nautilus doesn't like the drag operation, but anyway.)

Here, GNOME 2.x and KDE 4.x are suboptimal just as GNOME 3 and Unity are presumably suboptimal. Indeed, as I have publicly noted previously, KDE 4.x has an annoying iPod-like start menu that seems to confuse people, all to be able to squeeze a menu into a ridiculously small portion of the screen. Clearly, that isn't a good solution, either.

Seeking Enlightenment (The H)

Posted Nov 6, 2012 15:33 UTC (Tue) by ovitters (subscriber, #27950) [Link]

I did not say it was nonsense. I said it is not objective.

For instance, you bring up something on a mobile phone. There the text would be too small if not full screen. Something objective would state such criteria exactly and that criteria would be based on something subjective.

Obviously defining that e.g. a 4pt font is unreadable is quite easy. But when does it become readable? That is subjective. You can define e.g. 10pt and measure such a rule objectively, but rule itself ("10pt") was still based on something subjective.

I've read reviews about (IIRC) the Galaxy Tab, which is 5" or something. It also allows multiple windows (or something). Some reviews hate it (too small), some love it.

Note that GNOME is not very usable on a tablet. I suggest trying it before suggesting that it is a tablet interface.

Seeking Enlightenment (The H)

Posted Nov 6, 2012 20:54 UTC (Tue) by bojan (subscriber, #14302) [Link]

> I did not say it was nonsense. I said it is not objective.

Wow! If it is subjective, it is automatically nonsense. Otherwise, why should anyone follow design guidelines that someone found nice for himself only?

> You can define e.g. 10pt and measure such a rule objectively, but rule itself ("10pt") was still based on something subjective.

Surely, you have to be kidding. You obviously do not think that it is possible to measure at which font size majority of people can easily read the text, at which size majority of people can comfortably hit a button with their finger etc.

> Note that GNOME is not very usable on a tablet. I suggest trying it before suggesting that it is a tablet interface.

I did not claim Gnome was usable on a tablet. Only that Gnome developers insist on making my desktop behave like a tablet/phone. Which is unnecessary, because my desktop is neither of those.

PS. You have just reduced decades of usability research to nothing more than someone's whim.

Seeking Enlightenment (The H)

Posted Nov 6, 2012 22:18 UTC (Tue) by ovitters (subscriber, #27950) [Link]

Wow! If it is subjective, it is automatically nonsense.

In the definition I looked up, it states for nonsense: "Subject matter, behavior, or language that is foolish or absurd.". I was not claiming that what you said is foolish or absurd, I just don't agree with you.

Surely, you have to be kidding. You obviously do not think that it is possible to measure at which font size majority of people can easily read the text, at which size majority of people can comfortably hit a button with their finger etc.

You claimed you could measure similar things objectively.

I did not claim Gnome was usable on a tablet. Only that Gnome developers insist on making my desktop behave like a tablet/phone

So GNOME is not usable on a tablet. I'm a GNOME developer, and I don't insist on making your desktop behave like a tablet/phone. I don't really care what you use on your desktop actually.

I'd like to see some reference of those developers. Actually, a few designer would be more appropriate.

PS. You have just reduced decades of usability research to nothing more than someone's whim.
No need to get personal or to claim you're an expert on usability research. Seems like two different fallacy arguments.

Seeking Enlightenment (The H)

Posted Nov 6, 2012 23:34 UTC (Tue) by bojan (subscriber, #14302) [Link]

> You claimed you could measure similar things objectively.

> No need to get personal or to claim you're an expert on usability research. Seems like two different fallacy arguments.

I did not get personal. I was arguing that by your reasoning (i.e. that usability issues are purely subjective), you have reduced objective, scientific research to someone's whim.

I did not claim I could measure such things. Usability experts claim they can.

> I'm a GNOME developer, and I don't insist on making your desktop behave like a tablet/phone. I don't really care what you use on your desktop actually.

I think this just about sums up the attitude many feel is coming form Gnome developers' camp. I have been using Gnome for many years and helped with fixing of many bugs. According to you, for this I should just get scorn.

I'm not that easily dissuaded.

Seeking Enlightenment (The H)

Posted Nov 6, 2012 0:09 UTC (Tue) by liam (subscriber, #84133) [Link]

I did say "could".
I don't think the way it is presently implemented is either very user-friendly or particularly flexible but I think experimentation with modern-ish graphics are important in order to develop better UIs.


Copyright © 2017, Eklektix, Inc.
Comments and public postings are copyrighted by their creators.
Linux is a registered trademark of Linus Torvalds