Hacker Friendly Ranting

After GNOME 3.0 was released, I gave in and decided to take the plunge. I packed up my .xinitrc and .xmonad/ and decided to give it a fair shake. I used it for about a week of real work, as well as the usual recreational activities (surfing, coding, wine).

The Good

Overall I was impressed at how effective I became with GNOME 3.0, although it’s a far cry from Xmonad (as I’ve mentioned in the past, but I’ll try not to get into tiling evangelism here). I’m not as keyboard-centric as I’d like to be, so already I have my hands on the mouse more than necessary. Getting back into the drag and drop paradigm wasn’t hard. I will admit I used the drag-to-side half vertical tile functionality built into mutter a *lot* and, for the most part, I was content with it. Vertical tiles are by far the most common layout I use, and even though it was restricted to a 50/50 split (and sometimes I like a 70/30), it was nice.

The gnome-shell bar took a little getting used to (it’s hard not looking at the top right corner for a time), but the status icons contained in it were very sleek and their menu options were very useful. These little icons were one of my favorite changes in 3.0. Particularly on my laptop where the network manager icon is very useful.

As I’ve mentioned before, I love the keyboard friendly additions. Like the built-in run dialog with completion, and the ability to type two letters of a program name from the activities view and see it pop up in the icons. I like that even more than I like my Xmonad dmenu, which is saying something. I haven’t figured out whether you can easily address the “favorites” on the left hand sidebar, but the new use case is very nice. Also, I was able to make the transition a bit easier from Xmonad by binding some keys from the keyboard controls (like Alt+F2 to open a terminal, Alt+F3 to open the run dialog, Alt+1,2,3 to switch desktops). These worked well, although I would have liked the ability to Alt+Shift+1 to send a window to a desktop.

Everything with my usual work went fine (with some caveats in the next section), surfing was no problem. Multimedia worked just fine, sound with pulseaudio was no issue (although I’ve been using pulse for awhile even outside of GNOME). I was pleased that running things like smplayer fullscreen on my TV was handled just fine and didn’t interfere with working on the other monitor. I guess problems with using 3D and video at the same time have been worked out long ago, but I was concerned with the new desktop. My only qualm with that was that when someone switched to the activity view on the main screen, the full screen movie was shrunken a bit and enumerated like all the others (so it’s hard to use the main desktop without jarring the people watching the movie on the other screen. At least it’s consistent though and, because the movie was still playing in the slightly smaller window and with sound this is only a minor nitpick.

My wife actually really likes the new interface. She’s running it on her netbook and also on our “family” machine (RAID box), which is where I did the movie testing. GNOME is more her style of usage, being the type that uses a web browser and the media players most often. I was pleased that her little netbook had no trouble with the graphics either (it’s by far the lowest spec machine we have, unless you count phones).

Lastly, our Wine Diablo 2 sessions went very well. I found that if we run in a desktop, GNOME properly treats it like any other window, and it smart enough to full screen it on restore and hide its window bar. I attempted to run in unmanaged, full screen mode and it works. The bar was still shown for some reason, but that appears to have been fixed in the interim as I can’t reproduce now. The one issue that I had was that running unmanaged, if I accidentally hit the windows key, I had no idea how to get back to the Diablo 2 screen. If I hit windows key again, it would automatically take me back, but if I had focused another window it was a mystery.

All in all, I was happy with what amounts to the initial rough cut of the 3.0 interface. I’m really looking forward to seeing where it goes. I think that if I’m ever able to finish up my current side project, I might start taking a look at contributing some improvements. However, for now, I have some definite WTFs.

The Bad

Multiple Monitor Desktops

GNOME 3.0 has pretty decent support for taking advantage of more than one screen. As I mentioned above, playing a movie on one and working on the other works fine. The main problem is that, by default, your secondary screen can only have one desktop on it. Now this might be a sane default (i.e. switch tasks on the primary screen, keep static things like IMs / email / multimedia on the second screen), but it’s not how I want to use my monitors.

There is a gconf setting to allow the desktop to basically span both screens (so switching to desktop 2 changes both monitors) but after I enabled it, support was buggy. Things like closing empathy would cause windows to warp onto my primary desktop for no apparent reason. I guess there’s a reason this is hidden in gconf. Although I have no clue why it would still be a gconf key, as I though they were switching to dconf, but I could be wrong.

Anyway, ideally I’d like to see Xmonad style multiple monitor support. I.e. you have one desktop apiece and can change them independently. Xmonad enumerates them all and allows you to show any two, but I’d settle for having a per-screen desktop stack since that seems like it would fit the GNOME 3.0 paradigm.

Empathy / Integrated Messaging

I’ll admit, I was excited about the integrated messaging. It seems so useful and one of my nits about XMonad is that I always have to switch back to my IM windows, or send them to the desktop I’m working on.

This feature is entirely broken on release though. Empathy is just not ready, and neither was the integration. For example, you get a nice notification on a new message (or, configurably, on contacts going online / offline). You can even click that message and respond inline. Cool! But what you can’t do is focus another window and keep that inline chat window open. I want to hold a conversation while surfing Reddit with the inline window hovering on my desktop. I like to be able to read the last four or five lines of my conversation, not just see the next response show up for an instant in a notification and decide whether I want to respond.

In addition, the interaction between the integrated bar and the “traditional” empathy chat is absolutely bizarre. For example, if you have a real empathy chat window open (which you almost always will because of the inability to keep the integrated window open), and you receive a new message, you get a notification and a little chat blurb on your notification bar, but no new chat window in the actual empathy window. Some sort of foibling is required to get it to open there, and I’m not quite sure what.

The “traditional” empathy contact list is also underfeatured, like not being able to change the order of contact groups. It also doesn’t remember your status setting on return (just setting you to “Available”, whereas pidgin will automatically reuse that last status). Finally, because it supports some form of LAN messaging that never seems to connect the contact interface is constantly spinning as if it’s loading.

The last thing, and the true deal breaker for me using GNOME 3.0 is that the notification bar fails to show you urgent flashing notifications. I walked out of the room for a moment and missed a message. I had *no idea* I had missed anything until I randomly displayed the notification bar and say the empathy icon flashing and a two-hour-old message from my boss asking for an email. That doesn’t work ladies and gentlemen. I can’t take two hours to respond to a message. That’s ridiculous. Especially since I was sitting at my computer the whole time. Hell, I’ve even missed notifications because I was looking at my other screen.

I thought GNOME 3.0 was designed specifically for this workflow. I guess I was wrong.

Conclusions

Nobody argues with the fact that the GNOME 3.0 interface needs work. I was pleased with using it for awhile, but it’s sorely in need of improvement with a handful of things. Especially the chat and notification features, which are effectively broken.

The upshot though is that nothing is so wrong with the concept or the execution that these bugs couldn’t be fixed very quickly. These are not problems that should require heaps and heaps of code to fix. In fact, I wouldn’t be surprised if distros officially packaging GNOME 3 have already addressed and patched some of them.

NOTE: All of the above judgments were made using the current (as of 19 April) Arch GNOME 3.0 packages from gnome-unstable and testing. Particularly the gnome, gnome-extra, and telepathy meta-packages. Arch is great, but I’m assuming these packages are basically packages as-is, with only minimal distro integration patches.

EDIT: Suddenly my Evince 3.0.0 (according to Arch 3.0.0-3) has started naming bookmarks with the current / first new section on the page. Kudos to whoever fixed that, although the help still claims there’s no such thing as a bookmark =). 7 May 2011

Let it be known that I’m a bit of a PDF power user. Not because I wanted to be, but because my job suddenly started entailing reading 1000 page specs that are held together with nothing but blinding seas of jargon. Being that they’re usually new CPU user manuals and SoC programming references, they’re also extremely useful… when you can find the information that you need.

So, it was with open arms that I embraced Evince’s new 3.0 feature: bookmarks. It’s only been recently that I’ve started to need this feature, but a quick Google indicates people have been WTFing over Evince not supporting bookmarks since it was created. The F1 help (which as of this writing on Arch Linux still says that bookmarking isn’t supported) notes that you can use annotations as bookmarks, but it seems to me that they’re fundamentally different things and, even if they weren’t, the PDFs I’m reading seem to have hundreds of blank annotations. I’m not sure if that’s Evince’s fault (not all PDFs have it), if the annotations are real, or if the person generating the PDFs is just using a crappy PDF generator.

After installing Evince 3.0, I decided to give bookmarks a spin. I scrolled to a useful page in one of my specs (a page describing the minutiae of some bits in a hardware table that I’ve been using recently). Seeing the new “Bookmarks” menu, I simply clicked “Add Bookmark” (Ctrl+D would’ve sufficed as well). Then, clicking the same menu, I got a bookmark listed: “Page 250″. Right clicking it had no abilities, there was no apparent way to edit the bookmark. Gah. “Page 250″ is almost worthless as a bookmark, especially since I had like 10-15 other bookmarks I wanted to make. Bookmarks shouldn’t force a user to shuffle through them all just to figure out what they are.

But wait!

But I didn’t give up there. The “Bookmarks” menu makes it totally non-obvious, but in the Evince sidepane (hit F9 or View->Sidepane) you can now select a “Bookmarks” selection from the pull down. From that list of bookmarks, you can right click and actually rename / remove your bookmarks.

After renaming, they then show up properly in the “Bookmarks” menu.

For shits I decided to try it out on some other, non-work PDFs from random places on the internet. No dice for the bookmarks. Not that it tells you why it can’t create bookmarks, but from the look of it, Evince can only bookmark PDFs that support embedding them. I can’t be too disappointed because it doesn’t look like FoxIt or Acroread can do it either. Both of them seem to think “bookmarks” are chapter and section headings which is stupid. Nonetheless, I was hoping for a [gd]conf / ~/.config/evince solution that would work generically with all PDFs and versions.

So, in the end, the new feature is nice … after you learn to use the sidepane for it, and only if the PDF you’re using happens to support it.

For reference, the failing PDFs were (for some reason) all PDF 1.4, optimized, with no security and the work PDFs were 1.6, optimized, no security. I’m guessing that it comes down to the PDF version, but I could be wrong.

My previous post criticized modern desktop environments (like gnome-shell and KDE 4.6) because they lack anything resembling efficient window management. It’s not a surprising view coming from someone that’s been spoiled by Xmonad’s beautifully easy mouse-free window management and dynamic layouts.

I’m not going to talk about window management today though. I installed a fresh copy of the libraries from Arch’s gnome-unstable repository and, instead of giving gnome-shell another whirl (not yet at least), I merely restarted Xmonad with the usual suspects running in the background (gnome-settings-daemon, gnome-keyring-daemon) and continued to use my typical GTK apps.

I have to say that I was pleasantly surprised that the GNOME beta functioned well as a drop-in replacement and, aside from the different theme and widget style everything worked identically. The VNC client still remembered my passwords too, so I interpret that as the keyring being intact. I even like the new default font and anti-aliasing choices. I’m sure some people will complain that it looks muddy (as critics of anti-aliasing almost always do), but to each their own.

Overall I’ve been pleased with some of the non-shell improvements of GNOME 3. The control center is a good replacement for the system menu of traditional GNOME and on par with Windows and KDE. I probably appreciate it a little more than average considering that I don’t have the system menu under Xmonad and it’s easier to remember to invoke gnome-control-center than it is to remember which little GNOME subapp to use to tweak some behavior of the GNOME applications.

One thing that surprised me was dconf-editor. I’ve heard throughout the GNOME 3 bashing blogs, that a lot of the configuration was hidden in dconf-editor and that really sucks because it’s basically regedit. That may be true, but GNOME has always essentially had a registry (gconf) and dconf-editor has to be the nicest registry editor I’ve ever used. Not that it’s a passable replacement for a control center option, but as a developer and desktop explorer it’s nice to see a full on description of each possible toggle and tweak in the registry.

I understand that this is a very small subset of GNOME that I’m using, but it’s nice to know that even if the GNOME desktop’s next full release is the Linux desktop apocalypse, everything will still just work.

No doubt you’ve all been keeping (or have been kept depending on how you read the news) up on the GNOME 3.0 / gnome-shell / Unity stuff that’s been going on. In short, the new major version of the typical GNOME desktop, 3.0, has reinvented the way that you interact with your desktop.

Initially I was excited to hear the news because the new release is supposed to be much more keyboard friendly than “classic” GNOME, and I am all about efficiently using the keyboard. That said, I am wary of any sort of broad scale change like this because I haven’t been fully convinced this is a good idea. Gnome-shell takes some good ideas, like the concept of per-task workspaces and the aforementioned keyboard friendliness, but it falls short. I’ve been trying to compose this entry for awhile since I compiled the gnome-shell ecosystem and even tried the gnome3.org liveCD, but it wasn’t until I read this post about eliminating minimize/maximize that it finally struck me. What the mainstream DEs are missing now is decent window management.

I’ve been an Xmonad user for years now and that one fact has really kept me away from using GNOME or KDE (in it’s vanilla form) for about as long. It’s just unwieldy to move windows with the mouse and painstakingly arrange windows to be productive. It’s this sort of task that software should make simple. For the uninitiated, Xmonad arranges all windows on a workspace in a dynamic layout. If you have one window, it takes up the whole screen. Two, and it’s divided between the two, three and it’s divided between the three and so on. The layouts are very advanced, ranging from simple splits to spirals, to circles. Sending windows to another workspace, switching workspaces, switching layouts, all of these can be accomplished with simple keystrokes. I don’t know how I’d survive without it anymore, I just have too many active terminals and applications to juggle with a mouse.

To be clear, I’m not advocating that GNOME start using Xmonad (although Xmonad integrates beautifully with GNOME 2.x). Also, I am aware the latest gnome-shell builds support a rudimentary tiling by dragging a window to the left or right edge of the screen, but it’s just not enough. KWin (KDE’s window manager) recently added basic tiling support as well but its efforts are just as primitive and laughable.

If the GNOME devs want to improve the way we interact with our windows, like gnome-shell purports to do, we need way more control over them both automatically (like dynamic layouts) and with user interaction. Basic concepts like “maximize” and “minimize” cease to make sense when the user has so much more power over how windows are shown by default. Want to work with your windows like a stack of papers (all fullscreen)? There’s a layout for that. Want to have a desktop with a browser and a news reader split 70/30? There’s a layout for that. Floating desktop for something like GIMP? There’s a layout for that. Decide the current layout no longer works for your workflow? Change it and the windows are automatically rearranged. Any conceivable, efficient way to view your windows can be chosen and dynamically used by a window manager. Using layouts makes dragging windows with a mouse feel like stone age technology or a child’s puzzle.

Additionally, the technology is entirely backwards compatible. If developers are worried about scaring new users with their layouts (although since gnome-shell is already a radical departure that doesn’t make too much sense) then just make the default layout behave like the current default (floating). Meanwhile, the rest of us looking for a better way can benefit. Perhaps best of all for the GNOME community, the people looking for that level of efficiency are also usually developers.

In short, there needs to be a powerful set of keybinds with a powerful set of window actions (send to desktop n, switch to desktop n, change layout, give window layout focus, etc.). If a major DE like GNOME got behind the tiling/dynamic layout paradigm, that would truly be a leap forward. Until then gnome-shell will be, at best, a marginal improvement over 2.x or, at worst, a KDE 4 style debacle. For now, all of us Linux users have to choose between a desktop with A+ integration (like GNOME 2.x — although in reality it’s probably more in the B-/B range) and a desktop that gets out of your way and lets you manage your tasks efficiently.