The first night we met up with Steven Parrish and Rex Dieter for some dinner at an Irish pub near the hotel. We heard that there was live music there and indeed there was. Very loud and right next to our table. By the time we had eaten and returned to the hotel, lots of people were gathered in the lobby and were discussing various things. Rex and I talked with David Malcolm about what was happening with Python 3 in Fedora 13 and how it would affect the KDE side of things (more on that below). Seth Vidal also joined up and the topic moved between bdb and sqlite and how it affects rpm and yum, patches to python and rpm itself and other tangents that were all too easy to get off on.

After getting to sleep around 2am or so, we woke up for breakfast to get to the college at 9:30. The bus ride cost was a little crazy ($3.50 for one ride) but there were few viable alternatives. We all gathered in the room deemed to be Room 1 to discuss how things worked at FUDCon with the bar camp style of scheduling. Lots of people lined up to pitch talks for the day including Rob to pitch CHASM (something else which will be explained in more detail below). Everyone then lined up outside in the hall where the pitched topics were taped to the windows to mark which talks were the most interesting to people. CHASM got 22 and we got paired with Matt Domsch’s talk about Mirror Manager.

The first talk that I attended was Steven Parrish’s talk on reporting bugs effectively. I was the one who asked if anyone was logging the talk and in the process volunteered myself for it. The hardest part about logging was keeping up with the slides and the talk and typing things into the IRC channel fast enough but I got better throughout the day and it got a lot smoother. The logging of the talks in IRC and using zodbot to throw the results onto the wiki is a great way to help those who are not able to attend FUDCon to be able to participate as well. The talk itself was great. Even though I get CC’d on lots of bugs due to being a co-maintainer of the entire KDE stack and having to request logs and such for different things. Lunch was next and Rob and I talked with Adam Williamson about netbooks, computer specs and whatnot over the provided meals.

The next session was on MirrorManager and CHASM (a project Rob and I have been working on). Matt Domsch discussed what MirrorManager is and what it offers for Fedora. I never knew that it picked mirrors for me. Goodbye fastest-mirror plugin hackery (manually remove the results for any mirrors I don’t want). After that, Rob and I (well, mostly Rob) presented on CHASM, the Cryptographic-Hash-Algorithm-Secured Mirroring solution. It aims to ease mirroring large collections of software that change over time (such as distributions’ repositories). It was well received and we got good feedback from Seth Vidal and Matt. Some ideas that we had about what actually goes on with the mirroring of Fedora and what CHASM needs to do were updated and we have rethought some of the assumptions we had before FUDCon.

After that, I attended the GDB talk and learned how to make pretty-printers to get usable output when displaying values of the running program. There were also tips on debugging threaded applications that I wasn’t aware of before and should help with debugging some KDE applications that make use of threading.

The most interesting talk (other than CHASM) that I attended was Jesse Keating’s talk on how Fedora will be ditching dist-cvs for git. I am very excited about some of the ideas that were being discussed such as the automatic patch management (just some macros in the actual spec file), exploding source tarballs into a git repository to create patches with git format-patch, actual tools rather than hackish make targets, and actual support for branching (I’m looking at it being able to do snapshots of upstream outside of rawhide so that if things go really awry with release times, an epoch isn’t needed to fix things up when rawhide gets branched for a new release). Between this and KDE’s conversion to git, there should be plenty of documentation on how to convert repositories of huge scale and hackery over to git.

The last presentation I attended was on trademarks and how they affect Fedora. It was a great discussion between how lawyers see trademarks, how hackers see trademarks, and what needs to be done to make things clear on both sides of the discussion. After learning about the basics of the legal web of things that affect Free Software in general (copyrights, trademarks, patents, etc.) in the Open Source Software Practices class here, getting some input on how lawyers look at things was a great insight.

After the BarCamp was over (great format by the way), everyone headed back to the hotel to drop off laptops and such before heading over to Dave & Busters for FUDPub where I talked with Ian Weller, J5, Brennan, David Malcolm, Luke Macken, and others (many of whose names escape me) about various things related to Fedora, Python, and other technical topics. Towards the end, J5, Brennan, and I went to the game room to roam around and see what was going on (I watched since I prefer watching games than playing them). We ran into Spot, Jesse Keating, and others as well. Spot had a bunch of tickets and everyone donated what they had which, with a total of 670 or so tickets, got Spot a set of poker chips.

After FUDPub, I headed to the hackroom on the third floor and was told by Seth Vidal that if Rob and I can’t sell CHASM to John Hawley, the system administrator of kernel.org, we need to rethink everything. After hanging around and listening to some of the discussion, I headed to bed so that I could get up in the morning for the hackfests the next day. Rob says he lost the pool game, but I sunk the 8 ball in the wrong pocket, so to be pedantic, he won .

The next morning, after waking up, eating breakfast, and talking with…well I can’t remember exactly who I talked to, so many people. I’m sure it was interesting and productive . After that and the bus ride to the university, various pitches for things to work on throughout the day were done. I pitched the CHASM hackfest and invited anyone who would like to do GPG keysigning to somewhere in the room that was labeled as ‘E’ on the board. At the hackfest, I talked with David Malcolm about the Python3 happenings and how it affects the KDE side of things. He discussed what needs to be done to spec files to get them ready for the Python3 fun that Fedora 13 will sport. Adding PyQt4 and sip to the list of packages that need attention for the large majority of packages in KDE land to be ready for it. We went through the sip spec file and did preliminary work to get sip working with python3. Other than some of the black magic that it does, it went well and python3 on the KDE side of things should be done once sip, PyQt4, and the ever inimitable kdebindings support python3 parallel installation (other than the normal work of splitting the users and dependencies of these packages). John Hawley then showed up and Rob and I discussed what CHASM was, what kernel.org does, what it would need from CHASM to consider it, and other related things. He liked it, so it means that CHASM is on the right track.

Later, after some pizza got ordered, I hunted down Máirín Duffy to discuss a logo for CHASM. I had an idea in my head of what it would look like, but Mo was able to improve upon it and is capable of bringing it to fruition unlike myself. We also talked about some of the things about RPI since she is an alumni.

Towards the end of the night, a bunch of us gathered in room ‘A’ to work on different things. I didn’t get much work done there other than some fiddling with xmonad (rather than ratpoison) on the netbook. Haskell is very interesting and monads make my head hurt. Towards dinner time, those that were still at the university headed over to a mall to get some food to eat. I talked with one of the XFCE maintainers (sorry, can’t remember your name) about what CHASM was, differences between GNOME, KDE, and XFCE and the different use cases they fill.

After dinner, things such as “It’s a trap”, nethack, and “The Website is Down” were displayed on the projector. Not much work was done on my part after dinner, but it was fun in any case.

That night I went to the hackroom again and talked with Adam Williamson some more, debugged an issue where some kernel upgrade botched up the initrd creation for the new kernel making it fail to boot with Dennis Gilmore and John Hawley, and more.

The next day at the hackfest, Rex, Jaroslav, Steven, and I discussed KDE related things from what needs to happen with kpackagekit, the Network Manager frontend, updates, and other details. Rob and I did some documentation for CHASM, thought up of being aware of preferred files due to cache status for CHASM, and overall was a more productive day. Rob and I talked with Mo about coming to RPI to talk about the Fedora Project when Red Hat visits for the career fair in the spring.

Upon returning to the hackroom, we attempted to debug the network and determined that it was most likely due to the gateway overheating and shutting itself off once it hit its limit. Stephen Smoogen, David Malcolm, Spot, Mel Chua, and I went to play some Dungeons & Dragons. I first played this past summer, it was the first time Mel was playing, and the others have played much more. We got to the first fight, beat the tentacled arm thing and then called it a night.

The next morning was the last morning of FUDCon. The FUDBus left around 9:30, and Rob and I headed out around 10:15. We caught up with the FUDBus on 90 on the way back, but made a stop to stretch and lost it. I would like to thank Red Hat for my funding to go, Chis Tyler for getting the space at Seneca, Mel Chua, Paul Frields, and all the others who made this FUDCon possible. I’m looking forward to seeing everyone at the next one I can attend.

To Adam, I mentioned that I would be trying out GNOME Shell anyways, so no need to worry that I was trying alpha software at the whim of a comment . I tried gdevilspie, but couldn’t get it to actually do anything. The options indicate it can do what I want, but it just doesn’t do it. The GUI says the service is running, but neither Metacity nor GNOME Shell picked up on the rules at all.

To Nicolas, I know that the screen is actually 129 dpi, but the fonts are too big then. To get them to the size I want them to be, I have to set it to 6pt and 4pt for regular and small text, respectively. The font selections go to 6pt as a minimum, so anything meant to be smaller is just the same size. I’m sure I can edit it in gconf, but meddling in there is after I try doing it all through a GUI. The fonts getting bigger with higher dpi seems backwards to me since you should be able to show the same size with smaller physical size, but maybe I’m missing some key detail as to why the opposite is true. So the only way to get the font sizes I want is to set the dpi to a lower value, probably 96 dpi since I know the font sizes at that dpi.

To Alexander, pressing Shift while moving a window gets the effect I want, but only snapping is jarring and not really what I expect.

To twilightomni, the animations are something that I find dispensable. They aren’t necessary for things to work. The main reason I find fault in requiring 3d acceleration is that nvidia users are left out in the cold or forced to go for the blob. One of the reasons for this install was so that my KDE install with nouveau isn’t screwed up when I no longer need the nvidia driver. I have more on this issue today.

Second, the nitpicks that I’ve found with usage. First, I tried to use Guake which is supposed to be similar to YaKuake. Upon trying to start it, it just crashed (yes, a bug has been filed). I’ll look into it once the issue is fixed. This crash triggered abrt to report the crash. Now this is an issue that I’ve been finding all over the place, but dialogs are not given focus when they are created. I need to manually move the mouse (or alt+tab, but i have issues with GNOME Shell’s way of doing this as I went over yesterday) to give the dialog focus. In abrt’s case, it also put the dialogs underneath other windows or dialogs making me wonder why the application had become unresponsive. Having translucent windows as I normally do would help avoid the discovery that there is a modal dialog hiding among the others, but it would only ease the issue, not solve it. I also spotted a spelling error in the dialog that requests for credentials. The word “below” is misspelled as “bellow”. If this post isn’t enough to be considered a bug report, I’ll file a formal one.

There are also some migration pains with Nautilus. In Dolphin, F10 creates a new directory and prompts for a name. In Nautilus (and elsewhere), F10 gives the menubar focus (just like in the latest incarnations of Windows). Both KDE and GNOME also allow alt+<letter> to access the menubar, but in GNOME, F10 is a wasted key. This also conflicts with htop when inside of a gnome-terminal, but at least ‘q’ still works to quit as well. The screensaver also automatically locks the screen. I want it to lock, but I want some time to catch it and prevent it from forcing me to put my password in if I’m reading something (like my notes for these posts). I also found a quirk in PackageKit. After selecting and installing a set of packages, they are still checked in the interface, but the “Clear” action does nothing to change these selections which is a little odd, I thought.

For GNOME Shell, I do have issue with it not supporting a low-end setup without 3d acceleration. First, nvidia users are left to either stick with Metacity or use non-Free drivers which require some additional repositories to be enabled. Other chipsets have similar issues, but nvidia is the most noted one. Personally, I have found that the only thing that is used that really requires acceleration is the “Activities” view. Everything else is just animations and other effects that don’t really add any useful features to the experience (maybe I’m missing things, let me know). It is sort of necessary since there is no other way to easily launch applications that I can approve of for general use. With no other option to fall back to, there is no way to lower power consumption. KWin suspends its (well, PowerDevil triggers it, but it’s there) compositing when low on power to help prolong battery life. GNOME Shell has nowhere to drop to other than Metacity which loses essential and key features of GNOME Shell.

While on this subject, there is some lack of interoperability between Metacity and GNOME Shell. The activity count for GNOME Shell is not turned into a desktop count in Metacity and when going back to GNOME Shell, there is again only one activity. This was irritating as I had to reorganize my windows after each switch (though KWin would be able to pick up my rules and put important things where they belong). I know that window manager to window manager switching usually collapses all applications to one desktop anyways, but if GNOME Shell does have a power-save mode, Metacity would probably be it and they should work better together in that case.

The “lg” command that GNOME Shell has is nifty, but I don’t know what objects are available. If there was something like Python’s dir() function, discoverability of things I could do with it would be greatly improved[1]. The history output also seems to not keep the output scrolled to the bottom if it is there forcing manual intervention to see output.

Third, and probably the most interesting part of this post, is my test to try GNOME Shell and plasma-desktop at the same time. All I did was install kdebase-workspace and launched plasma-desktop. To my surprise, it worked and there were no dead_kittens. Plasma picked up GNOME Shell’s activity count and pager happily displayed them all. However, the workspace switcher in GNOME Shell was confused and only saw one workspace (with the ctrl+alt+left/right keys, going back to 1 worked, but you were locked in once there) even after plasma was killed (switching to Metacity and back fixed it). Plasma also stole the notification area (status bar) and GNOME Shell didn’t get it back until the Metacity switch as well. After finished with plasma-desktop (and KRunner), killing them brought back the nautilus desktop as one would expect. The only thing that didn’t really fit was GNOME Do which still stuck to the bottom edge behind the plasma panel, but it still worked despite being there. I also removed kdebase-workspace (with yum history undo) to prevent future temptation to just use them instead.

Tomorrow, I’ll have some RAM usage statistics I gathered last night.

[1] As a note, I just figured out what the “Evaluator”, “Heirarchy”, and “Properties” labels do, but that is, again, more mousing than I like.

First, things that got fixed from my last post are that the audio applet showed up after the first upgrade, so I’m guessing the PulseAudio issues that were there caused all of the issues and the upgrade fixed that all up. The nvidia driver works, but it has those issues that I had last time I used it every now and then (waking up, it goes right back to sleep, some jerky animations, screwed up DPI (129 instead of 96), etc.). In response to Adam’s comment on my last post for the driver, the dual booting with another Fedora install makes the kernels a touchy subject and not something I want to deal with. I ended up “rebuilding” the RPMFusion SRPM myself against this kernel. Thanks to the help with getting the “focus follows mouse” setting and the opening terminal entry for the desktop. I did end up getting my keyboard layout settings fixed, but gnome-shell has broken them again (see below).

This morning I had to poke around with some code, so I fired up gedit, played with some settings (mainly tabulators to spaces and 4 spaces for indentation). After some mistakes with typing, I found out what kept causing me to screw up: ctrl+bksp behavior is very different from what I am used to in KatePart. In KatePart, it goes to the beginning of the next <word> and since KatePart is used pretty much anywhere non-trivial text editing is needed, it’s consistent everywhere. In gedit, however, its movement is very different. Going forwards, it goes to the end of a <word>. Numbers are considered a different <word> which adds to the confusion (I’m getting used to looking ahead in vim for command counts, the numbers adding to it throws the counts for key presses off). Going in reverse, the cursor is placed at the beginning of a <word>. This duality is disorienting at times. There is also some odd behavior near the end of newlines. In KatePart, ctrl+bksp will stop at a newline effectively just clearing it. Instead, gedit also deletes the last <word> on the line before which causes me to manually hit Backspace to delete what I want. Word jumping over newlines follows similar patterns and is hard to get used to. For actual coding (pretty much anything beyond what that class requires), I don’t know how well gedit works out, but it’ll take some getting used to. I was pleasantly surprised to see a Kate color scheme in gedit.

I discovered some odd behavior with the volume keys as well. If something is selected, changing the volume with the keys makes it use the “hovered” color while the button is pressed. Another oddity I found was the “desktop effects” configuration selection for Compiz had quick options for “wobbly windows” and “the cube”, two of, what I consider, the useless effects that have been made. I usually use “Present Windows”, “Zoom”, “Desktop Wall”, “Thumbnail Aside”, and “Translucency” in KWin. Other than that, it’s just some visual cues that I could live without like a minimization animation. The rest I see as eye candy (well, the other desktop switchers can be useful, just not my preference) and not really necessary, so why two of the more “useless” effects are front and center befuddles me.

I also found some inconsistency with dialogs. Sometimes Escape works as if I hit “Cancel” and other times it just does nothing. One such dialog is the “Add/Remove Software” window. Visually, it screams “I’m a dialog” to me since it lacks a menubar and toolbars (though these aren’t always applicable) and yet Escape does nothing. Just something I’ve come across and I’m sure there are more around.

How am I supposed to eject flash drives in GNOME? I kept trying to do “Safely Remove” but it kept coming back that “this file could not be stopped” (though it is a funny error message, fairly for those who don’t know about processes holding file descriptors open). Eject works, but it’s not what I’d choose with an option of “Safely Remove” right there.

I also installed GNOME Do and tried it out. Unfortunately it is also not KRunner by a long stretch. I search for “naut”, GNOME Do suggests “Livna Display Configuration” and only that. Why not Nautilus? I know what it’s called, let me launch it. KRunner also searches my bookmarks, contacts, recent documents, path names, Konqueror web shortcuts (very handy), Kate sessions, offers a calculator, and lots more. GNOME Do doesn’t do any of that (as far as I can tell) out of the box. I haven’t searched for plugins yet, so any pointers on good ones for the things I listed would be nice. GNOME Do is also not a standard text entry widget, so ctrl+bksp doesn’t work at all.

Window management is one thing that I’ve gotten touchy about after getting used to KWin. Metacity and GNOME Shell both have the following issue that bugs me: windows only snap to edges when going towards the edge and not when the two windows previously overlapped. I have high sensitivity on my inputs (but smooth, which is where the Windows drivers start to annoy the hell out of me with their jerky scrolling) so I overshoot placing windows often and “reverse snapping” is nice. KWin also snaps corners to corners and when edges line up (the top will snap to the neighbor’s top line if the two windows “touch”) which Metacity and GNOME Shell are both missing. Resizing windows is also weird with it depending on where the mouse starts within the window. In KWin I don’t really resize things much since I lock windows’ position, desktop, and size for persistent apps (and size for common ones like Konqueror or KWrite). I also set shortcuts for persistent apps (meta+alt+i for Konversation for instance). As far as I can tell, neither Metacity nor GNOME Shell has this ability, so I’m left to manage such things manually and since default sizes are way too small for 129 DPI, I have to resize things.

While on window management, I’ll report my initial findings for GNOME Shell. The first thing I noticed is that its alt+f2 tool is even more featureless than the default GNOME one. It only accepts complete command names, no tab completion, no ctrl+bksp, forces focus, and other things. This mix of things it’s missing makes it pretty much useless in my book since opening up a terminal and letting bash auto-complete for me is much faster. In addition to this, the GNOME Do shortcut gets clobbered somewhere in GNOME Shell so it has to be manually started.

The alt+tab switcher in GNOME Shell is also hard to get used to. If there are two windows for an application, they cannot be separated without the mouse or hitting the down arrow. I was able to get it to display the separate windows with nothing but alt+tab a few times, but it’s not reproducible reliably. Window thumbnails are also hidden in this submenu. Being used to “Present Windows” (with two variants, all desktops and just the current one) which has filtering, this is a step backwards for me anyways, the awkward navigation just adds to it

The list of applications next to the “Activities” menu doesn’t make much sense to me. It only lists the currently focused application so of what use is it? In addition it says “File Manager” for the desktop. How many people know that the desktop in GNOME is just a Nautilus view of a folder with a wallpaper? I didn’t until today. I kept trying to find this “File Manager” that was open so I could close it.

The “Activities” area is weird. There’s no keyboard navigation unless you’re searching (which also does not support ctrl+bksp by the way). You can add and remove desktops here as well. Adding them is no problem. I find issue with the way desktops are removed. You can only remove the last one and the way it works, the position of the last desktop keeps moving. I was also unable to find keyboard access for this. The other issue with these desktops is that the layout with the “Activities” visible is 2d. When back on the actual desktop, it is 1d. On KDE I have my 9 desktops set up in a 3×3 square with shortcuts that treat it as 2d and 1d (if I had 6 more arrow keys and shortcut actions to attach them to, there would be 27 in a 3×3x3 logical layout ). GNOME Shell has no way to treat them as a 2d grid. In addition, I did not find any shortcuts for moving windows between these desktops. You have to get the window menu (alt+f3 with KWin, click in GNOME Shell) and choose to move it right or left. There is also no wrapping, so to get a window from desktop 1 to desktop 9, I have to do this 8 times. There’s also the ability to move windows when in the “Activities” area, but this also requires a mouse and lots more of it.

GNOME Shell also likes to play around with shortcuts. Since the “Activities” area is triggered by the Meta (also known as “Super”) key, my shortcuts to the desktop through VNC are chomped when they use the Meta key. Not fun. If the “Control to show mouse position” is active, the Control key modifier is also impossible over VNC (now using TigerVNC since vinagre is really annoying). My keyboard settings for CapsLock-as-Backspace and RtAlt-is-AltGr, Shift+RtAlt-is-Compose are also completely ignored. Setting them manually with setxkbmap works as expected, but the keyboard settings are completely ignored. The GNOME Do access key (Meta+Space) also seems to have no effect when in its dock mode forcing the mouse to access it there as well.

With GNOME Shell active, GNOME Do defaults to a Macintosh-like dock. It’s nifty and looks nice, but not really my thing. Unfortunately, with what other ways there are to launch applications, it is the least annoying (though I just found the “Sidebar” and it looks a lot better, will report on that tomorrow). Similar to the alt+tab switcher, the dock is hard to use with multiple windows for one application. Right click on the app and then click on the entry you want. No thumbnails to differentiate them if the titles are the same or keyboard navigation.

Reflecting, GNOME Shell gets a “meh” from me so far. I hope that it’s still alpha so that more features can be added (assuming standard alpha and beta definitions). If it isn’t, GNOME Shell’s first release will probably feel like KDE 4.0 did to many. In either case, KWin and Plasma have been much more powerful from my experience and they aren’t as much as a detachment from how things have worked for the past many years.

Experiment for tomorrow: gnome-shell with plasma-desktop. The /proc/acpi/dead_kittens file will probably be reporting numbers in the dozens.

Luckily the install I’d like to keep clean (the KDE one with the nouveau driver) had the kernel for the Beta still installed. When installing the GNOME instance, I didn’t install a bootloader since the possibility of overwriting the good settings was too risky. It turns out that Anaconda did nothing with /boot in this case (it made a directory for it on its /, but that’s all I can tell). Maybe it did and I didn’t notice, but there was no GRUB entry for the new install and the kernels both went to the first install (though the settings on the menu and timeout do seem to have been modified, not sure what’s going on there).

After hacking the GRUB listing to boot the GNOME install, I did an upgrade. I skipped the kernel since I don’t want to see how the two installs would get along with each claiming ownership of the files in /boot, so the GNOME install will keep its kernel and the KDE install won’t uninstall the version that the GNOME one is on.

I think that in order to fix issues like this, there needs to not be one file in charge of GRUB entries. One way I can see this working better is for GRUB to have global settings (and a GUI to control these) that all installs obey and to have a directory with one directory for each installed system on the machine. This way the Ubuntu magic incantations can stay in one file and the Fedora lists can each be separate. GRUB would then scrape all these lists and boot whichever it needs to. This is much more extensible and allows the distros to each do their own thing and not step on each others’ toes with defaults and whatnot.

I must say that the GNOME frontend for PackageKit is much more advanced, but KPackageKit allows for a more pain-free way of dealing with cerdentials. I don’t know how many times I entered the root password to fiddle with packages, but it’s not fun with my password. The dialog for the password request also doesn’t get focus without clicking forcing me to act to deal with it. This with no setting for “Focus Follows Mouse” that I could find annoyed me. I also had issues with PulseAudio popping the speakers every other second. Luckily the update fixed that, but it was really annoying while it was going on.

As for GNOME itself, I feel lost. Since PulseAudio was popping the speakers, I wanted to mute them. However, when I went to the Sound entry under settings it just sat waiting for PulseAudio and never did anything. I couldn’t find any mixer like KMix, so I started up a terminal to do alsamixer. When I ran it, it gave an error about PulseAudio not being reachable. Fine, at least there’s a reason that the Sound configuration doesn’t work. But it then crashed nm-applet somehow. Unfortunately, abrt had failed to start for that session and there’s no trace for it.

Some other little things I saw were that the icons in the menus sometimes “popped” into existence after the menu was displayed. The Alt+F2 tool (not sure what it’s called) is similar to what KDE3 had and not the KRunner I’m used to and the combination of that, the menus being foreign to me, and me not using menus normally made it awkward to launch applications. KRunner can do substring suggestions and also searches descriptions so that Konsole matches any of “term”, “kons”, “sole”, and others. To launch the GNOME terminal it’s “gnome-terminal” (tab completion helps, but KRunner is still far superior here). Since terminals are a dime a dozen and I usually close things as soon as I’m done with them, this is unproductive for me (I normally use YaKuake).

I also coupled my /home partition with the new install, but firstboot required me to use a different name or chance that the uid for my user on the new install and the old install were the same. I’d like a button to say “Yes, it’s my $HOME, set my uid/gid to what that directory’s ownership is” to help with migration and reinstalls.

With this, it shared my other install’s xdg-dirs settings, which has everything pointing to $HOME. This is the first time in a while that I’ve had icons on my desktop and I can’t say I miss it. That with the overlapping (which I was able to fix after fiddling with the context menu) and previews, it makes my desktop feel cluttered, but if I could get rid of the previews, I’d feel at least a little better about it, but I’ve been unsuccessful in my attempts to get that to happen yet.

I was also unable to find any GUI control to fine tune my keyboard settings (I set the Caps Lock key to be an extra Backspace and RtAlt to AltGr and Compose with Shift). I know the commands to do it, but would have liked to set it in a GUI somewhere. Speaking of configurations, it took a while to find the control panel (similar to systemsettings in KDE) and it was an accident. I had kept going to the menus (which I find awkward with, but if there’s keyboard access to them, I’d feel better) so configuring was another annoyance.

I use VNC to access the session running on my desktop, so I editied my client script (it shunnels the VNC session) to use the client shipped by default. The biggest annoyance is that there’s no cursor shown at all, so I have to rely on effects in the display to guess where the cursor is (I’d use the keyboard if the local X session didn’t eat all the important ones).

Another thing I am missing from KDE is the context menu on the desktop. In KDE, there are entries for Konsole, logging out, shutting down, and locking the session all in one convenient place. Instead, I have to go to those menus (requiring a mouse) to do any of this. The ctrl+alt+del menu also doesn’t have logout, but the other entries being there is nicer than going to the menus.

After playing around a bit, I had to install the nvidia driver so that I could use the install for the class I have to use Panda3d for (this was the main reason for a second install; so that the KDE one wasn’t polluted with this thing). The kernel the GNOME install has is older than what the RPMFusion repositories carry, so I had to “build” it myself. That wasn’t hard. THe gotchas were allowing it to execute code on the stack and rebuilding the initrd since nouveau was in the one from the KDE install. The nouveau and nvidia drivers really do not work together at all. With nouveau, you have it taking the frame buffer and nvidia not being able to grab the device and uninstalling the nvidia driver is just a pain. It also does not do 96 dpi out of the box and all my fonts are huge (physical dpi is around 129 or so).

I’ll be playing around with this install for a while (3 projects to go with Panda3d), and I’d be grateful for any tips on how to do anything I failed to do and described here. I’d say that most of it is that I’m just missing where things are for half of it, but I wouldn’t be surprised if I had to go app searching for some of it (like VNC). I’ll also be trying out gnome-shell later since I have 3d acceleration with the nvidia driver (one reason I think it shouldn’t become default until nouveau supports 3d or nvidia is miraculously released as Free software).

Another gripe I have is that it puts everything down two levels of dereferencing. Shapes use a gradient, which is in the defs, that’s okay. But then that gradient references another gradient for its stops. Why this separation? I’ve also seen objects defined that aren’t referenced at all (I imagine gradients that were made then ditched later).

Inkscape also likes to export attributes that aren’t necessary. If the fill is none, the fill opacity, color, and rules mean nothing, so don’t bother writing it out. The id tags on objects are just numbers, making use of search necessary. Everything is put into style=”foo:bar;” instead of foo=”bar” format. I’m sure this makes parsing slightly faster, but it’s irrelevant since how many extra attributes are set (not to mention Inkscape-specific ones). It’s harder to read and edit since syntax highlighting tends to be “string blob” and not “string of key/value pairs”.

I’d like to see Inkscape have an “Export to SVG” menu option so that a nice, minimal SVG is output. Without the Inkscape-specific extensions, without the unneeded attributes, etc.

Maybe it’s just me, but anyone else do their drawings in a text editor (including, but not limited to, SVG, MetaPost, pic, TikZ)?

I’m not sure how long the table will take to propogate and I am only up to page 60 or so of the things on the kde-apps website, so keep an eye out for any applications or libraries you use or would like to maintain.

This machine has proven to be very useful. Light and agile for class. Quiet, suspend/wake up in a decent time (in KDE, there are…faster ways however). A full KDE install fits on the machine (though I have nm-applet, but I removed basically everything GNOME-ish from the thing, even nodoka so it looks like X days) and I still have about 1GB of space left to clone repositories and hack around. RAM usage (it has 512MB) is amazing. KDE fits and operates in under 160MB of memory.

However, I do have something else I have been attempting to try out. I set up the antidesktop on it using ratpoison. After screen starts up, I have a grand total of…60MB of memory used. And that includes the htop instance used to view usage. I’d like to get some console apps running to try and stay in screen as much as I can, but mutt seems to be a pain to get to do what I do with a basically stock KMail (5 SMTP/IMAP accounts side-by-side), so I’ll end up using KMail at least and probably KNode as well since it’s the same package. I have Bilbo and Choqok installed for blogging and denting, respectively, as well. I wish there were command line clients for these (full read/write, not just posting), but this seems to be a dead end.

Besides the apps, I do have some things that I have been unable to get working nicely. If anyone else has a similar setup the current issues I am having with the ratpoison setup:

• Losing session DBus connections from the screen instance when I restart ratpoison for config tweaking
• wpa_supplicant magic (nm-applet dies with X and cnetworkmanager can’t find nm-applet configs to use those)
• locking X and screen on lid close
• sleeping on lid close (currently using a script to execute acpitool -s, which is very fast)

Any tips? My Google-fu has not turned up anything.

I am also learning to use the Colemak keyboard layout on the eee. The keyboard is different enough that switching between the two is easy. The biggest change (for me) will be that there aren’t any dead keys. I use us_intl normally and tend to type out “‘ ” a lot on non-us_intl keyboards.

Today I wrote up the documentation (along with screenshots) for the overall data structure for Sigmods.

I also did some wiki work for the KDE SIG. Seeing as my motivation for coding is still low, I figured I’d do some more mundane stuff to help. I’ll get around to cleaning up the other pages in the coming week.

Luckily (and unluckily), the laptop started having performance issues and VNC was utterly unbearable, so now I’m on the desktop where I did all of the work anyways (plus being able to type ‘<’ again is nice), so now there’s one less layer between working and what I’m seeing (keyboard shortcuts are eaten locally if possible before VNC gets them, so the mouse was being used more…which is not really my thing). Now I’m using synergy between the two and it works well, though I’m finding that it’s hard to use both. About the most I can do is fire off a mock or rpmbuild on the laptop and then get back to IRC or whatever.

The company seems to be working to get out of the binds of being proprietary. Supporting system copies of dependencies is apparently one of the top priorities, but changing from “python is here” to “run python” isn’t all that easy (there’s $(THIRD_DIR) all over the place in the .t files and the patch would be…monstrous). The mailing list does not have an nntp access point (such as GMane) and I was unable to find any public archives, but that should come in time I imagine. I’ve joined the ML for developers in the meantime to try and help with the system dependency stuff.

I also put Bilbo Blogger (which I am using to write this post) up for review today (I’ll swap reviews). The post previewer is great as well as the support for editing older posts and their details.

In other news, I’ve updated some stuff around here. WordPress is now the latest version and the comment validation plugin has been modified be more confusing to bots that try to post here (but not to humans).

• Robot Visions by Isaac Asimov
• Disclosure by Michael Crichton
• Sphere by Michael Crichton
• Frankenstein by Marry Shelley
• The Catcher in the Rye by J.D. Salinger
• Smoke and Mirrors by Neil Gaiman
• Mathematical Fallacies and Paradoxes by Bryan Bunch

and those that I am currently reading:

• Gödel, Escher, Bach by Douglas Hofstadter
• American Gods by Neil Gaiman

It does feel good to be back reading again. I only got through 4 or 5 of the 60 I took to college last year. So now I’ll be taking less and reading more so instead of 6%, I hope to be closer to 80% of those I take up. I may write stuff about them in the future.

On the coding side, I’ve cleaned up some stuff in Sigen from fixing Sigmodr to behave better with minimum values of some variables showing descriptive text instead of having -1 be -1, it says “No limit” which is how the engine interprets it where applicable.

I’ve also started to work on the Kross fixes I’ve been wanting for a while in KDE’s playground. Not sure how long it will take with the current schedule I keep. I do have time to work on it today, so that’s where I’ll probably be until everyone else gets back.

I also started to learn the Colemak keyboard layout. It’s proving to be interesting and I like it better since the home row is more useful than QWERTY’s.

I’ve started to package Kobby for Fedora. Should be up for review by the end of the weekend (with dependencies libinfinity and libqinfinity). I’ll have it in my personal repo until Fedora proper has it.

* One thing that I’ve always hated was the touchpad. It’s too finicky and usually just gets in my way. However, I’d like to have one on whatever I get next since I use it for scrolling (everything else is disabled: tapping, movement, buttons, the works). I don’t think I’d be able to work with one (especially if the buttons were on the side) since I’ve gotten so used to the trackpoint. I thought at first that I would be doomed to live with a trackpad until I saw the Sony Vaio P. It has a trackpoint and all three mouse buttons. Unfortunately, it is overpriced and only comes with software I’d rather not give money to.

* I also would like at least 8 hours of battery life. I can get 4-5 hours from the good battery for the T61 and on powersave, so I don’t think this is too outrageous for a netbook.

* It has to come with Linux (or nothing). Even if I’m wiping it for Fedora, I’m not giving money to companies that don’t deserve the money.

* The Insert/Delete/Home/End/PgUp/PgDn keys must exist on their own. Having the latter 4 be modified arrow keypresses is not excusable for me since I use them so often (this kills the Asus machines). It’d be preferable to have them in the standard 2×3 arrangement, but it’s excusable if it means fitting on the machine (laptops don’t have an excuse).

* It has to be less than a kilogram. This is pretty common, but the larger ones tend to approach this limit.

* Sleep and hibernate should work. This is a no brainer, but it currently fails on the T61. It’s not really that much of a problem since boot time is low and the thing isn’t exactly agile either.

* Solid state drive. I have a desktop to tether it to, so 20G would be good, but I could probably work with a 12GB, 8G would be cutting it close. My current installs work on 10G and 15G / partition sizes (/home is the only one that’s separate) and are usually more than half free depending on the debuginfo and mock setup at the time, both of which won’t be happening on the netbook.

* Ethernet port on it. You’d think this would be a given, but the Vaio P is too slim for it and has some weird adapter that’s probably overpriced and defeats the purpose of the thing being so small.

* A webcam would be nice, but it’s not a showstopper.

* It would be nice to have 2G of RAM, but 1G should do it; currently I’m using on 650MB on the laptop and a little over a gig on the desktop with lots of extra stuff running.

Unfortunately, I think it may be a few years until I can find anything on the market that meets these requirements. Even laptops have trouble making them (except for the weight thing which I’d waive for them). Unless there’s one I’m not aware of. Thoughts?

I also updated the front page of the wiki so it isn’t the stock one. I hope that I can get the project some kind of visibility with the 0.2.0 release, but since I have yet to actually get it to run reliably on Windows, it’ll be a chore. Hopefully making it available will be easy and the KDE4 Windows installer will be able to handle it.

I’d say not. I’m really not sure I’ll ever be using Chrome (at least according to this page). They forget one thing in their list of what not having a 64-bit version available means. You have to duplicate the disk space for libraries that I have 64-bit versions of already. So yes, you may have larger pointers and executables, but by forcing 32-bit, I am forced to take up lots more disk space. Just installing gtk2 here would be 19MB of space. This is much more than the difference between a 32-bit executable and a 64-bit executable for a browser. The difference in disk space for KDE 4.3b1 i586 and x86_64? 9201809 bytes (9.77MB). This is over 1.4GB of rpms for each (this includes debuginfo, srpm, binary rpms, and repodata for yum). So for installing a 32-bit version of gtk2 to be cheaper than just offering a 64-bit version (on-disk), the executable has to be (assuming linearity) around least 3GB. I don’t think that would be acceptable. In fact, why it’s 511MB (according to a quote in this article) is even beyond me. That’s 1/3 the size of KDE, so Chrome already sounds bloated beyond belief. So the arguments they list for not providing a 64-bit build are flaky. The most defensible one is the V8 not having x86_64 code generation, the rest are poor excuses.

Now, I know that disk space is cheap, but not so for my / partition, which I usually keep around 10-15GB. This is enough for all of the typical stuff I install, room for new stuff apps out, and a couple debuginfo packages. The only thing that tends to hit the limit is large mock builds and even then only after not clearing the cache and theres lots of old, unused packages there. I like to keep it minimal, and 32-bit libs on my 64-bit machines is usually a show-stopper to installing something these days.

Then there’s the decision of which look’n'feel to use. Now I’d prefer Qt/KDE, but it doesn’t really matter as long as you pick a side and stick to it. Sitting in the middle just makes you a weird application that doesn’t really fit on either side. Mozilla has a nasty habit of doing this (on Linux at least, Windows is already a misshapen landscape in this regard) and making me squirm when their apps fail to work with the other apps I use. Mainly, Firefox needs to grow some support for shared-mime-info and stop trying to be special. I can make GTK look like my KDE apps (though size hinting for widgets is off, that’s minor; the real miracle would be replacing the horror that is the GNOME file dialog with KFileDialog), but if an app chooses to use some of GTK and then custom draw other things, it’s going to look even stupider when the KDE theme is attempted on top of it. Don’t be a sore thunb on someone’s desktop. Theming and skinning suck, just deal with it and use the system colors.

Speaking of Firefox, the one-process-ever deal. I don’t like it. One window crashes and they all crash. I understand what Google’s doing with the one-process-per-tab, but I think it’s a little too much. One per window is enough and session restoration should cover the rest. Which Firefox has also had issues with. At times it has been one or two tab changes behind and forgotten history when restored. Generally makes that a pain. Konqueror has remembered form entry data as well, which I haven’t seen Firefox do yet (though it does crash less, ted.com was reliably doing it, but it’s Flash so there’s nothing anyone can really do). Ctrl+Tab is a disgusting key combination, but I have yet to find a way to get away from it. I like Ctrl+, and Ctrl+. better. They require moving two fingers, Ctrl+Tab requires rotating my hand. Yes, my shortcuts matter to me, but Firefox has them hard-coded it seems.

I think the browser wars are good. Keeping other vendors on their toes and having to make new things instead of settling with what’s done is good. The proliferation of JavaScript engines however, I’m not all that thrilled about. Hopefully there won’t be the need to do browser checks once things move onto The Next Big Thing. They suck and I certainly won’t be doing any favors for non-standard browsers. Custom extensions to the standard also suck for the same reason. WebKit also has its issues, such as not using system widgets and instead opting to draw their own and look different (at least in QtWebKit; I couldn’t care less about Safari or Chrome since I probably won’t be using them in the forseeable future).

By the way, I use Konqueror. It runs lighter than Firefox (12MB vs 29MB at startup with extensions that brings Firefox to just below accessible feature parity). It has no need for the basic extensions. Adblock, sane downloading, noscript, and on-demand plugin denial are all with it. The only major thing that is missing is some kind of GreaseMonkey equivalent. It Just Works with my other applications. No need to customize “Open With…” bindings and all that stuff.

It’s been over a month since I’ve paid attention to this nook of the Internet. Another year of school ended, got my wisdom teeth removed, gotten into playing some Dungeons & Dragons, hacked a bit on Sigen, building KDE 4.3 beta 1, finally getting VNC working, and some other miscellaneous stuff.

I was able to replace 10,000 lines of code with about 1,500 lines in the tree widget in Sigmodr. It now no longer needs right clicking to add and delete items in a game. It has lost drag’n'drop and copy/paste support for now, but the maintainability of the new code is well worth the loss.

The map editor has had a few bugs squashed as well, though I may have to switch over to a new way of doing it since there’s a test case I hadn’t thought of that the current algorithm is choking on. To visualize the case, imagine a π and then placing that on a platform. There is now an outline and an internal area. The current algorithm screws up the internal outline since it doesn’t know that the other leg (on the same polygon) is the target; it only looks at the other polygon for a target point. So it’s broken and fixing it would take a lot of time. I’ve been trying to think of a data structure scheme which will nicely accomodate the vector idea I was tossing around even while making the current algorithm. A map would work, but I don’t want to have to fiddle with pairs and whatnot.

I’ve also built KDE 4.3 beta 1 for Fedora 11 (i586 and x86_64). I don’t know when they’ll hit the kde-unstable repo at kde-redhat. I have a list of bugs to submit, which I’ll work on submitting in the next few days.

The Kopete runner I was working on is still at the same position it was before, but after talking with Matt Rogers, it looks as though Kopete’s DBus interface needs to be fixed up and redone for it to work. This means that it will have to wait until KDE 4.4 and I’ll be learning myself some DBus goodness. It doesn’t look all that bad.

I would also like to announce that I have my server up with a public address now.It has HTTP and git services. The important URLs are:

• Trac pages
• gitweb

Most of the git repos are there to keep old code in VCS if I ever hack on them again. The older stuff is scary, not really for the faint-at-heart.