KMail’s Akonadi migration in openSUSE

In openSUSE’s KDE team, we’ve recently planned the migration to Akonadi, the groupware caching solution that will be the base of upcoming KDE PIM versions, notably KDE’s address book, email client and calendar app. With the release of KDE SC 4.4, we’ve seen the first component being ported to Akonadi, KAddressbook, spearheaded by its fearless maintainer Tobias König. In the KDE SC 4.5 cycle, we’re seeing more components in their first Akonadi-incarnation. As this means a big step for these applications, some attention needs to be paid to users who will, over the coming months, migrate to the Akonadi-based PIM components. The PIM team has decided to go with a stepped approach, and not introduce all applications in their Akonadi version at once. This is a sensible decision, as it allows you to learn from problems in the migration path, and fix these in the next wave of ports. PIM in SC 4.4 brought the address book migration, which wasn’t completely smooth from a user’s point of view. While in most cases, the fix was as easy as "point Akonadi to your contacts (or .vcf) file", we can (and will) do better with the migration of KMail. KMail2 (which is akonadi-based) will not arrive with 4.5.0, though, but is planned become part of the next monthly SC update, 4.5.1. This decision has been made by KDE’s PIM team in order to get a little more time to stabilise and test the release. This is also our first line of defense :-)

As users’ needs vary, we decided to make the Akonadi port of KMail opt-in for the 11.3 cycle. openSUSE 11.3 is based on KDE SC 4.4, and as such will install the "mostly traditional" PIM suite. Users will not automatically be upgraded to KDE SC 4.5 (which is due in August), but in all likelihood it will be easily installable. As there are many people who follow KDE releases closely, many will install SC 4.5.0 and followups from the Factory repositories, so when 4.5.1 is released, these users’ email would get migrated to Akonadi automatically. That might come as a surprise, as it’s unconventional to make such a big technological leap in what looks like a bugfix update. So in openSUSE, we will keep shipping 4.5.0’s KDE PIM even in Factory, but also make available packages that replace KMail1 with KMail2. Users will be able to opt-in to the Akonadi migration, so they can do this upgrade when it fits for them. From a discussion with the PIMsters, it also looks a lot like you can try the Akonadi-based KMail, and if it’s not ready for you yet, you can switch back to KMail1 without losing your config. That’s a great achievement by the PIM team, and shows that they’re developing with end-users in mind.

For the user, this means that for the innocent nothing will change automatically in the upcoming cycle (other than bugs getting fixed). The effect is that there will be a roughly 6 months long window in which users have the choice whether they just want their KMail to not change, or to jump on the Akonadi bandwagon into the future.

This upgrade also gives us (upstream KDE and downstream openSUSE) the opportunity to make the migration and workings even smoother, and deliver some icing on the Akonadi-cake with the openSUSE 12.0, which show why Akonadi is a darn useful thing to have. I know many people (and I am one of them) who are looking forward to make full use of Akonadi, not only in the applications you’d expect it to be, but also integrate all the interesting information from Akonadi also in other apps. I’m sure some very interesting features will crop up after Akonadi is fully upon us.

Epic Moment: Free and fast graphics at last

Well, epic for a Free software geek. Kim and other normal people just chuckled, and at best grinned when I told them.

I’m now running openSUSE on my desktop, which has an AMD RadeonHD 4350 card. That card is fast enough for my graphics needs (compositing window management, a dated game once in a while) and still passively cooled and providing two DVI outputs for my dualhead setup. I’ve been using an NVidia VGA before, but switched to AMD/ATi when I last upgraded my desktop workstation (to an Intel Quad core, meaning new mobo, graphics as well). I picked a RadeonHD card instead of an NVidia chip because of AMD’s open dealing with specs, something which I deeply despise in NVidia. So, NVidia, there you go: You lost me as customer because you’re too closed a company.

Now bitching about NVidia is not the (primary ;)) goal of this post. The goal of this post is, that with openSUSE 11.3’s graphics stack, I’ve been able to run a composited desktop with the Free radeon driver finally. While there were some hickups in the installation procedure (normal for a beta release), it now all fell into place, and I’m enjoying fast, beautiful graphics with a completely Free software stack. That is a lot of work from the Xorg people that has finally come to fruition: There’s the new DRI framework in the Linux kernel, along with drivers supporting compositing for many of the "newer" RadeonHD models, the compositing support for newer chips that has landed in the latest Xorg, all on top of the EXA acceleration infrastructure. If you’re interested what you can expect from your upgrade to 11.3, check the feature matrix compiled by the Xorg devs.

I’ve tried the setup on Ubuntu Lucid Lynx in the first place, since that’s what was installed on the workstation before, but didn’t have much luck. The kernel shipped with Lucid is just one version too old it seems, so lacks the functionality necessary to accelerate graphics suitable for compositing. For that reason, I’ve upgraded the kernel to 2.6.34, but only to find out that my Xorg is not good enough. Tried installing newer packages from the xorg-edgers repo, but that just resulted in my system hanging solidly during boot. (That’s to be expected if you’re using bleeding edge Xorg, and upgrade your own kernel, but still a shame it didn’t just work ;-)). I assume that new upcoming Ubuntu will provide this nice out of the box experience as well, but I didn’t try it.

For many users, the most interesting thing is probably that the Free radeon driver (and to some degree the Intel graphics driver as well) provide the best user experience for running a KDE Plasma desktop. Using the NVidia driver, the desktop always feels sluggish. That’s supposedly due to NVidia not accelerating some calls that are made while switching windows, so a couple of milliseconds are added there and the whole thing feels less responsive. It’s not as bad as it was, about 2 years ago, but there’s still a notable difference in snappiness when using a low-end (non-poulsbo ;)) Intel chip compared to anything using the nvidia.ko binary driver — and that’s just pathetic given the NVidia hardware is supposedly much more powerful. The same, even if to a lesser degree, is true for ATi’s binary driver, the (in)famous fglrx.ko. This one, while it works OK-ish, also suffers slightly from this lagginess. Switching to the free driver makes the whole thing just very snappy. I could well imagine that those who are complaining about a perceived slow system are suffering from just this problem — bad graphics drivers. If you’ve recently compared the binary and free drivers for RadeonHD cards, please leave a comment so we can see if this — rather vague — theory of mine (binary drivers suck, Free ones feel faster) is true for more people.

So, what’s the epic moment? Well, the epic moment for me was seeing KDE Plasma start up with the Free driver, enable compositing automatically and by that delivering a much more beautiful and functional desktop to me, out of the box. I’m happy that with the new graphics stack in openSUSE 11.3, the same will be the case for many users out there.

Jumped!

Last weekend, a couple of friends and I went to an island in the Dutch Wadden Sea to celebrate the bachelor party of a close friend of mine. It was all pretty awesome, I picked up the poor guy around noon last Friday (when he thought he’d have another good 3 hours of meetings that afternoon), so it was a good start. The night, we spent bbqing on the beach (had Jerk pork) and the next day we want to the small airfield on Ameland — Chris still not knowing what would be happening.

The next couple of hours we spent hanging out with the parachuters from Paracentrum Ameland, learnt a bit about safety, how you’d not break your legs, and what happens if you lose your contacts in mid-air. Then we boarded the small plane, one after the other, and climbed up to 3000m above the island, giving a fantastic view over all Dutch Wadden Sea islands. Then I was secured to my tandem master more tightly, put on the safety glasses, and open went the door. At that point, I just stepped out of the airplane onto the small step outside the airplane, stood there for two or three seconds, and jumped. We fell freely for a good 30 seconds until the parachute opened (probably wouldn’t be writing this if it hadn’t ;-)) at ~200km/h, and then spent another 5 to 6 minutes slowly decending onto the airfield again. Landing went smooth, as you can see on the photo.

I must say, that was an awesome experience, it’s a really weird moment when that door opens, but the view, the experience (and certainly the rush of adrenaline) make it absolutely worth it. I wonder if I’ve left behind my slight acrophobia on that flight. As to Chris, I think he still likes me. His face, in terms of Big Smile looked about the same as mine right after the jump.

If you’ve got friends that have you thrown out of a plane in mid-air, you don’t need enemies.

The KDE e.V. Board, May 2010

As we’re having a board meeting in the KDE e.V. office in Berlin, we thought we’d take a photo for posterity.

Left to Right: Cornelius Schumacher (President), Frank Karlitschek Celeste Lyn Paul, Sebastian Kügler, Adriaan de Groot

Claudia Rauch, KDE e.V.’s business manager took the photo. Normally, I’m not the overdressed one (but I do wear my KDE 4.0 Release Event t-shirt underneath. On the photo, it’s not very well visible, but Celeste does have part of her hair blue.

Trinity and the Challenges of Continuing KDE 3

This morning, while having my usual Cafe Latte (albeit this time in Berlin instead of at home sweet home in Nijmegen), I read about the Trinity project, which is an effort to revive KDE 3. I think this project nicely shows the advantages of Free Software. While the vast majority of KDE contributors agrees that KDE 3 is a dead end, technologically, these two guys (according to the somewhat sparse information on the website) are trying to continue to support and feature development on KDE3. Now I see a couple of real challenges for this project:

  • Maintainance – KDE 3 is a large codebase. You need a good amount of people with domain knowledge of many different areas to effectively maintain a project like KDE. I see some of the first roadmap tasks for Trinity are updating the build system to deal with all the updated developer tools (e.g. newer autotools versions).
  • Upstream Support – Official support for Qt3 ended on July, 1st 2007. nearly three years ago. Now this is not necessarily a problem (as Qt, too is Free Software), but it puts additional stress on the maintainance team — you have to provide support for Qt as well, not only KDE. This is also true for many other upstream components
  • Community Support – KDE has many applications, and their developers take conscious decisions wether or not to support a specific version. Moreover, developers take great pride in their creations. Patches made to those components (even if it’s just maintainance) should be reviewed by their developers (who also happen to know the codebase best, have domain knowledge, and so on). Now the resources of these subprojects are also limited, and many decided to effectively end of life the support for their KDE 3 versions and fully concentrate on the KDE 4 version.
  • Interoperability – In KDE 4, we introduced a couple of new technologies, D-Bus as our inter-process-communication mechanism, and the FreeDesktop icon naming scheme for Oxygen, KDE4’s artwork pillar. This was (and is) not possible with KDE3, and is actually only the tip of the iceberg of many improvements that have gone into KDE4, and which cannot be taken advantage of by programs based on KDE 3.
  • Testing & QA – While the whole KDE project switched to KDE 4, we realised that releasing KDE 3 would be a lot harder in the future, since we heavily rely on the community testing the code before releases. This assures that the code is of sufficient quality and actually works. This process is not quite trivial, it needs testing, but also reviewing of changes that go into the codebase. For reviewing code you need familiarity with the APIs used by it, the actual codebase that’s being changed, and on top of that domain knowledge of the area your program is used in (for example if you’re writing a mail client, you should know mail protocols, and so on.) Again, you need a larger team to assure sufficient quality of the code you release.

For the above reasons, the KDE community has decided to not do releases of KDE 3 anymore. Technologically, it’s a dead end since many architectural, structural and environmental aspects of KDE 3 couldn’t be solved in KDE 3. If we just kept releasing KDE 3, we wouldn’t be able to assure sufficient quality and enough support so people wouldn’t be left out in the rain. Or put in other words, it would give KDE a bad name. The KDE3 branch in our SVN source code repository remains frozen for new features, bugfix patches are allowed but make very little sense, since there’s no plan to release new versions of KDE 3.x. That’s why the Trinity developers are using a work branch.

Those are challenges the Trinity team is facing now. Some of them can be solved by pouring enough manpower, tender love and care into it. Others aren’t easily solvable (think IPC) within KDE 3. Whether or not the Trinity team will succeed in maintaining and developing KDE 3 remains to be seen, but it’s certainly not an easy task. Then again, easy tasks are boring, and the Trinity team is validating the Free Software development and licensing model. In any event, the KDE community supports the efforts (for example by offering infrastructure in the form of source code repositories), and would like to see Trinity as another blooming subproject inside the KDE community.

New challenges.

I’ve resigned my job at KDAB last month in a swift move towards more KDE-time. This all came pretty suddenly, but it felt like The Right Thing to do for me personally and for KDE, which I care a lot about. Since May, I’ve been working for Open-SLX, a German company that makes and supports the openSUSE boxed version. My focus in that work is the user experience of the product. The idea is to work upstream (in openSUSE and KDE / Plasma) as much as possible. While Open-SLX benefits directly from my work done in KDE, this is also a nice way to give back to the community, by making sure I get to spend enough time on things that are not directly related to the product. So now I’ve settled into my new job, and up until now, it’s been great. I’ve been able to catch up with a couple of areas in KDE, I didn’t get to spend as much time as I wanted in the past, and I have started working on some ideas I was dragging around in the back of my brain for a while). One of those things is Project Silk, which is a Project to boost and deeply integrate the web into KDE Plasma and applications. Its motto is no less ambitions than "Freeing the Web From the Browser", so there’s lots of work to do. ;-) Others have already shown off their cool creations, so I’ve got some catching up to do. I’ll share more detailed information about Silk in the next weeks, so if you’re interested in that, hang on just a little bit longer.

With this new job, I’m also able to spend a bit more time on KDE e.V. things. I’m a Board Member for some time already. Being able to sneak in a bit more of that structured desk time for things that need doing in the near future is surely a good thing. Regarding the e.V., I’ll travel with Ade to Berlin on Friday to meet Celeste, Cornelius and Frank there for an extended weekend of board work (and fun).

Blog moved.

I’ve moved my blog (and the rest of my personal website, vizZzion.org to a new host. Since we were having problems lately with the stability (in fact the disks’ stability) of the machine running api.kde.org, our online source code documentation, I’ve decided to move my blog to a virtual machine on slicehost.com, so we’d get rid of another detail in the setup of the disks on the machine running api.kde.org. Setting up the server at slicehost was painless, within 10 minutes after registering, I had a clean-slate virtual machine running Linux, and a bit later, after making myself comfortable — mostly setting up zsh and my usual set of aliases and bells and whistles in place, I installed and configured apache, mysql, migrated my data and was good to go. The DNS change has now trickled down into the DNS caches of the world.

More Plasma Network Management Updates.

Since there were quite some changes in the Network Management machinery in KDE Plasma over the last days, I thought I’d sum them up here, as to update the brave that already use the as of now unreleased Plasmoid.

  • Plugin name changed – We’ve changed the name of the networkmanagement plasmoid plugin to from "networkmanagement" to "org.kde.networkmanagement" in order to avoid possible future naming conflicts. This means that you might need to re-add the NM widget to your notification area (do that using the config dialog and checking the Network Management plasmoid)
  • SVN switched from KNM to the Plasma Widget – I’ve just committed a set of changes that enable the NM Plasma widget by default, this means no more pesky manual loading of the kded service module. It also means that the autostart file for knetworkmanager is not installed anymore, since we’re using the Plasmoid now. If you’re compiling from SVN trunk, you can use cmake -DINSTALL_KNM_AUTOSTART=ON to get it back. Make sure you prevent the kded module from autoloading (or just unload it), if you want to go with knm (Alternatively, tell us why you prefer knm above the Plasmoid and we’ll try to fix it).
  • New artwork – As you can see, we have been working hard on improving the visuals of the NM widget. While it’s not perfect yet, it’s definitely getting better in terms of consistency, layout and overall visual appearance. With a bit of help from notmart, the Plasma NM widget in trunk now uses artwork from the Air theme, which is installed by kdebase. For backwards compatibility (I like all those people testing on 4.4!), we’ve chosen to install the artwork for "older" versions (last week and before), but we might be lax updating it (as it happens with copies of files in source trees).
  • New souls – Over the past days, two new hackers have started sending patches for our network management machinery. One is Lamarque Vieira Souza, who has been working on modemmanager integration, that stuff that gets you on the Internet when relying on 3G, HDSPA and the like. The other one is Andreas Demmer, who has been continously testing the Plasmoid over the last months, already contributed some artwork, and today finally got his SVN account and started committing fixes. Welcome, guys!
  • Many others – There were also many other changes, some fixing bugs, others small layout improvments, those will keep coming.

As you can see, we’re on the home stretch for a first release, fixing up all kinds of small issues, testing, reviewing things. After that, we’ll move the NM Plasmoid to KDE Extragear and release it from there. Before or with 4.5. Promise.

Blog back online.

Thanks to my unrestful ex-colleague and current coffee-buddy Ade, my blog is back online, as you can see. Now I have to figure out why Blogilo works fine on one machine, but has problems getting my blog’s metadata on another one (my laptop). Blogilo is another one of those applications from the category Internet I’d rather not live without. Writing blogs offline (on trains, for example), adding media (screenshots for example, those seem to be quite popular) to blogs is just so much easier if you don’t have to deal with the uploading yourself but get it done in the background. And now I’ve mentioned screenshots, I should probably also show one (since it’s that easy). I’ll pick the new interface details in the Network Management plasmoid. In the screenshot, you can see what you get when you click on a network interface, it’ll show you additional information about the interface, and if connected some basic traffic stats. For this, I’ve used the systemmonitor dataengine and the Plasma::SignalPlotter widget, so the patch to add this nice little feature weighed in at only about 50 lines of code. The widget only updates when the details are shown to save power cycles when it’s not in use, but you can switch to the details and then dismiss the popup (which happens automatically with our focus policy), and it will keep collecting data. If you switch back to the normal view (using the back button), it’ll stop updating. I’m also happy to have received Andreas Demmer’s first patch today, which looked good right away and has been merged already. He changed the back button in the details widget to be more consistent with other buttons, so that’s a nice addition. Andreas has been keeping an eye on the development of the networkmanagement plasmoid, and has been providing useful feedback in the process. So there’s another step on the ladder towards hacker-heaven. (Yes, Plasma development is *that* cool. :-))

Another thing that plays a rather important role in my daily workflow is the Quassel IRC Client, which is an IRC client that allows you to easily use different machines for your IRC needs, without the need of logging in and out all the time and losing the history of your conversations. So if you found my IRC presence to be wonky during the last days, it should be better now.

I hear you asking "The real relevance of this post is…?" Well, api.kde.org is also back online, so your Konqueror shortcut "kde:<classname>" works again — although I’d expect that by now, every sensible hacker has loaded this file into Qt Assistant (Edit | Preferences | Documentation | Add…) and have both Qt and KDE docs available in one place.

Tokamak.finished()

Last night was the last night of Tokamak 4, our semi-annual Plasma sprint for me, this morning I boarded a train to Düsseldorf (Germany), where I’m staying for another two nights with K.

Tokamak 4 was again a total blast. The Suse/Novell offices in Nürnberg were the perfect place to get a (for a sprint, anyways) rather large group together to hack day and night. I think I averaged about 6 hours of sleep, not bad. On the other hand, I didn’t get as much done as I wanted, but got a number of other things to take care of, so that’s OK.

You’re probably most interested in the progress of the network management Plasma widget, so I’ll explain a bit what we’ve been working on. We is Will Stephenson and Alex Naumov from openSUSE, Aurelien Gateau from Kubuntu (who took part remotely), Martin Zilz of the Oxygen team and me. The goal of the network management plasmoid is three-fold: it should be very easy to get your network set up, it should be technically robust and flexible, and it should also look good doing all that. We started of with fixing some bugs, Will got a bug in our DBus service fixed which would put bogus connection items on the bus, which then showed up in the list of offered connections. Aurelien took care of polishing the user interfaces of many of the dialogues the network management infrastructure uses, for example for editing connections. He also nailed a bug which would pop up two connection details dialogues when trying to connect to a wireless network.

We also did some design work, in order to determine which functionality we’ll offer, and how. I think the result of that is really nice, but not finished yet, so I won’t bother showing a screenshot. The design session we held showed one thing rather nicely, all the ideas we had fit very well into the current, two-paned UI concept without adding additional clutter, we’ll even be able to make the UI less complex and offering more functionality. I think that this is really the Plasma way, and in extension KDE 4 way of designing. Offer great functionality in a smart way, do not clutter the UI but create streamlined workflows.

So, what’s coming? Well, first off the networkmanagement plasmoid supports of course all different kinds of networks, wireless, VPN already. We haven’t yet tested mobile broadband, but are definitely planning to also make this a first-class citizen. You’ll also be able to see detailed information about your connections (no matter which way you choose to connect). You can choose to "just connect", but also to use a specific interface for a connection, so people with multiple network adaptors of the same type get full control.

I’m especially pleased that Martin, a.k.a. mofux jumped in to provide us with some kick-ass new artwork for the wireless icons. The artwork we had been using up until last week had pretty bad contrast issues with the Air theme, the new artwork is much better in that regard, and also a lot clearer and easier on the eye at small sizes. We’ve also added theming support here (through Plasma’s SVG theming mechanics), which is a nice bonus.

On particular pet peeve of mine is the connection status. I want to see how long I still have to wait until my network is up and I can do all those cool Internet things. Tooltips aren’t very well suited for this, and they have the tendency to go away just before I want them to. Also opening the connection dialogue is not good enough for me, since it involves manual interaction. I want to resume my notebook and see at once how far along my connection attempt is. What I have right now is I think already pretty close (though only 90% there from the beauty point of view). The idea is to show the progress as an overlay in the network management icon in the panel’s notification area. It’s now implemented as a pie chart, that starts at 12 o’clock and completes as the connection goes through its different stages, like connection at 802.11 level, authenticating (you’ll get a nice hint that a password is required in the panel as well) and configuring the IP address. For not so technical users, this is readable since it re-uses concepts well known from the real world, for the more technical folks (I tend to refer to them as propellerheads ;)), it provides more information than just a busy-spinner. As an example, if it "hangs" at 7 o’clock, you know that the DHCP config is taking its time. This is something that has been bugging me in other wifi configuration tools forever, so hurray, another step taken towards the best connection management tool in the world. :-)

So, is it finished yet? Well, no. It’s in a better shape than last week — which is not always a given, because sprints are often used for larger redesigns, and indeed we did improve some UI concepts that required parts of the code to be changed. Most importantly though, we’ve been making great progress at hard-to-fix bugs, and what’s left now is tying up a couple of loose ends. I’ve been hacking away on my list of changes to the plasmoid itself and I’m pleased with the progress so far. It’s not done yet though, but I expect that within a couple of days’ hacking, we’ll get there. I don’t see any particular blockers from my side right now, which is good. I’m confident that we can release a beta version shortly and give it some time to stabilise until it’s officially released as part of KDE SC 4.5.