Today was quite a day! I attended many interesting sessions and it’s going to be quite difficult to keep this post to a reasonable length. I make no promises that I will succeed at that. Here’s a set of highlights, mostly my stream-of-consciousness notes from the sessions I attended. I’ll pare down the best details in a final wrap-up post on Saturday or Sunday. For now, consider this an exhaustively detailed set of live notes published as a digest.
This has gotten quite long as I write, so I’m going to break these up into morning and afternoon reports instead. Enjoy!
Free and Open Source Software in Europe: Policies and Implementations
The first talk I attended today was the keynote by Gijs Hellenius, Free and Open Source Software in Europe: Policies and Implementations. It was an excellent overview of the successes and issues surrounding the use of Open Source software in the European public sector. The good news was that more and more politicians are identifying the value of Open Source, providing lower costs and avoidance of lock-in. The bad news is that extensive lobbying by large proprietary companies has done a lot of damage, thus preventing more widespread adoption.
That is all in terms of usage. Unfortunately, the situation soured somewhat because developers of public sector code are hesitant about making their code open-source (or contributing to other open-source projects directly). It was noted that the legal questions have all been answered (and plenty of examples of open-sourced public development exist). Unfortunately, a culture of conservatism exists and has slowed this advance.
A highlight of the talk was the Top three most visible open source implementations section. The first of these was the French Gendarmie who switched 72,000 systems over to Ubuntu Linux and LibreOffice Desktops. While this author would certainly prefer that they had selected an RPM-based distribution, I have to bow to the reported 80% cost reduction.
The second major win was the conversion of 32,000 PCs in government and healthcare in the Spain Extremadura areas (mostly Debian).
The third one was the City Administration of Munich, who also converted 14,800 workstations from Windows to Ubuntu Linux and LibreOffice. Excellent quote on his slide (paraphrased because it went by too quickly): “Bill Gates: Why are you doing this, Mr. Ude?” “Ude: To gain freedom.” “Gates: Freedom from what?” “Ude: From you, Mr. Gates.”
Fedora QA – What Why How…?
The second talk I attended today was an introduction to Fedora QA from Amita Sharma, a Red Hat senior quality engineer.
Amita started by talking about why Fedora QA is important, citing that Fedora’s fast schedule would otherwise lead to a very buggy and unpleasant system to use. She showed us a few bugs in Bugzilla to explain how easy it can be to participate in QA. In the simplest sense, it’s no more than spotting a bug and reporting it.
Next she discussed a bit about the team composition, meeting schedule and duties. This involved things like update and release testing, creation of automation and validation testing, filing bugs, weekly meetings and scheduling and execution of Fedora Test Days.
Some of the interesting bits that Amita pointed out were some of the various available resources for Fedora QA. She made sure to call out the daily Rawhide Report emails to the email@example.com mailing list, as well as Adam Williamson’s Rawhide Watch blog and the regular Security Update test notice.
She next launched into a detailed explanation of the karma process for Fedora testing updates. She explained how it is used to gate updates to the stable repository until users or QA testers validate that it works properly.
Amita moved on to describing the bug workflow process, from how to identify a bug to reporting it on Bugzilla and following it through to resolution.
The next phase of Amita’s talk focused on the planning, organization and execution of a Fedora Test Day, from proposing a Test Day to creating test cases and managing the process on IRC.
I was originally planning to attend the Taskotron talk by Tim Flink, but the room was over capacity, so I went over to the Wayland talk being given by Matthias Clasen (which was my close second choice).
I arrived a few minutes late, so I missed the very beginning and it took me a bit of time to catch up. Matthias was describing Wayland’s capabilities and design decisions at a very low level, in many cases over my head, when I came in, so I really regretted missing the beginning.
One of the fundamental principles of Wayland is that clients are fully isolated. There’s no root window and there’s no concept of where your window is positioned relative to other windows, no “grabs” of input and no way for one Wayland application’s session to interfere directly with another. (Author’s note: this is a key security feature and a major win over X11). Wayland will allow certain clients to be privileged for special purposes such as taking screenshots and accessibility tools. (Author’s note: I had a conversation earlier in the day with Hans de Goede where we talked about implementing Android-style “intents” for such privileged access so that no application can perform display-server-wide actions without express user permission. I think this is a great idea and we need to work out with user-experience designers how best to pull this off.)
Another interesting piece that Wayland will offer is multiple interfaces. For example, it will be much easier to implement a multi-seat setup with Wayland than it was with X11 historically.
Matthias then moved on to describe why we would want to replace X11. The biggest reason is to clean out a lot of cruft that has accumulated in the X11 protocol over the years (much of it unused in modern systems). Having both the compositor and renderer in the same engine also means that you can avoid a lot of excess and duplicated work in each of the window managers.
One very important piece is the availability of sandboxing. In X11, any application talking to the server has access to the entire server. This potentially means that any compromised X application can read anything made available to X from any other application. With Wayland, it becomes very easy to isolate the display server information between processes, which makes it much more difficult to escalate privileges.
Matthias also covered some of the concerns (or arguable disadvantages), the biggest of which was that compositor crashes will now destroy the display session, whereas on older X11 approaches it was possible to restart the window manager and retain the display. Other issues involved driver support; for example the nVidia driver does not support Wayland. (Note: the Nouveau driver works just fine, it’s the proprietary driver that does not.)
The next topic was GNOME support of Wayland, which has come a very long way in the last two years. GNOME 3.10 and later has experimental support for running on Wayland. The remaining gaps on Fedora 21 are mainly drag-and-drop support, input configuration and WACOM support. (Author’s note: I’ve played with GNOME 3.13.3 running atop Wayland. Much of the expected functionality works stably, but I wouldn’t yet recommend it for daily use. Given the rate of improvement I’ve seen, this recommendation may change at F21 GA release).
Matthias made a rather bold statement that the goal to hit before making Wayland the default in Fedora is that end-users should not notice any difference (Author’s note: presumably negative difference) in their desktop.