Day 2 Notes, LinuxCon 2010

Wednesday August 12, 2010
Boston, MA

Prepared by Alison Chaiken. offered under cc-by-sa

"MeeGo: Where are We Now"

Dawn Foster, MeeGo Community Manager
Intel Corporation

Foster described the merger of the Maemo and Moblin projects into MeeGo as a logical joining of efforts with similar objectives. Foster explained that MeeGo will feature netbook, handset and in-vehicle infotainment (IVI) "user experiences" plus others that haven't been considered yet. Central to all the UE is a common core of code that is the heart of MeeGo. The intention is that developers can use the core to create applications and then easily deploy to all the UEs. Porting apps from the Linux desktop is also intended to be straightforward since MeeGo is rather straightforward GNU/Linux, unlike (for example) Android. The netbook UE is already available for download and the "in-vehicle infotainment" one has just been published. While MeeGo can be installed on Nokia N900 handsets, the combination is considered rather experimental. MeeGo-based handsets should be shipping by Christmas.

Foster showed a rather impressive demo of MeeGo on a Toshiba Atom-powered netbook. MeeGo is pretty and appears to be both stable and snappy. The boot time of the Toshiba netbook to the MeeGo desktop was fairly impressive -- maybe about 10 seconds from power-on. The home screen is arranged as a window with tabs representing different kinds of activities, and navigation is intuitive. At one point, about a dozen error messages popped up about WiFi dropping. The dropped network is hardly surprising in the context of a technical conference, but the profusion of duplicative messages was more concerning. MeeGo is a bit cute for hard-core Linux users but may well appeal to people who want to use a web browser and answer email. It's a shame that MeeGo did not arrive in time to ride the wave of netbook sales growth that occurred a couple of years ago. Watching Foster's netbook-based demo, I'm curious to see how usable the handset "user experience" will be.

"Freedom to Innovate: Can MeeGo's Openness Change the Mobile Industry?"

Tom Miller, Nokia
Dean Speed, Intel

Intel's AppUp store is for libraries and other middleware as well as full-blown applications. OEMs can rebrand AppUp and host on Intel's platform if they wish. Nokia's Ovi website is more of a traditional app store.

The Qt project is now hosted on gitorious and is fully open source. A port of Qt to Android is underway.

An audience member asked about the situation with DRMed media content on media and Miller answered that music buyers would not be restricted to Nokia's music store. Clearly the questioner was hoping to get a higher-level answer about the situation with proprietary codecs on MeeGo.

An audience member asked if other CPU vendors and handset manufacturers can now ship MeeGo. The two representatives agreed that there were no barriers to other vendors competing on MeeGo. A follow-up question about locking of MeeGo devices was not answered, implying that at least some of them will in fact be locked.

"Open Development of Medical Devices"

Karen Sandler, Software Freedom Law Center

The fact that the firmware of medical devices is not only secret but untested by the FDA is appalling. The case of voting systems has illustrated how important the openness of societally critical software is. FDA has no staff with expertise to evaluate firmware and has promulgated no standards with regard to it. Furthermore, FDA jurisdiction effectively means that patients who are injured by defective but approved devices are severely constrained in their right to sue. Before getting her own medical implant, Sandler carefully considered the firmware safety of the various types of devices before making her choice. To get even the limited information she received, Sandler had to contact the manufacturers directly, and even there she was largely shut out.

"Saving Energy with the intel_idle cpuidle driver"

Len Brown, Intel

Slides

Brown, following Wysocki and Garrett, presented the third in a series of excellent talks about recent power management developments in the Linux kernel. Brown's work has focussed on successive idle states of a laptop CPU (C states and P states) as opposed to Wysocki's description of mainline's approach to peripheral device power management and Garrett's thoughts about handset power states and the divergent Android approach.

Power management is an area where Linux has recently been innovative and is just plain outperforming Windows.

C and P states are specified in the ACPI protocol, whose purpose is to make Linux work with Windows-intended hardware. The names of the states vary a bit by architecture. Not all states are supported on all architectures. Poorly designed criteria for entering and leaving the states can actually waste power, as the transitions cost energy. Getting out of P states (where devices and buses are powered off) is more expensive than getting out of C states, where just the processor is partially or fully powered down.

Intel's power management software never promotes the system to a deeper sleep than is explicitly requested, while AMD's sometimes does.

When Brown's improvements to cpu_idle were first complete, the first Dell product to ship with the new software had a Windows-motivated BIOS bug which prevented the kernel from taking advantage of the new features. The BIOS soon had a patch, but of course very few consumers update firmware. The situation was frustrating to Brown, to say the least!

Wysocki's newish PM_QOS allows dynamically loaded drivers to prevent sleeps. BIOS typically forces changes in power state settings up transition from mains to battery power. Using PM_QOS is more flexible when control of the transitions can be wrestled away from the BIOS.

cpu_freq for dynamically changing the processor clock has a huge table of parameters. By comparison, intel_idle is relatively simple. Users can check dmesg for an mwait statement to see if intel_idle is ready to run.

Despite difficulties that lack of control over BIOS presents, coreboot project is not a solution, as, according to Brown, it "will never work with commercial machines," only with embedded systems.

PCIe has better power-saving potential than PCI. Apple does an excellent job in power-saving on laptops since they control the whole stack. Fedora 13 often doesn't suspend when Brown expects it to; he hasn't yet discovered why.

turbostat is a useful diagnostic tool for viewing per-core C states, available as part of pmtools on kernel.org. powertop is a similar tool but tends to be buggier. The goal of cpu_idle is to spend as much time in the higher P states (PC3 and PC6) with peripherals turned off as possible.

"Doing What it Takes: Current Legal Issues in Defending FOSS"

Eben Moglen, Software Freedom Law Center

Moglen believes that Linux is well-established as a viable business and argues that it is now clear that commercial and "sharing" models of commerce can peacefully coexist. He emphasized that he personally is for capitalism if not of it.

The threat to Linux from copyright attacks is now largely passed. Unfortunately, after the Bilski decision, it's clear that software patents are with us indefinitely, and they continue to be a threat.

Back to Top