Linux Wireless Summit II

Tux
Earlier this month the Linux Wireless Summit II, following Summit I from 2006, took place in London. Main topics were the current state of Linux’ new WLAN stack as well as the reasons for the still hesitant adoption of Linux by hardware vendors.

First the background

The WLAN situation on Linux sucks – at least from a technical point of view. Every driver has its own subsystem and is integrated into the kernel on its own. There is no common ground. That’s dumb and makes it quite difficult to import new drivers because first of all a new driver developer would have to decide which of the current possibilities of the other drivers he adopts for his own driver. Also, a specific driver which sucks even more is the Intel 3945 driver, which needs a proprietary user space daemon. So far the current state.

Quite some time ago the developers decided to change that, and looked for a new subsystem. Suddenly, the company Devicescape released their WLAN stack as GPL – and the developers decided after some discussion to integrate that stack as the new bay for all WLAN drivers in Linux. That was around April 2006.
Until now, the users haven’t seen anything new about that.

The summit

The summit took place in London, and I have the feeling the news didn’t really got around about that. Anyway, Stephen Hemminger and Daniel Drake have two detailed summaries of the summit.

The first big issue was why the vendors do not pick up on Linux. The main problem seems to be here that the FCC has rules that aren’t specific, but imply that you might need a proprietary part in your drivers. The idea behind it that you should not be able to change the radio frequencies of your device. And since the kernel needs open drivers, that becomes a problem. The Freedom Law Center was only able to add that the regulations are not specific enough and that someone has to ask them explicitly. However, that takes time.

The other big thing was the WLAN stack. As already mentioned, the stack from Devicescape is there, but still not integrated into the kernel mainline. And it looks like it will take some more time, possibly until 2.6.22, till it will be merged into the main tree. Atm there are still locking problems, also there should be more developers since all the drivers have to be ported. Additionally, there have to be configuration utilities ready, and they are not yet in a releasable state. But they are on their way.
One big advantage for the new stack is clearly that Intel decided to merge drivers into that new stack – starting with the mean 3945. And with the new stack Intel is finally able to provide a driver which only needs some microcode disk – and therefore no daemon or stuff like that. That’s a huge improvement over the current situation. You can get more information on that topic on Intel’s new Linux-WLAN website: intellinuxwireless.org.

And, last but not least, the WLAN Linux team got their own, polished homepage: linuxwireless.org/. It didn’t really start off at the moment, but it shows that the people really do care for all involved people. I’m very happy that this page is now up because it will be a central point for WLAN hardware vendors, developers and users at the same time: such a central place has the possibility to power up all WLAN related things around Linux. And if it merges other, similiar pages as well, even better.
And, one new idea you can already see on that page: certifications. It was mentioned that there should be some “Linux Certified” certificate available in the future. Interesting times!