Weekly Engineering News 36/2008

[Wolfgang Spraul]

Hi everybody,
last week had two important new developments in Taipei:

1. Samsung 6410
2. Marvell 8688

We decided to start working on a new hardware design using the Samsung  
6410 SoC. Samsung's documentation is unfortunately closed, which puts  
them at a disadvantage compared to TI or Freescale, for example.  
However, they are releasing a Linux driver set which we will forward- 
port to the latest mainline Linux kernel and make more mainline  
compatible. Also, we will use our very supportive contacts at Samsung  
to see whether we can convince them to open up their documentation.  
Staying with Samsung allows us to reuse more of our current Linux  
driver work, for example the PMU (NXP 50633) can stay the same.

Similar with Marvell - we decided to seriously evaluate Marvell's Wi- 
Fi and Bluetooth chips, specifically the Marvel 8688. As with Samsung,  
documentation is currently closed, but we got high-level support at  
Marvell telling us they are actively working on opening up their  
documentation, and GPL drivers are available from Marvell or already  
in the mainline. Our work with Atheros will continue, we want to  
remove several abstraction layers from the driver Atheros released, or  
switch to the mainline driver right away. We will also evaluate  
Atheros' 6002 chip and are very satisfied with the engineering support  
we get from Atheros. Some big Openmoko fans over there :-)
The reason we want to work with both Atheros and Marvell is that we  
have so many plans around Wi-Fi that we cannot afford to not consider  
both options. In the long run we are even thinking (well, it's more  
'dreaming') about designing our own Wi-Fi baseband chip, but give us a  
few years for that...

I want to add a disclaimer here: What I am talking about are  
engineering activities. They may or may not translate into products,  
whichever products. Speculating about GTA03, GTA04 or some totally  
different acronyms won't help and I won't participate. A 'design and  
development' process is naturally full of surprises, and marketing has  
the upper hand still when it comes to products.
Any product we are making, we want to be the best in kernel.org (and  
other upstream) compatibility. Ideally 100% of the product's features  
should be available through the mainline. That's our goal, for  
existing and future products.

Best Regards,
Wolfgang