[Battlemesh] What hardware still works?

[Daniel Golle]

Hi Jonathan,

On Thu, Mar 10, 2016 at 02:32:03AM +0200, Jonathan Morton wrote:
> 
> > On 7 Mar, 2016, at 15:24, Daniel Golle <daniel at makrotopia.org> wrote:
> > 
> > Apart from that, I'm not sure if modularization will truly solve the
> > problem as it would still require that devices are assembled by
> > end-users (and not re-sold as a whole) as well as suitable hardware
> > components (especially SoftMAC WiFi mPCIe modules) being available...
> 
> That’s not what happens with laptops or all-in-one PCs.  The last time I found a laptop where I'd have to install the wifi card myself was over a decade ago, and ath9k chipsets are still appearing in 802.11n devices.

Yes, ath9k does, finding good low-cost modules with ath9k has become
more difficult.

> They have to be relying on the FCC certification of the module for intentional-emission compliance, so they only have to do the normal unintentional-emission tests on the complete laptop.  And laptops are *fundamentally* devices on which you can replace the entire software stack on a whim.  They just wouldn’t be useful as PCs if you couldn’t.  That’s why I’m focusing on laptops as the primary use-case that the FCC has to preserve intact.

The easiest way to comply is using a FullMAC or HardMAC which does some
DRM on the wifi module itself -- then OS and driver doesn't matter...

> The Raspberry Pi 3 also makes an interesting case.  The Broadcom chipset it uses is a FullMAC, attached via SDIO.  It’s really not a high-performance solution, but in the volumes Pis sell at, it could be added without increasing the price.  And its approach to spectrum allocation is interesting - it will use channels 12-14 if it can hear other wifi devices already using them, but remain silent on them otherwise.

Devboards are generally excluded from certification, at least in the
EU. brcmfmac is as ugly as it can get for community networks, though...

> This is a sensible strategy for a client device, but it does mean that if a Japanese student brings their wifi router with them to the US, suddenly lots of devices will think that transmitting on channel 14 is okay, making them non-compliant with US regulations.

Funny... When I last saw brcmfmac, they had a text-file loaded from
/lib/firmware for EEPROM-less devices, containing the rules of the
intended target market...
For device in the 5 GHz band, the story looks a way more complicated,
as the allowed bands in different regulatory domains are disjunct and
additional requirements may apply.
Even for the 2.4GHz band this is bad for community networks in eg. the
EU, which often use CH 13 exactly because noone else does...
So things like the RasbPi 3 can participate, if surrounded by an
existing setup somewhere on 2.4 GHz, but not more than that.


Cheers


Daniel