Re: [Ipw2100-devel] For James : WE-19 preview

[Jean T. <jt...@hp...>]

On Tue, Mar 29, 2005 at 05:38:32PM +0200, Marcel Holtmann wrote:
> Hi Jean,
> 
> > > > 	Some of the ideas I'm toying around is to add support for
> > > > 802.15.4 (aka ZigBee) in Wireless Extensions.
> > > 
> > > I also looked at the ZigBee stuff. Do you think that the WE are the
> > > right place for it?
> > 
> > 	I'm obviously biased about this subject. 802.15.4 is a simple
> > CSMA/CA MAC with retry, the Phy is DS with various channels. I expect
> > the usual setting of frequency, retry, tx-power, sensitivity,
> > etc... There is simple symetrical encryption. There is the concept of
> > cells with a PANId and Scanning. All that sounds very familiar.
> > 	Now, on top of 802.15.4, there is all the ZigBee stack. This
> > is currently closed, so I have not looked at it, but it may look more
> > like BlueTooth, with service discovery and similar high level
> > stuff. Anyway, at this point it look like most people go for 802.15.4
> > without the ZigBee stack.
> 
> I tried to convince the ZigBee Association to open their protocol
> specification, but then the communication got stuck.

	They have already opened 802.15.4, which is already a good
thing. They are making money out of selling the spec, so don't expect
them to change course.

> The linux-802-15-4.sf.net project is using character devices for the
> interface and this is definitive wrong from my view.

	There was the same debate at the very beggining of
Linux-IrDA. Currently, all 802.15.4 chipset have a SPI interface, so
if you just want to expose the raw chip, serial make sense. But, if
you want to have a generic/uniform driver interface, then I would
expect a net model might make more sense.
	Remember, everything start as a quick hack before generic APIs
are put in place. Remember OpenBT ?

> Do you have any
> concret plans for IEEE 802.15.4 support in WE?

	Not yet. Just an intent.

> IEEE 802.15.4 without ZigBee sounds like a good option.

	There are pros and cons, it depend what you want to do.

> > > Do it really fit into the network card model?
> > 
> > 	WE is wireless configuration, and doesn't assume a network
> > model. To me, 802.15.4 look very much like a network, but obviously
> > the higher layer could change that.
> 
> I must admit that the last time I read the specification is quite some
> time ago and so far I haven't got any chips for testing.

	Most chips are SPI, and are quite available. End-user dongles
are unfortunately not.

> Regards
> 
> Marcel

	Jean