[Prism54-devel] SIOCGIWNAME usage

[Luis R. Rodriguez]

On Wed, 7 Jan 2004, Jean Tourrilhes wrote:

> Luis R. Rodriguez  wrote :
> >
> > Since we already have cardmodel identification implemented, we can still
> > keep it and perhaps use it for verbosity at card probe time. I'll ask
> > Jean to see what he thinks about integrating specific card model
> > detection/displaying for the API and/or what to do in cases where IEEE
> > data transfer specifications can alter for cards.
>
> 	That must be me ;-)
> 	The card model is not wireless specific, for example Ethernet
> card also have model name, so it should be handled at the level of a
> more generic API, such as lspci or Ethtool. Actually, "ethtool -i"
> almost has what you need.

Agreed and understood. Thanks :D

> 	The intent of SIOCGIWNAME today is to show the capacity of the
> card, so that a application knows what it can do with it. That should
> be fairly static. Listing the capacity of the driver or the current
> mode of the card are good approximation.
> 	The idea there is that if two users want to know if there
> respective card are compatible, they would look at this value and
> check if they have some letter in common.

I've added this. Our iwconfig output is now in accordance to this.

> 	You may want to check quickly what madwifi decided to do here.

I can't get to sf at the moment. Glad our project is not hosted there ;)

> 	With respect to SIOCSIWNICKN : This IOCTL should probably
> never been added to the API, but kept as private, as it has zero
> impact on the protocol. But, so many of the 802.11b cards were
> implementing it... So, don't feel too bad and allow yourself to abuse
> it as you wish to return useful information. The TI-ACX100 already use
> it to display the driver version, but displaying the card model number
> is a better idea.

OK perhaps we will do this. For now I'll just display it at probe time or
so for dmesg output.

Thanks for the reply,

	Luis