Hostap_pci vs. orinoco_pci

[Jar]

> The drivers only list devices they support.  The drivers is where this
> information is stored for the kernel and userspace utilities.  Drivers
> don't scan buses and don't choose whether to support a particular device
> or to leave it to another driver.
>
> Not having some supported devices listed in the driver would be a bigger
> problem, because if the driver doesn't know whether it supports a
> device, how can more generic utilities know that?

Thanks for this clarification! I have always (wrongly) believed that the driver is
guilty for this. According to you there is nothing wrong even if there is 99
different driver candidates for one single device, right?

> You should talk to hotplug/udev developers about it.  The mailing list
> is at linux-hotplug-devel at lists.sourceforge.net

Seems bug to me or at least something important has been forgotten when implementing
hotplug/udev. I am not familiar with hotplug/udev, but sure the developers know the
problem?

> Blacklisting the driver doesn't look like a flexible solution.  You need
> a way to "blacklist" a driver for a device, or better yet, "endorse" a
> driver for a device to make sure that no other driver is allowed to use
> it.

Blacklisting is better than nothing, now we have nothing. And it really sounds bad
when someone says "delete a module".

> By the way, orinoco_pci only support Prism based cards.

Is there some particular reason to keep the orinoco_pci now when we have hostap_pci?

-- 
Best Regards, Jar