Merging with Linus's kernel

[Roman Kagan]

On Fri, Apr 01, 2005 at 12:32:20PM +0200, Duncan Sands wrote:
> (2) the new setup requires that the ATM layer generate hotplug events.

Umm, not exactly "requires"...  Rather the opposite: it makes it easier
for the ATM layer to generate hotplug events at the right time.

Basically, the difference of the CVS version and the in-tree version is
that with the former you have to wait until the associated ATM device
appears, while with the latter it appears immediately, but you have to
wait until it's usable (otherwise opening it will give -ENODEV).  Given
that people don't seem to have any problem with the latter (presumably
using sleep in their scripts), I would expect even less with the former.

> Roman posted a patch, but I don't know what happened to that.

I'm not yet used to the modus operandi of Chas Williams, so neither do I
know...

> (3) a small patch will need to be applied to modem_run, for those weird
> people who don't want to use the in-kernel firmware loader.

Given that we're talking about 2.6 only, I believe you can pretty much
rely upon hotplug firmware agent being available, and the recently
discussed "select FW_LOADER" makes sure that the kernel has it too.  So
I'm not sure if continuing to support modem_run is justified.

Cheers,
  Roman.