>
>> > Revision 1.34 of cats/conf/INSTALL seems to have quietly put the cat=
s
>> > install kernel back to using tlp0 again, which still isn't finding i=
ts
>> MII
>> > properly (for unknown reasons).
>> >
>> > Is there any reason why this shouldn't be restored to the functional
>> de0
>> > driver?

Hmm, someone reversed that then, I had changed it in the install kernel o=
n
purpose.  Looking at CVS I think nick changed it accidently.  I had
planned to start testing 2.0 cats in the next few days, so would have
spotted it soon.

>> Is the de driver really more functional in general, or just for your
>> card? The tlp driver has been getting a lot more care and feeding than
>> the de driver in the last, oh, four years.

But it's still not working for some cards.  I've no idea why, I'm not a
pci expert or debugger, all I know is that the de card in my cats works
fine with tlp on i386, but not on cats, I got told port-specific bug, and
to go fix it.  I asked for advice/help on where to start, but I didn't ge=
t
any.

cats is inconsistent, it uses de for INSTALL and tlp for GENERIC, I
couldn't decide on the right thing, perhaps an INSTALL.de and GENERIC.de?=
=20
with the main ones using tlp?

>> As I mentioned last year when you brought this up, "tlp works for me",
>> both with and without MII:
>>
> The tlp driver doesn't find its MII, consequently the machine will pani=
c
> as soon as you try to do anything with the driver (such as run 'ifconfi=
g
> -a').  That's fatal and means you can't install anything over the netwo=
rk.
>
> The de driver may have its faults, but at least it doesn't bring the
> machine down the first time you try to use it.

Indeed, this is rather fatal.

Chris