On Mon, 14 Oct 2002 kuznet@xxxxxxxxxxxxx wrote:
> Hello!
>
> > Actually the extra flags are only valid when IFF_RUNNING is not set.
>
> Even so...
>
> Well, then I am inclined not to agree to give even one of those valuable
> 16 spare bits for this. :-)
>
> I cannot imagine how much should I drink to consider states descibed
> in this rfc as a valid abstraction. Device can be working and can be dead
> by thousands of reasons. The best which I can propose is to show a string
> somewhere in /proc (well, or as a _string_ attribute in RTM_NEWLINK),
> explaining why device is not alive and let snmpd to translate this string
> to these bogus states to generate traps.
;-> Believe it or not people use these things to draw nice GUI
representations (search for lowerLayerDown at cisco for example);
how we represent them shouldnt matter whether its via ifi_flags, /proc
etc. Infact it should also be fine if we dont propagate them to user space
for now, but the abstraction should stay in the kernel at least.
cheers,
jamal
|