Le 18/02/2019 à 20:04, J. Fahrner via Dng a écrit :
> Am 2019-02-18 18:45, schrieb Didier Kryn:
>>     If you reject the software because you distrust the author, try
>> netplug intead, or write your own. The function is the same: react to
>> plugging/unplugging the Ethernet cable. I dunno how these daemons get
>> notified, but they do, and configure/deconfigure your interface.
>> That's what they do and what they're supposed to do. I never tried
>> netplug, but I guess it invokes ifup/ifdown, just like ifplugd.
>> Alternatively, you can also invoke ifup/ifdown by hand (-:
>
> I don't think there is any advantage to work around the dhcpdiscover
> boot delay by using daemons that bring up networking later. Several
> init scripts (like ntpd) rely on networking and have a dependency on
> this. Now when we bypass it, they simply fail and try later. This is
> not a clean concept.


    The only way to work around the dhcpdiscover delay is to configure
the interface statically. But this isn't the issue here.

    The delay we are dealing with is the delay until a person
eventually plugs a network cable into the computer. If ifup is invoked
at start-up in the absence of a carrier, it will give up after some
delay and the boot process can proceed. Not only this causes a delay,
but then nothing will bring automagically the interface when the cable
is plugged in, you will need to invoke ifup/ifdown manually, restart
networking, or even reboot.

    The solution is to not bring up automatically the ethernet
interface when you don't know if there is a carrier - which avoids the
boot delay, and have a daemon which will do it when there is a carrier -
which avoids rebooting after plugging the cable.

    dhcpdiscover is the very last step; it is invoked after alll the
preceding has succeeded.

       Didier