Subject: Re: null interface implementation
To: Luke Mewburn <>
From: Andrew Brown <>
List: tech-net
Date: 08/22/2002 09:53:53
>  | >>>I attached a small patch to use null0 network interface
>  | >>	IIRC you can use lo1 for this (maybe with RTF_REJECT).
>  | 
>  | 	but anyways, i'm okay with adding it.  if i hear no major objection
>  | 	for a couple of days i will commit it.
>is it really necessary to create Yet Another Device (null0 network,
>which namewise could be confused with "null" (as in /dev/null)), when
>it seems that using lo1 (with "pseudo-device loopback 2", or possibly
>by making "lo" a cloning device?) should suffice.

creating a null interface is easy.  even i can do it (though i called
mine nul).  :)

seriously, though, i think having two loopback interfaces is wrong (or
let's just make the loopback interface itself a cloning interface
which is also very easy), especially when one of them isn't going to
be used to loop anything back.  for one thing, they both get a ::1
address.  at least, they did the last time i tried it.

the cisco null interface implementation, which is where i originally
got my idea, can have routes pointed out of it (we don't really do
interface routes), and can be configured either to send or not to send
icmp unreachable messages.  you are also allowed only one (so we
differ there because i can conceive of wanting more).

otoh, ciscos can have up to 2^31-1 loopback interfaces...

>or am i missing something?

probably not.

|-----< "CODE WARRIOR" >-----|             * "ah!  i see you have the internet (Andrew Brown)                that goes *ping*!"       * "information is power -- share the wealth."