netdev
[Top] [All Lists]

Re: routable interfaces WAS( Re: [PATCH] hashed device lookup(DoesNOT me

To: Michael Richardson <mcr@xxxxxxxxxxxxxxxxxxxxxx>
Subject: Re: routable interfaces WAS( Re: [PATCH] hashed device lookup(DoesNOT meet Linus' sumission policy!)
From: jamal <hadi@xxxxxxxxxx>
Date: Sun, 7 Jan 2001 19:56:20 -0500 (EST)
Cc: "netdev@xxxxxxxxxxx" <netdev@xxxxxxxxxxx>
In-reply-to: <200101080026.f080QE120377@sandelman.ottawa.on.ca>
Sender: owner-netdev@xxxxxxxxxxx

On Sun, 7 Jan 2001, Michael Richardson wrote:

>
>   The nicest thing about routable interfaces (vs what FreeSWAN and many other
> IPsec's use now) is that it makes the choice of outgoing IP address (the one
> inside the tunnel) behave like all other multihoming.
>   I think the same criteria applies to VLAN interfaces as well.
>

As well for MPLS (on an LER), L2TP, IP/AAL5, all other tunneling
proptocols etc. anything that needs to use IP and be "forwardable".
I think the net_device by design is a routable interface because of IP
influences.
Multi-homing means just adding another IP alias; unfortunately i wouldnt
call an alias "routable" (if you agree with my definition of a routable
interface). Making IP aliases net_devices would solve the problem;
thats what VLANs do today. I would also think there are people who run
4000 aliases. urgh, maybe i am not making sense and should write up
something.

cheers,
jamal



<Prev in Thread] Current Thread [Next in Thread>