On Tue, Jun 20, 2006 at 08:04:18AM -0700, Ben Greear wrote: > >Or do you mean that people run dhcpd on eth0 but expect it to > >still serve VLANs? I don't think that should work. > > People run dhcp (and other programs I'm sure) on eth0.7 and send it > raw ethernet frames and expect that the vlan driver will do the > work of adding the vlan framing. Yep, I'm one of them. > But, it would be equally valid to send a frame with the vlan-7 > header already in it to eth0.7, so that must continue to work > (without re-encapsulating it into eth0.7.7) This is very unintuitive IMHO. I would indeed expect tagged packets sent to a VLAN interface to come out as qinq on the wire. But never mind. It could work the old way by default and then require vconfig to be used to choose the new (qinq) mode. On Tue, Jun 20, 2006 at 05:15:59PM +0100, Alex Zeffertt wrote: > >But, it would be equally valid to send a frame with the vlan-7 > >header already in it to eth0.7, so that must continue to work > >(without re-encapsulating it into eth0.7.7) > > This is the bit which confuses me. Why would anyone want to do > this? Why wouldn't they instead just sent the frame to a raw socket > on eth0? Perhaps it's another part of the kernel? Maybe bridging or something else has some optimizations that lead to such behavior? //Peter
