"Remco van Mook" <remco@xxxxxxxx> writes: > Most switches and routers I work with (that have q-in-q capability) > tend to have a configurable ethertype for a second (or outer) vlan > tag. Look at http://www.google.com/search?q=ethertype+0x9100 to see > what I mean. 0x9100 is starting to look like a more-or-less standard > for metro ethernet providers. Vendors supporting this include > Foundry, Cisco, HP and Juniper at the very least. What is the reason for using a new ethertype for q-in-q instead of just reusing 0x8100? I can't see any gain in this except complicating matters and cluttering up existing software. If I understand the whole q-in-q issue correct, the ehternet frame would look like this, for an IPv4 datagtam from host aa:aa:aa:aa:aa:aa to bb:bb:bb:bb:bb:bb vlan id 0x0111 in vlan id 0x0222: aa aa aa aa aa aa bb bb bb bb bb bb 91 00 02 22 81 00 01 11 08 00 45 ... Is this correct? urs
