Hej Jørgen,
What changes from CEF256 to dCEF256 is primarily where the forwarding descision is made, not so much how the packet is actually forwarded.
When running CEF256 (Sup2+SFM but no DFCs), the supervisor is involved in the forwarding *descision* of each and every packet: The packet (or rather, the header of the packet) is forwarded via the D-Bus to the Supervisor which then makes a forwarding descision. The result of that descision is sent on the R-Bus.
The actual forwarding of the packet in case 1) is then done via the SFM, whereas case 2) is handled on the line card.
If you install dCEF256 line cards, the forwarding descision will also be handled by the line card, because the DFC is basically a local PFC. This is why the forwarding is rated at 30 Mpps per line card in stead of 30 Mpps for the entire chassis. When running dCEF256, not even the header will be sent over the bus.
In your scenario 1) the packet is of cause still forwarded over the SFM. In scenario 2) the packet is actually still forwarded via the SFM *if* one port is on one Medusa ASIC and 8G serial channel and the other port is on the other Medusa. If both ports are local to the Medusa, the packet is also forwarded locally.
In a mixed system (both classic and CEF256/dCEF256 systems), things obviously get a little more complicated...
There's a C6500 architecture white paper somewhere on Cisco.com. Drop me a note if you can't find it.
HTH,
-A
