generic v6 tunneling

[Steve Deering]

At 8:51 PM +0100 2/6/02, Francis Dupont wrote:
>   > - there is no reason to encapsulate such packets (they can be sent
>   >  directly).
>   
>   That's not the case if the tunnel entry point node is not the original
>   source node.
>   
>=> this changes nothing: in both cases there is a route to the
>destination and packets will follow it. The difference is they can be
>encapsulated or not, the visible source address doesn't matter...
>
>PS: differences can only be in source address based policies (QoS, IPsec,
>policy routing, etc).

Francis,

Perhaps we are saying the same thing.  If the tunnel entry point is
other than the source node, that intermediate node has some reason
to perform the encapsulation (security, performance, policy, whatever).
Simply omitting the encapsulation -- which is what I thought you were
suggesting -- may fail to satisfy the reason for the tunnel.

Or in other words, encapsulation adds information to the packet
(additional address(es), possibly different traffic class or flow label
values, possibly extension headers).  Simply removing that information
changes the semantics of the packets, in possibly harmful ways.

Steve