Review: A network architecture providing host migration transparency

The Sony team proposes a network architecture in order to provide host mobility. Even though the article was written back in the 1991, we are still facing the same problem when dealing with host migration transparency.
They propose to separate the network layer to two sub-layers:

• The virtual network (VN) is a logical network that exists on the physical network. Each host in always connected to a VN just as it is connected to a PN. A host never migrates from its VN even if it migrates from its PN. A host has a VN address (permanent) and PN address (temporary).
  
• The physical network (PN): is the IP layer where the PN address indicates the location of the host in the PN which is used for routing.
  

The transport layer specifies the target host by its VN-address no matter where it migrates in the PN. The PN and VN have the same format.
When a host migrates, only its PN-address changes. Thereby the path to deliver data and not its identity (VN-address).
If host A wants to communicate with another distant host B, A uses B-VN-address. Host A-PN layer will translate Host B-VN-address to B-PN-address.

The Sony team uses the propagating cache method in order to reduce the overhead of address conversion. In this method each host and gateway hold a cache for address conversion. it is actually an address mapping table between the host's PN-address and the VN-address. Every entry in the table hold a VN/PN address, a source address timestamp used to determine whether the entry is obsolete and an Idle time (a Time To Live).
When a host changes a physical network it sends an update message "control packet" to its native network. This packet includes its VN-address and the new PN-address. When the packet travels to its native network, gateways on the path updates their entry with the new PN-address.
When a host A sends a request to another distant host B without any knowledge of host B PN-address, A just sets the host B PN-address with host B VN-address. The packets will be redirected to its native network and then will be redirected to its actual PN because on the path to the native network an address mapping of VN-PN will be hold by gateways.

This approach is interesting and seems to reduce overhead traffic comparing to Central server and broadcast methods.
However since packets sent from host A to host B don't always take the same path. Some of the routers and gateways will be updated and others will still hold an old mapping information. This would definitely lead to obsolete cache entries and loss of packets due to routing error.

link to the article