Re: [mobile-ip] Re: FW: I-D ACTION:draft-tsirtsis-dsmip-problem-00.txt

[Fred Templin <ftemplin@IPRG.nokia.com>]

George,

Tsirtsis George wrote:

> Now, when the mobile is  not at home but in some other network, there is an
> issue about what kind of v4/v6 support there is in the foreign network. The
> nice thing about Mobile IP is that it is based on tunnels and thus it
> provides natural way of tunneling over networks that are not compatible with
> the mobile ...if only Mobile IP could configure v4 over v6 and v6 over v4
> tunnels (as opposed to just v4 over v4 and v6 over v6 tunnels). A dual stack
> mobile in a v4 only foreign network would then be able to create a v4&v6
> over v4 (forward and reverse) tunnel with its HA and thus maintain all its
> connectivity.
When a dual-stack MIPv6 MN encounters a v4-only foregin network,
I see (at least) three possibilities for the v6-in-v4 tunnel endpoint:

1) the tunnel endpoint resides in the home network; perhaps
   even co-located with the HA (potential use case for configured
   tunnels)

2) the tunnel endpoint resides in the visited netwok (potential
   use-case for isatap)

3) the tunnel endpoint resides in some 3rd party network
   (potential use-case for tunnel broker)

In any case, the v6-in-v4 tunnel should present an MTU large enough
to encapsulate 1280 bytes PLUS the size of the outermost IPv6 header
so that the inner MIPv6  IPv6-in-IPv6 tunneling does not incur harmful
fragmentation (see RFC 2473, section 7). But, most v6-in-v4 tunneling
specifications cap their MTUs at 1280 bytes. Does this seem like a
potential performance issue waiting to bite us?

Fred
ftemplin@iprg.nokia.com