[e2e] source code release

[Zoltan Turanyi]

Christian,

See comments inline.

On Mon, 10 Mar 2003, Christian Huitema wrote:

> 
> There is a whole family of protocols that work by overlaying a unique
> name space on top of the existing IPv4 infrastructure. These protocols
> are deployed by adding software to the end stations (to understand the
> new overlay convention) and also optionally in some relays, to overcome
> connectivity limitations. Your proposed 4+4 experiment has exactly those
> characteristics: change the hosts so they understand the convention;
> change the relays so they can do some form of "cross realm" routing.
> This is not radically different from other experiments, like HIP, or
> indeed like the IPAE proposal of 1992. 

There is one important difference. In HIP (and IPAE)  the "upper"
namespace is independent of the underlying "address space". In 4+4 the
"upper" address space includes the lower address space. The current IPv4
address of a host is part of its future 64-bit 4+4 address. This enables
the continued use of v4 routing; the goal is not to replace those routers. 

Transition to v6 happens by introducing 6to4 (and ISATAP) tunneling, and
then removing it. Transition to 4+4 happens by introducing tunneling,
there is no removal. The tunneled packet format is the final one, the two
addresses in the tunneled packet collectively constitute the final 
address.

> You can achieve exactly the same result with a combination of two
> currently deployed IPv6 services, 6to4 and ISATAP; 6to4 provides an IPv6
> overlay on top of the connected IPv4 backbone; ISATAP provides an IPv6
> overlay on an IPv4 private realm. The requirements of the 6to4/ISATAP
> combo are exactly the same as the requirement of any other overlay
> solution: upgrade the participating hosts, and upgrade the enabling
> "NAT". You can in fact achieve an even better result by using the Teredo
> approach, which overlays IPv6 on top of UDP, and as such does not
> require upgrading most NAT.

The result is very different in the two cases. By deploying 6to4+ISATAP 
you end up in the middele of a transition period. By deploying 4+4 you 
finished the transition and extended the address space.
 
> Using IPv6 to structure the overlay has the big advantage of a clean
> exit strategy: as some point, we can quit overlaying, and just route
> IPv6.

The point is, that with 4+4 you do not need an exit strategy. IPv4 routing 
is kept "forever" and no new routing needs to be installed to finalize 
transition. 

In 1997 in his Sigcomm keynote John Postel predicted that IPv6 transition
might take longer time than we expect. After nearly 6 years this is still
true. I fear that we have to maintain a parallel v4/v6 Internet for quite
some time (well, there is some relcutance even to start it). 4+4 aims to
avoid the "two Internets".

Zoltan