RE: IPv6 Home Use to stimulate deployment over IPv4-NAT

["Bound, Jim" <Jim.Bound@hp.com>]

Erik,

Good thing to parse and the traversal draft thanks.

Forget the odd proto-id that was bug in my busy processes below :--)

Thanks
/jim

 


> -----Original Message-----
> From: Erik Nordmark [mailto:Erik.Nordmark@sun.com] 
> Sent: Friday, February 21, 2003 7:56 AM
> To: Bound, Jim
> Cc: v6ops@ops.ietf.org
> Subject: Re: IPv6 Home Use to stimulate deployment over IPv4-NAT
> 
> 
> > Assume home routers want to support IPv6 and will 
> eventually but won't 
> > move until they believe it can be used over provider networks.
> >  
> > Assume there is not enough Ipv4 address space for providers to give 
> > out to all subscribers or cannot at reasonable cost.  But they can 
> > give the subscriber an IPv6 prefix.  This means 6to4 or 
> ISATAP won't 
> > work in this scenario in the users home.
> 
> I think what is needed are four things:
> 1. A method for the actual encapsulation
>    IPv6 over UDP over IPv4 might be the easiest
> 2. A method to keep the NAT state up to date
>    ICMPv6 echo's over the tunnel can do that
> 3. A method to detect when the NAT state has been lost or changed
>    so that it can be restored (perhaps using the same 
> mechanism in #4) 4. A mechanism to determine the tunnel 
> endpoint (IP address and port)
> 
> Note that draft-ietf-mobileip-nat-traversal-07.txt specifies 
> how to do this when #4 is the Mobile IPv4 registration protocol.
> 
> I think much of that draft can be used, and for #4 we can use either
>  - TSP as for tunnel broker (makes sense when some authentication of
>    the client is needed)
>  - a DHCPv4 option (makes sense when DHCPv4 is already used by the ISP
>    and no special authentication is needed for the tunnel)
> 
> > The home user network encaps the IPv6 packet at NAT with 
> Protocol ID 
> > equivalent to "6".  The provider then takes that packet and 
> decaps at 
> > their edge and uses native IPv6 or 6to4 to encap that 
> packet to where 
> > the IPv6 service is located.  I realize this has many 
> assumptions and 
> > I would work on those with some other folks interested in 
> this problem.
> 
> Using a separate protocol ID implies that the NAT box has the 
> functionality. Using UDP tunneling provides more flexibility 
> since one can run tunneling across a NAT box one can't 
> modify/control (like the one I have at home).
> 
> Thus until the home router has been modified one could do 
> this tunneling from the host at home.
> 
>   Erik
> 
>