Re: SIIT/NAT64 is similar to RSIP

[Alain Durand <Alain.Durand@Sun.COM>]

[ post by non-subscriber.  with the massive amount of spam, it is easy to
 miss and therefore delete mis-posts.  so fix subscription addresses! ]


Jun-ichiro itojun Hagino wrote:

> 	no, the above description is incorrect.
>
> 	SIIT/NAT64 environment:
> 	For an FTP client on IPv6 only kernel (on the lefthand side of the
> 	diagram) to be able to contact and transfer files across SIIT box,
> 	the FTP client has to implement and understand IPv4 FTP commands,
> 	such as PORT and PASV.  this is not a normal requirement to IPv6-only
> 	FTP client.  IPv6-only FTP client normally implement EPSV/EPRT, for
> 	protocol #2 (IPv6) only, and that's all.  remember SIIT/NAT64 box will
> 	not rewrite the payload of packets.
> 	so, the FTP client knows about IPv4 (= outside world from the SIIT box).
> 	this is why i call SIIT/NAT64 similar to RSIP.
>
> 	NAT-PT environment:
> 	FTP client on IPv6 only kernel (or IPv6/v4 dual stack kernel) talks
> 	EPSV/EPRT, with protocol #2 (IPv6).  NAT-PT box will translate the
> 	content of the FTP control traffic into IPv4 ones (PORT/PASV).
> 	therefore, FTP client is seeing IPv6 peer from FTP protocol point
> 	of view.
>
> itojun
>
>
> SIIT/NAT64 environment
>
> FTP client
> ========== AF_INET6 API
> IPv6 only kernel		SIIT box	IPv4 FTP server
>  |				 |   |		  |a.b.c.d
> ==+==============================+= =+============+==
> 	IPv6 traffic			IPv4 traffic
> 	(::ffff:a.b.c.d)
>
> SIIT/NAT64 environment
>
> FTP client
> ========== AF_INET6 API
> IPv6 only/dualstack		NAT-PT box	IPv4 FTP server
>  |				 |   |		  |a.b.c.d
> ==+==============================+= =+============+==
> 	IPv6 traffic			IPv4 traffic
> 	(to some fake address)
>  
I think I now understand your problem.
What you say is true if and only if the ftp client has internal
knowledge that ::ffff:a.b.c.d is a special address that is in fact an 
IPv4 address.
That is, FTP looks at it in a sockaddr_in/AF_INET and not a 
sockaddr_in6/ AF_INET6
structure.

This does not have to be the case. Remove that knowledge from the 
application/kernel
and then you do not need anymore PORT/PASV.

   - Alain.