Re: [v4v6interim] Comparison document

> -----Original Message-----
> From: Mikael Lind [mailto:mikael.lind@hexago.com] 
> Sent: Sunday, September 21, 2008 8:47 AM
> To: Dan Wing; v4v6interim@ietf.org
> Subject: Re: [v4v6interim] Comparison document
> 
> The client side might be more or less identical, at least for 
> ip in ip 
> tunnels, but the concentrator will be quite different. The DS-lite 
> concentrator doesn't need any establishment or preexisting 
> states in order 
> to receive and translate packets, as it is all done per 
> session, while the 
> SNAT has to to establish and maintain tunnels for all clients 
> at all time.

I am not an expert in tunnels, but it is my understanding that
tunnel establishment is trivial; RFC2003 doesn't require messages
to be exchanged to establish a tunnel.

As for tunnel maintenance, the problem described in Section 5 of 
RFC2003 is due to routers along the tunnel path returning short
ICMP errors.  Since RFC2003 was written, it is my understanding
that most routers return longer ICMP errors.

If I am wrong, I know someone will correct me.  Any such drawbacks
should be described in draft-touch-intarea-tunnels-00.txt and we
will certainly discuss such drawbacks at the interim meeting and
I expect the revised SNAT+DS-Lite draft will discuss the positive
and negative aspects of using IPv4-over-IPv6 versus IPv4-over-tunnel.

-d

> >> 3.1.2 describes the two proposals Dual-stack lite and SNAT as
> >> more or less
> >> the same. Even if there is discussion about merging them I
> >> still think it
> >> would be worth pointing out in the document that there is a
> >> big difference
> >> between the two proposals.
> >> DS-lite doesn't require a tunnel to be set up in order to
> >> function as it
> >> keeps the IPv6 source address as part of the NAT state, much
> >> like a NAT-PT.
> >> In the case of SNAT there has to exist a tunnel per user
> >> before traffic is
> >> sent as tunnelling and NAT are more or less separate.
> >
> > The 3 authors of SNAT and DS-Lite all told me to collapse the
> > text of their proposals because they are consolidating the two
> > proposals.
> >
> > SNAT uses the tunnel identifier to distinguish each subscriber;
> > DS-Lite uses the IPv6 address to distinguish each subscriber.
> > This is a pretty minor distinction in the overall scheme of
> > things.   In both cases, the concentration endpoint must be
> > configured (or otherwise learned, e.g., via a DHCP option)
> > and in both cases the NAT44 function needs to differentiate
> > each subscriber (such as VLANs, tunnel ID, or IPv6
> > interface) so that subscribers can be given overlapping
> > IPv4 addresses.
> >
> > -d
>  
> 

_______________________________________________
v4v6interim mailing list
v4v6interim@ietf.org
https://www.ietf.org/mailman/listinfo/v4v6interim