RE: 3gpp-analysis document and automatic tunneling

[<juha.wiljakka@nokia.com>]

 Hi, Karim and Pekka!

In my opinion, the simplest (and most straightforward) way is to use IPv6 PDP context, if you use IPv6 services and use IPv4 PDP context, if you use IPv4 (legacy) services. And that should be the main way, how things are handled.

However, it is possible that in the early phases of IPv6 deployment, there is no IPv6-capable GGSN (or SGSN/HLR) in the network (in the home or visited 3gpp network as Karim explained below) and the user still wants to use IPv6 - that's the special case we document in 3.1. I think that the current text "...the UE may activate an IPv4 PDP context and tunnel IPv6 packets in IPv4 packets using a tunneling mechanism. Tunneling in the UE requires dual stack capability in the UE. ..." is quite good. We can mention some tunneling mechanisms that can be used in that case (informative references), or just mention some dynamic tunneling mechanism alternatives in 2.2, as we do in the current document. Anyway, I feel that we cannot explicitely state, which tunneling mechanism should be used.

Cheers,
	-Juha W.-

-----Original Message-----
From: ext Karim El-Malki (EAB) [mailto:Karim.El-Malki@era.ericsson.se]
Sent: 28 May, 2003 13:06
To: 'Pekka Savola'
Cc: Fred L. Templin; Wiljakka Juha (NMP/Tampere); Soininen Jonne
(NET/MtView); v6ops@ops.ietf.org
Subject: RE: 3gpp-analysis document and automatic tunneling


 > >  > I haven't read the analysis in detail, but the latest 
 > >  > Scenarios draft at
 > >  > least *seems* to very explicitly mention that GGSN supports 
 > >  > both IPv4 and
 > >  > IPv6.
 > > 
 > > I am pretty certain that's not what the doc says.
 > > First, the fact that the GGSN can support IPv6 and IPv4 
 > routing does
 > > not mean that it has support for IPv6 native connectivity 
 > to mobile hosts.
 > > Secondly we cannot guarantee that all GGSNs from all operators
 > > will support IPv6 at the same time. Some GGSNs will only support
 > > IPv4 as they do today.
 > 
 > It seems to that the clear implication from the draft is:
 > 
 > "If you want IPv4, use an IPv4 GGSN"
 > "If you want IPv6, use an IPv6 GGSN"
 > 
 > These do not need to be the same.  This seems clearly stated 
 > in Scenarios 
 > sect 3.1

You can't really choose to use one or the other GGSN. You request
an IPv6 connection and if it doesn't work you revert to using IPv4
and some automated tunnelling like ISATAP. There are other mechanisms
also that I don't think we can exclude yet, which is why at least
mentioning the few applicable mechanisms in the draft is important.

 > 
 > PDP contexts could be opened to different service boxes in a service
 > provider's network (whichever support IPv6); and if the 
 > service provider
 > does not support *any* IPv6, the user is on its own *anyway*.
 > 
 > (of course, the UE could support a transition mechanism 
 > itself but I find 
 > that a bit questionable approach -- the goal is to make 3GPP devices 
 > simple not complex.)

My view is very different as you have realised already. I think we need
to have transition mechanisms for early v6 introduction stages and special
roaming cases which allow the user to access those v6 e.g. peer2peer
services even when a v6 pdp context is not available. I've tried it in
real networks and a simple mechanism like ISATAP works fine. Anyway we
seem not to be taking steps forward with this discussion so I'll
stop here.

/Karim