Re: 3gpp-analysis document and automatic tunneling

[Fred Templin <ftemplin@IPRG.nokia.com>]

Pekka Savola wrote:

> On Thu, 29 May 2003, Fred Templin wrote:
>  
> > > On Thu, 29 May 2003, Fred Templin wrote:
> > >
> > >
> > >      
> > > > > The client is
> > > > > in different admin domain as the router
> > > > >
> > > > >     
> > > > >          
> > > > No. The client gets an RA from the router that contains a non-zero
> > > > router lifetime and adds it to the default router list. The client is in
> > > > the same admin domain as the router.
> > > >   
> > > >        
> > > No.  AFAICS, the router belongs to the 3GPP operator.  The client belongs 
> > > to the 3GPP user (user device).  Right?  These are different entities.  
> > > Whether RA is used makes no difference.
> > >      
> > OK, but there is no difference between this and the model in which the
> > GGSN is serving the mobile a native IPv6 PDP context (i.e., the GGSN
> > belongs to the 3GPP operator and the client belongs to the 3GPP user.)
> >    
> Careful: there could be a lot of difference; that is not the original 
> ISATAP model.
OK.

> > > > > (tradeoffs of this might be
> > > > > manageable), as is in a different admin domain as all the other clients
> > > > > (which would also appear to be on the same link).
> > > > >
> > > > >     
> > > > >          
> > > > All the other clients would also be on the same link as the router. 
> > > >   
> > > >        
> > > But also on the same link as other clients in the same 3GPP operator (or
> > > maybe the particular GGSN), right?  In any case, it's *NOT* a 
> > > point-to-point link between the router and one client, right?
> > >      
> > No - it's a multiaccess link just like any other multiaccess link used 
> > for IPv6.
> >    
> Thanks for confirmation.  (It isn't 100% clear what you refer to with
> "No", but I'm assuming the latter sentence.)
>
> So, other clients are reachable on the same multiccess link, using
> link-local addresses.  That seems pretty significant to me.
>
> All in all, it seems ISATAP is applied in a scenario which it hasn't been
> originally designed for, but consideration to the differences in scenarios
> are overlooked.
We are discussing this application for the first time (as far as I can 
tell),  so I
believe the term "overlooked" is a bit premature.  I believe I have 
convinced
you that there are no functional barriers that keep an isatap client from
connecting multiple hosts. I'll grant you that issues in the trust model 
need
to be dealt with in more detail, however.

Fred
ftemplin@iprg.nokia.com