On Wed, 2004-05-05 at 13:17, Florent Parent wrote:
> -- On Wednesday, May 05, 2004 16:59:32 +0200, Jeroen Massar wrote:
>
> > To rephrase:
> >
> > The configuration protocol can easily see if it is a NAT or not.
> >
> > Clients MAY request ANY, UDP or Proto-41 as encaps.
> >
> > When the encaps is ANY the client sends the IP it think it has globally
> > to the configuration server, which compares it with the actual
> > connection origin, different -> NAT -> use UDP. Otherwise it can safely
> > assume proto-41.
> >
> > Additionally:
> >
> > Clients MAY also check if they have an RFC1918 address, they then also
> > know that they are behind a NAT.
>
> In fact, I would not make any assumptions about the IPv4 addresses. This
> service could be offered inside a large privately-addresses network, where
> both the clients and server are using rfc1918 addresses, therefore no NAT
> in the path. IMHO, its best to let the server do the detection of whether
> an address translation occured.
That is true and another scenario could be:
{Internet} --- {ISP} --- | NAT | --- [TB/6to4/ISATAP/...] ---- {clients}
Thus I revoke that thought ;)
Greets,
Jeroen
Attachment:
signature.asc
Description: This is a digitally signed message part