RE: Issues about Teredo draft 8

> =A0=A0=A0=A0=A0=A0 But in my opinion, If the destination address of a T=
eredo=20
> packet is not a Terdo IPv6 address (Eg. global IPv6 addr), this=20
> packet should be discarded during step 2 because it's not a bubble=20
> or an ICMPv6 message! Do you think it's a=A0self-contradiction, or=20
> just a difference of the word "MUST'=A0and=A0"should"?

According to the specification, Teredo clients will sent to the server=20
ICMP packets or bubbles whose destination is a native IPv6 address.=20
The server will relay these packets to IPv6.

> =A0=A0=A0 2. I=A0want to implement=A0the Teredo server (also with Tered=
o=20
> relay function, since the burden of a Teredo server is not too heavy)=20
> on=A0the linux platform. Do you have some suggestion for=A0us during ou=
r=20
> implementation? Any kind is appreciated. :-)

Sorry, but I don't have much experience developing on Linux.

You can however certainly develop a Teredo server as a socket application=
, using a regular UDP socket for receiving packets from clients and relay=
s, and using a raw IPv6 socket to send packets to IPv6. Since the Teredo =
server never initiates transmission, the only thing you have to do is loo=
p on "recvfrom", get a packet, process it and send the response either ov=
er the UDP socket (sendto) or over the raw IPv6 socket.

>=A0=A0=A0 3. Do you think it's reasonable to add mobility support to=20
> Teredo=A0specification?=A0You know I come from China, where most of=20
> the=A0terminals behind NAT are GRPS terminals. I mean that maybe mobile=
=20
> users are the most urgent to get IPv6 connectivity bypass NAT(s) in=20
> the=A0near future. I think about=A0the mobility issue of=A0Teredo=A0ser=
vice,=20
> but it seems=A0too=A0complicated. :-( Do you have some comment?

AFAIK, most GPRS system provides you with an address that remains constan=
t for the duration of the GPRS session. This should normally also result =
in a Teredo address that remains constant for that session.

-- Christian Huitema