Re: Connecting to remote IPv6 addresses from Linux machine requires local scope id

On Mon, 2006-07-24 at 18:36 +0530, Dhiren Chandvania wrote:

> Configuration details: 
> eth0      Link encap:Ethernet  HWaddr 00:30:6E:38:EA:EF
[..]
>           inet6 addr: fe80::230:6eff:fe38:eaef/64 Scope:Link

[..]
> sigfs     Link encap:IPv6-in-IPv4
> 
>           inet6 addr: fe80::c0a8:1df/128 Scope:Link
[..]

> Solaris:
[..]
> hme0: flags=2000841<UP,RUNNING,MULTICAST,IPv6> mtu 1500 index 2
[..]
>         inet6 fe80::203:baff:fe02:ccee/10

[..]
> Issues that I have: 
> 
>  
> 
> I]
> 
> All outbound traffic from a Linux machine needs scope id. Why? 

Because you are using link-local address space.
In short, those are only available on the local link. If you have
multiple interface they all get assigned out of fe80::/64, which is
actually part of fe80::/10 which is reserved for Link Local.

You should either use ULA (http://www.kame.net/~suz/gen-ula.html /
RFC4193) or use global addresses, see
 http://en.wikipedia.org/wiki/Tunnel_Broker for a long list of them or
of course contact your local ISP. If you don't want to ULA for whatever
reason, you can always abuse 2001:db8::/32 which is the documentation
prefix, but do note that a lot of things break when using that.

Please read http://noc.sixxs.net/faq/ipv6/?faq=whatisipv6 or a real IPv6
book to get into this topic.

Greets,
 Jeroen

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.3 (GNU/Linux)
Comment: Jeroen Massar / http://unfix.org/~jeroen/

iHUEABECADUFAkTFFk8uFIAAAAAAFQAQcGthLWFkZHJlc3NAZ251cGcub3JnamVy
b2VuQHVuZml4Lm9yZwAKCRApqihSMz58I5M4AKCbrPcERyuXv43sxcwfOeZPTAEB
nwCgtwlEgUjIoIkgIHl06nSSzfYUtTQ=
=CYDj
-----END PGP SIGNATURE-----