[6bone] link local for tunnel endpoints

[SHIRASAKI Yasuhiro]

On Mon, 27 Oct 2003 08:47:14 +0200 (EET),
Pekka Savola <pekkas@netcore.fi> wrote:
> On Sun, 26 Oct 2003, Dan Reeder wrote:
> [...]
> > Now because my local router and the remote router also have valid 2001::
> > global addressing (on mine for the /64 on another interface, on the remote
> > for other purposes), so traceroutes back and forth are going through just
> > fine. I realise that every device needs a globally reachable ip set on it
> > somewhere, even on a loopback interface, to be reachable.
> > But are there any operational down sides or gotchas that would prove this
> > type of addressing to be unsafe or impractical for use?
> 
> A few minor points I'm aware of -- should not be show-stoppers:
>  - when doing a traceroute, you can see which nodes the packets go 
> through, not which interfaces (the latter may be interesting e.g. with 
> backbone routers and their multiple interfaces).
>  - you can't ping the point-to-point address remotely, meaning, if the 
> other end-point has hosed its static route towards you, you can't isolate 
> the problem except from your border router, pinging the link-local 
> address.

Some old bgp4+ implementations couldn't work with link-local address.
Though peers over tunnel link seemed unstable and should be avoided,
yet we can see many bgp4+ peers over tunnel link.

--
SHIRASAKI Yasuhiro @ NTT Communications
t: +81-3-6800-3262, f: +81-3-5365-2990