suspect routing?

[Guy Davies]

On Tue, 4 Mar 1997, Alain Durand wrote:

> On Mar 4,  8:23am, bmanning@ISI.EDU wrote:
> > Subject: suspect routing?
> >
> > Hi,
> > 	What is the general take on mapped IPv4 address prefixes being
> > 	propogated within 6bone, e.g.  ::192.0.2.10
> >
> > 	And I've found a recent "feature" that allows the propogation
> > 	of the loopback address  ::1
> 
> 
> My opinion is that:
> 
> 	- IPv4 compatible routes should not be advertized

I certainly agree with this.  The injection of IPv6 compatible IPv4
addresses is bound to cause confusion and, as far as I can see, serves no
useful purpose (please correct me on that last statement if I'm missing
something obvious ;-)

> 	- Static routes should be injected in RIPng
> 	  only if they are for directly attached links or tunnels

This needs to be slightly widened to allow for any routes which are
connected via a static route to an organisation using static routing to a
RIPng site (I know this is unlikely but it's possible and so we should not
preclude the injection of such routes into RIPng).  I think the main aim
of Alain's statement here is to avoid multihop statics which conflict with
RIPng routes.  I can say from first-hand experience that this can cause
very strange problems which confused me and several others for most of a
day.

> 	- No default route nor 5f00::/8 routes shld ever be advertized

Not onto the 6bone at large.  You may want to advertise defaults to
internal networks running RIPng which have access to only one router.

> 	- more generaly, no routes with a prefixlength < 32 should
> 	  ever be advertized.

Again, on the 6bone, this should _never_ be necessary.  When mapped onto
the "future Internet" running IPv6, it will clearly be feasible, given the
proposed address usage (pre 8+8), for routes < /32 to be necessary. 

Just my thoughts.

Guy

> 
> - Alain.
>