Re: Failover for a multihomed site with unreachable ISP

[marcelo bagnulo <marcelo@it.uc3m.es>]

[...]

> > 
> > I understand that the difference would be that RFC 2260 proposes that
> > the end-site injects the alternative route information and in the other
> > case is the ISP that injects the address space assigned by the other ISP
> > to the end-site.
> > However, i am not sure that this is a feature of the alternative
> > proposal. I mean RFC 2260 does not requires coordination between ISPs
> > while the other proposal does.
> 
> No, neither does our proposal. The ISP A reacts on signals from the DFZ 
> (vanishing of PB/32), the other ISP B is not directly involved.

Ok.
But the ISPB has to know which prefix to inject, but this may be
acceptable i guess...

>  
> A feature may be, that it is not nesseccary to establish a BGP peering to the 
> customer and therefore the customer doesn't need to have a non-private ASN. 
> 

Yes, i think so.
However, the end-site needs some mechanisms to known how to route
packets. I mean, if there is an outage above ISPa, the end-site needs to
know that it must send packets through ISPb, how do you achieve that if
you do not use BGP?

Regards, marcelo