>       I think the sending host should just drop the packet when the
>       destination is not covered in its prefix list and there is
>       no default route installed. Otherwise IMHO this last attempt 
>       would lead to erroneous packet generation, and allows 
>       misconfigured host or host running a buggy implementation to 
>       continue with bad behavior. 

Imagine a case when host1 and host2 are on the same link, but
DHCPv6 has given them different /64 prefixes for some reason.
The routers have been configured to not advertise any on-link prefixes
i.e. the hosts send packets to the router and get redirected to
the on-link destination.

The communicate fine as long as the router is around.
When the router fails this communication local to the link would fail, unless
the fallback to treat all destinations as on-link is in place.

A less esoteric example is when the hosts are in the same prefix,
and either no on-link prefix is advertised (no 'O' bit in the prefix
advertisement) or the on-link prefix expires in relatively short time
i.e. the hosts quickly forget which destinations are on-link.

>       Also the code put in to supporting this looks and feels 
>       like hacks.

That seems like a comment for the implementors of whatever
implementation you are looking at, and not a comment relating to
the RFC.

  Erik

--------------------------------------------------------------------
IETF IPng Working Group Mailing List
IPng Home Page:                      http://playground.sun.com/ipng
FTP archive:                      ftp://playground.sun.com/pub/ipng
Direct all administrative requests to [EMAIL PROTECTED]
--------------------------------------------------------------------

Reply via email to