pTLAs missing in action

[Steve Deering]

At 11:30 AM -0700 6/30/99, Richard Draves wrote:
>OK, some progress... now the list of pTLAs for which I have no addresses to
>ping is
>
>	3ffe:800::/24 (ISI-LAP)
>	3ffe:d00::/24 (ANSNET)
>	3ffe:1500::/24 (UO)
>	3ffe:1600::/24 (NUS-IRDU)
>	3ffe:1700::/24 (MREN)
>	3ffe:1f00::/24 (NETCOM-UK)
>	3ffe:2700::/24 (ERA)
>	3ffe:2d00::/24 (GRNET)
>	3ffe:3500::/24 (REGIO-DE)
>	3ffe:3700::/24 (ABILENE)
>	3ffe:8010::/28 (ICM-PL)
>	3ffe:8030::/28 (QTPVSIX)

It would be very nice if, for every piece of topology identified by an
address prefix, the border routers of that piece of topology would
automatically assign to themselves an anycast address consisting of
the relevant prefix followed by all zeros.  The anycast address would
be assigned to the interface(s) attached to the prefixed topology.
This would allow you to ping a prefix and get an answer from the nearest
border router of that prefix.  This particular use of anycast addressing
does not impose any additional load on routing outside the piece of
topology (since it aggregates with all the other addresses for that
piece of topology), and adds only one entry to the routing inside the
piece of topology (for the zero-valued next-level-down prefix).

Note that we already require this for the border routers of subnets
(effectively, all routers), and I recall that the IPv6 address allocation
policies that the regional registries are proposing/using requires this
for border routers of sites.  I suggest we require it for the border
routers of all prefixes.

Comments?

Steve