Here are a set of questions we would like to ask to router vendors,
when we make purchase decisions :-)
which interface types do you support IPv6? ether, p2p (BRI/PRI/OC48/...), dialup, 802.1Q VLAN, ...
SNMP IPv6 transport? IPv6 MIBs?
ssh/telnet IPv6 support, incoming/outgoung
ipv6 forwading performance: normal case (fast path), and with routing header case (slow path).
any negative impact on IPv4 performance, if we have IPv6 traffic? or, when we configure IPv6?
do we have a good orthogonal user-interface for ipv4 and ipv6?
some vendor routers do like this:
show ip bgp summary
show bgp ipv6 summary
and is confusing.
can we establish BGP peering over IPv6 link-local address? what happens if we have multiple interfaces?
can we configure multiple link-local addresses to a single i/f?
can we configure multiple address from the same prefix, onto a single i/f: like P::a and P::b to the same interface?
can we configure lifetimes on RA messages? what is the default value? (the value suggested on RFC, or what?) validation against RAs from other routers?
acl? filters? performance drawbacks if we configure filters?
router renumbering support?
memory footprint growth?
stability when you enable IPv6? can we ensure SLA agreement for IPv4 while operating the routers in IPv4/v6 dual stack config?
support? no support, special contacts, or normal support contact?
when do you going to support IPv6 by default? do we need a special firmware for IPv6, or is IPv6 integrated to normal firmware?
can we operate routers IPv6-only? can we disable IPv4?
performance with tunnels?
support for ping6 -w (ICMPv6 node information query, node name)?
DNS over IPv6?
DNS record types? (AAAA is the way to go)
multicast routing? which multicast routing protocols? IPv6 PIM dense/sparse is a must, as we will not be putting any DVMRP tunnels for IPv6.
can we use the box as site border router? can we support multiple site routing domains?
which IGPs do you support? RIPng, ISIS, OSPFv3? (OSPFv3 is important for many Japanese networks)
tftp over IPv6?
how can we disambiguate linklocal peers on multiple different interfaces? do you support "fe80::1%ether0" syntax?
which kind of tunnelling encapsulations? static configured tunnel, 6to4, auto-tunnel? (no need for auto-tunnel)
# of IPv6 routes you expect to take?
# of IPv4 routes you expect to take, when IPv6 is enabled?
(ether switch) MLD snooping for multicast flooding optmization
(ether switch/wireless basestation) per port/source-MAC-address filtering for router advertisements, to prevent malicious routers from showing up
have you carefully check the procurement specification by JGN (Japan Gigabit Network)? it should be useful for you.