[c-nsp] another take on outbound bgp load sharing, now with ipv6

[Joe Pruett]

i have had a good working setup where i tweak med values (with
always-compare-med) to help balance outbound traffic to two upstreams
that have almost identical as path lengths.  what i do is use the
oddball cisco non-continguous wildcard magic on the 9th bit in the
address and just use a simple even/odd match to set the metric.  as in:

! match if bit 9 is 1
ip access-list extended bgp-magic-isp1
 permit ip 0.128.0.0 255.127.255.255 any
! match if bit 9 is 0
ip access-list extended bgp-magic-isp2
 permit ip 0.0.0.0 255.127.255.255 any

route-map isp1-in permit 10
 match ip address bgp-magic-isp1
 set metric 1
route-map isp1-in permit 20
 set metric 2

route-map isp2-in permit 10
 match ip address bgp-magic-isp2
 set metric 1
route-map isp2-in permit 20
 set metric 2

this gives me a decent 50/50 split on equal length paths for outbound
traffic.

now that i've just turned up ipv6, i wanted to do the same thing, but
cisco dropped the wildcard stuff for v6 and only allows pure mask length
logic (as far as i've found).  and the way that v6 addresses are being
handed out by the rir's doesn't make it easy to come up with a way to
create a few rules to help split things like i've done before.

does anyone have any good ideas?

i've been reading up on the bgp maximum paths magic, and it seems like
it would be a good solution for me, but cisco seems to only insert
multiple routes if the as paths are identical, not just equal length.  i
even tried some magic found on the net to prepend my as to all incoming
routes based on the claim that would trick ios into doing multipath, but
no luck there either.  have i missed something about maximum path?

this is on 7206vxr with vanilla 12.4, if that makes any difference.