[afripv6-discuss] First tests with IPv6

[Nishal Goburdhan]

On 16 Jul 2015, at 15:35, Willy Ted MANGA wrote:

> Hi,
>
> Le 16/07/2015 10:53, Nishal Goburdhan a écrit :
>> On 16 Jul 2015, at 8:52, Willy Ted MANGA wrote:
>>
>>> Hello Nishal,
>>
>> hi,
>>
>>> Yep, Camtel has at least 8 IPv6 prefixes and I will bug them soon
>>> on that purpose :)
>>
>> not according to afrinic:
>
> It's a bit strange .. I was looking on hurricane here [1]
> 1. http://bgp.he.net/AS15964#_prefixes6


ok, so, if you look closely, you’ll see that 2001:4268::/32 is the 
covering prefix for all of these prefixes.
someone then though it was a good idea to advertise another 7 
de-aggregated prefixes  (all of which are contained within that single 
/32)

if you deconstruct that you’ll see:
(with some help from: 
http://www.cidr-report.org/v6/as2.0/aggr.html#AS15964)

AS15964: CAMNET-AS,CM
   Prefix  (AS Path)                   Aggregation Action
2001:4268::/32       6939 6453 15964
2001:4268::/48       6939 6453 15964      - Withdrawn - matching 
aggregate 2001:4268::/32 6939 6453 15964
2001:4268:1::/48     6939 6453 15964      - Withdrawn - matching 
aggregate 2001:4268::/32 6939 6453 15964
2001:4268:2::/48     6939 6453 15964 15964 15964 15964 15964
2001:4268:200::/48   4777 2497 6453 15964
2001:4268:210::/44   4777 2497 6453 15964
2001:4268:220::/48   4777 2497 6453 15964
2001:4268:2a0::/48   4777 2497 6453 15964

2 of these prefix announcements are unnecessary;  they’re the ones 
marked as “withdrawn”
the others appear to be … some attempt at traffic management.  it’s 
difficult to believe that it would have any nett effect on the end 
network (15964) though!

at a rough guess, to me, it seems more like someone took at bad habit 
that they had used in IPv4 (“hey, let’s de-aggregate”) and simply 
applied that here.  that’s a Bad Thing to do;  more so in v6, because 
the chances that you (or rather, everyone) has to contribute to routing 
table bloat are much, much more.  if de-aggregated, a typical IPv4 /16 
(which is a large amount of address space for most ISPs in this region), 
that will yield (0..255)+1 IPv4 /24 prefixes that could be visible in 
the internet routing table.  if you apply that, frankly, quite stupid 
network configuration, to v6, you could de-aggregate an IPv4 /32 to 
(0..65535)+1 IPv6 /48 prefixes, which could possibly appear in the 
routing table.  so, much worse.

it pays to think about what routing policy you *actually* need to have; 
in the long run, it’s better for everyone.  i really doubt that this 
policy is doing CAMTEL any good;  but perhaps they have their own 
(valid) reasons..

—n.