[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: Ideas for IPv6 BGP and tunnelling



Adrian Kennard wrote:
[..]
> What is not clear to me is what this the best practice and workable IPv6
> next hop to specify. Seems to me it could be:-
> 
> ::x.x.x.x

That one was deprecated by RFC4291. I used to use ::/96 for storing IPv4
addresses inside IPv6, but then you don't know if :: is 0.0.0.0 or IPv6
::, as such I am now using the ::ffff::/96 in the places where I was
using that and the code that I have updated. Then ::ffff:0.0.0.0/96 is
for sure the IPv4 variant.

> ::FFFF:x.x.x.x
> 2002:xxxx:xxxx::
> 
> The latter seems to express that we want to use simple protocol 41 IPv6
> over IPv4 tunnelling. The first two seem to me to just indicate an IPv4
> address as the next hop without saying how the traffic is to be sent to
> it (e.g. GRE, protocol 41, whatever).

I would go for ::ffff:x.x.x.x, as then you have the properties you
describe above, but also know for sure that the 2002::/16 prefix can't
be hijacked by some routing entry, next to it being clear that this is
really on the wire as IPv4 and not as IPv6.

> FYI, I'll make our routers understand any of the above as a next hop to
> send over protocol 41 when received, but need to know what I should used
> when generating this as a next hop to send.

I do sincerely hope that you will be looking heavily at the security
concerns here, especially a line saying "only accept packets from known
prefixes" and "filter those prefixes out at the border" aka BCP38.

Greets,
 Jeroen

Attachment: signature.asc
Description: OpenPGP digital signature