Re: [Idr] Fw: Re: New Version Notification for draft-xie-idr-mpbgp-extention-4map6-00.txt

[Robert Raszuk <[e-mail]>]

Hi,


> *[Chongfeng]: *Let's make a preliminary quantitative analysis, suppose the size of IPv4 BGP table of Internet is R, and that of IPv6 is S, the values of R and S are about 1M and 200k respectively now. Just from the perspective of IPv6 BGP routing, introduction of new address-family will have new cost, but from the perspective of the network as a whole, this is equal to move IPv4 reachability information into IPv6 domains by mapping, there will be no native IPv4 reachability information table in multi-domain networks anymore.
>
> And that is exactly one fundamental problem with your proposal.

Assume you have PEs with attached sites which will need to connect to IPv4
Internet. So in order to provide this you will need to *DUPLICATE* entire
1M of IPv4 routes into your new SAFI so now the capacity of your control
plane doubles.

If sites do not need to connect to the Internet there are other proposals
which do not require a new BGP mapping SAFI.

Example:

https://datatracker.ietf.org/doc/draft-mishra-idr-v4-islands-v6-core-4pe/
**

** I do have concerns about language of this draft or it's place in IDR,
but I am bringing it here to illustrate that what you are proposing is much
heavier and complex control plane wise.



> Furthermore, the forwarding of IPv4 services in P routers will be based on IPv6 FIB, the size of which is
>
>
In all cases forwarding in P routers will be based on IPv6 FIB so I do not
understand why you are highlighting it here.


*[Chongfeng]:* I agree with you that it is local configuration to
place mapped address as source in the outer encapsulation.  But it is
not true for destination address, the creation of outer IPv6
destination address needs to get the IPv6 mapping prefix of remote
end-point in advance, which needs the coordination of different nodes.
>
> Not really.

Destination address used as egress tunnel endpoint can be /96. So remaining
bits if needed can be filled with IPv4 destination.

IPv4 to IPv6 mapping is algorithmic and does not need to be signalled. The
routing to egress will still be done based on /96.



> That alone is highly questionable as such mapping's algorithm is well known
> and each node can do the mapping by itself. All it needs to know is the
> AFI/SAFI 1/1 and tunnel endpoint. There is no magic to it. This really
> proves no justification to invent new SAFI for it.
>
> *[Chongfeng]: *Can you show which draft or RFC specifies the technology or procedure to get the mapping prefix of remote-end?
>
>
https://www.rfc-editor.org/rfc/rfc4291

Section 2.5.5

Regards,
R.