Re: [Idr] MP-BGP Extension and the Procedures for IPv4/IPv6 Mapping Advertisement

[Acee Lindem <[e-mail]>]

Hi Robert, 

> On Mar 18, 2024, at 11:08 AM, Robert Raszuk <[e-mail]> wrote:
> 
> Hi Acee,
> 
> Apologies if this has already been covered but it was clear from yesterday’s IDR meeting chat  that I’m not the only one who questions the necessity of this BGP extension.  
> 
> Why can you just use one of the IPv6 addresses ranges reserved for embedded IPv4 addresses as described in section 2.2.5 of RFC 4291? 
> 
> You man 2.5.5 ? 

Yes - I had the right link but wrong reference. 

> 
> Well the issue here is that this section describes a special address block to be used for mapping and last time I checked is hardly routable interdomai  



> 
> Moreover the goal of "IPv4-mapped IPv6 address" as explained even better in RFC4038 is to provide a way to talk to IPv6 services itself for IPv4  only sites/hosts. This proposal just aims to assist to interconnect IPv4 only sites. 

I got that but I don't know why it the advertisement/translation/routing wouldn't work as long as the IPv4 blocks aren't private addresses.  



> 
> So authors here are proposing to signal arbitrary prefix within and overlay between interesting (typically non directly connected ASNs) which of their prefix should be used to reach remote IPv4 sites. 
> 
> I think from this perspective the proposal is sound. 
> 
> What is however questionable and what we have recently discussed with Ketan - why not do encapsulation instead of double translation ... to accomplish the very same. 

Agreed - and this is generally cheaper than translation in the data plane. 

Thanks,
Acee


> 
> Cheers,
> R.
> 
> PS. 
> 
> If you use these embedded IPv4 using the standard mappings there no reason to advertise explicit mappings. There is already precedence for doing this - https://datatracker.ietf.org/doc/rfc6992/
> 
> This is intradomain only. 
>