Re: [v6ops] What ie a site? (was: WGLC: draft-ietf-v6ops-dhcp-pd-per-device-02)

[Brian E Carpenter <[e-mail]>]

Owen,

A comment right at the end:

On 08-Nov-23 14:29, [e-mail] wrote:
> 
>>> Most likely, more FIB memory and faster RPs.
>>> However, that doesn’t seem like a realistic scenario given how long ARIN’s policy for PI has been running without anywhere near that level of growth in PI prefixes.
>>
>> Indeed - because a combination of RIR policies and the costs of operating a multihomed PI site have prevented explosive growth. But I have some sympathy with Jordi's view: there are at least 10 million small businesses in the world that would benefit from the improved reliability of a cheap, scaleable multihoming solution. We've been trying to solve that problem (without NAT) for 22 years
>> (the multi6 WG started in March 2001).
> 
> I would argue that it’s not RIR policies so much any more. RIPE policies don’t make PI hard to get, indeed, they’ve made it even easier than in ARIN, you just have to get it from a sponsoring LIR (~90EU/year).
> 
> Costs, well, yes, but really, those aren’t that high if you have someone on staff that knows what they are doing. That’s the only major cost remaining in the process. You can run a multi-homed site in the DFZ effectively on a nanopi R6S running Free Range Routing. Transit for a gig in a datacenter is down to pennies per megabit. Indeed, I can get DIA gig service in most urban areas of the US for ~$1k/month per circuit these days.
> 
> In terms of arguing against Jordi’s position on this particular subject, I’m the wrong guy to talk to… My house is AS1734 and in the DFZ. I’m taking full dual stack tables at home and one of my upstream connections goes to my cabinet at HE FMT2 where I’ve got 10G to FCIX and 1G transit from HE. I’ve also got upstream to a VM running on NetActuate with BGP to NetActuate at both FCIX and on my VM as well as set of upstream connections to Act USA in Las Vegas.
> 
> To control the costs of all that, those upstreams are all run over GRE tunnels that use Comcast Business and Ridge Wireless connections to provide the layer 2 transport to the upstreams.
> 
> Hardware wise, I’m running a pair of MX240s in the garage and an MX240 at HE.
> 
> I’m also the guy that wrote the first RIR PI IPv6 for End Users policy proposal (ARIN) and while that one didn’t get adopted, it was the genesis for other proposals that got merged together (for which I was also a contributing author) and eventually were adopted as the first RIR policy to allow end-site PI direct assignments for IPv6. David Farmer and I also worked extensively together when we were on the policies which are still essentially in place at ARIN for both ISP and End-User IPv6 allocations and assignments.
> 
> So I’m 100% behind Jordi’s position on this particular issue. End users that want to multi-home should be able to do so.

I didn't say anything to contradict that. But I've been waiting for a cheap, scalable solution without NAT since 2001.

    Brian