Alternative Re: ipv4/25s and above Re: 202211210951.AYC

Tue Nov 22 01:34:47 UTC 2022

> On Nov 21, 2022, at 7:18 PM, Joe Maimon <jmaimon at jmaimon.com> wrote:
> … Further, presentment of options in this fashion presumes that we have some ability to control or decide how engineering efforts across the entirety of the internet should be spent.

Joe - 

In the snippet above you allude to a very important aspect of the Internet that is rather germane to this discussion – ii.e. that we really don’t really have any "ability to control or decide how engineering efforts across the entirety of the internet should be spent” –, but then you don’t really work through what that fact means for realistic outcomes of class E space re-utilization…

First, I want to be really clear:  I don’t particular care one way or the other regarding the proposal to “de-reserve” 240/4… I don't run a network (nor has the ARIN community discussed the matter and directed that ARIN take a position either way.)  However, I do think the operator community should be thinking hard about how such de-reserving and redefinition into general purpose space will impact the Internet operations community and whether such space can realistically ever be utilized in production manner in the public Internet. 

As you alluded to, we really don’t have any "ability to control or decide how engineering efforts across the entirety of the internet should be spent”, and the practical implications of this fact is that there will always be many devices out there in production that will not pass IP packets with class E addresses in them…   (just as there’s always going to be some devices, somewhere that don’t know about IPv6.)

Of course, the difference is that with IPv6 we can attempt a connection and then fall back to IPv4, and further that devices out there either support and are configured for IPv6 routing, or they are not - networks rather quickly learn not to announce (via routing & DNS) IPv6 connectivity for devices without it actually being in place and operational or having solid IPv4 fall-back and relying fast fallback/happy eyeballs. 

With your using repurposed class E address space in the headers, your customers with such addresses are rather unlikely to ever know why a connection won’t establish – or why existing connections sometime fail mid-stream – as it only takes a single non-conforming device along the ever-changing path through any number of network operators to resulting in the silent drop of that packet.  That may (or may not) lead to you experiencing what you consider reasonable support costs for your customers, but as we all know, everyone else has customers who are the other ends of those connections who will call their ISP’s customer support line trying to figure out why they can’t get your customer (or can only get there intermittently) – so it appears that your proposed use of de-reserved and repurposed class E space has some real interesting implications about imputed support burdens on everyone else – if indeed the intended use case is includes providing connectivity to the public Internet.   

If you’re not proposing public Internet use, and rather just within your own administrative domain, then feel free to do – talk to your vendors, get them to support it, and turn it on.   As you already noted, we really don’t centrally decide how everyone runs their own network – so using it locally is fine since it doesn’t presume others will diagnose connection problems with your customer traffic that quite reasonably is categorized as invalid. 

Thanks,
/John

p.s. Disclaimer:  my views alone. Note: contents may be hot - use caution when opening. 

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.nanog.org/pipermail/nanog/attachments/20221121/0f8327af/attachment.html>