nanog mailing list archives

Re: AWS Using Class E IPv4 Address on internal Routing


From: Enno Rey <erey () ernw de>
Date: Tue, 9 Mar 2021 16:49:10 +0100

Hi,

On Tue, Mar 09, 2021 at 06:53:33AM -0800, Fred Baker wrote:
The "RFC" you're looking for is probably https://tools.ietf.org/html/draft-wilson-class-e-02

there was/is also 'IPv4 unicast extensions project' (https://github.com/schoen/unicast-extensions) with a similar idea 
& approach.
I for one think that from an operations perspective those addresses would be pretty much unusable in pretty much all 
complex networks (except for corner cases like, maybe, AWS' one), see 
https://theinternetprotocolblog.wordpress.com/2019/10/06/some-notes-on-ipv4-address-space/.

cheers

Enno





, which was not agreed to and so has no RFC number. The fundamental issue that was raised during that discussion was 
that while repurposing class e would provide a few more IPv4 addresses and so delay the need to replace the IPv4 
protocol for some period of time, APNIC's experience with a new /8 in 2011 (it was given the /8 in January 2011, and by 
April had largely distributed it to its members) suggests that the address space would be used up almost immediately if 
distributed publicly, and if used privately doesn't benefit the many networks that really honestly wish that we could 
squeeze more than 2^32 addresses into a 32 bit container.

I'd really suggest using IPv6. Networks like Reliance JIO in India, which has turned off or never turned on IPv4 for 
most of its services, find that they don't need IPv4 apart from customer preference.

On Mar 9, 2021, at 6:36 AM, Douglas Fischer <fischerdouglas () gmail com> wrote:

So, if an organization wants to use that, they will require from the vendors the compliance with that RFC.



Em ter., 9 de mar. de 2021 ??s 11:00, Forrest Christian (List Account) <lists () packetflux com> escreveu:
Back a little bit ago when the thread about running out of RFC-1918 space was going on, I was going to make a 
suggestion about repurposing the Class E space in the case where one ran out of space, assuming one could get the 
vendors on your network to support this address range.

I sort of discarded the suggestion just because of the whole issue of that range being hardcoded as invalid in so 
many implementations that this didn't seem all that useful.

On the other hand, if you're large enough that you're running out of RFC-1918 space you might be able to exert 
enough power over select vendors to get them to make this work for selected purposes.   Router-to-Router links, 
especially between higher-end routers seems to be one of those cases that it might be useful.     It might be the 
case that Amazon is already doing this....


On Mon, Mar 8, 2021 at 12:07 PM Douglas Fischer <fischerdouglas () gmail com> wrote:
Has anybody seen that also?

P.S.: I'm completely in favor of a complementary RFC assing FUTURE USE exclusively to "Between Routers" Link 
Networks...

--
Douglas Fernando Fischer
Eng?? de Controle e Automa????o


--
- Forrest







-- 
Enno Rey

Cell: +49 173 6745902
Twitter: @Enno_Insinuator


Current thread: