Search RPD Archives
Limit search to: Subject & Body Subject Author
Sort by:

[rpd] Report of the Soft Landing isuue

Owen DeLong owen at
Mon Apr 3 23:55:49 UTC 2017

> On Apr 3, 2017, at 4:41 AM, Noah <noah at> wrote:
> On 3 Apr 2017 12:55 a.m., "Owen DeLong" <owen at <mailto:owen at>> wrote:
> I am not calling for softening and depleting IPv4 at this stage, but I don’t see any advantage to tightening it, either.
> The last i checked, the IPv4 FIB is handling over 600k aggregate prefixes vs IPv6 FIB that stands at close to only 40k aggregate prefixes a compeling fact that the internet is still largely dependent on IPv4 today.

This is absurd in the extreme. The 600k prefixes are not aggregate, but actually disaggregate and a clear example of the untenable mess that IPv4 has become. The 40k prefixes in IPv6 cover a MUCH MUCH larger address space than the total 600k prefixes in IPv4.

Admittedly, they still cover a somewhat smaller number of hosts, and nobody is denying that the internet still depends on IPv4. In fact I’m saying that is a problem we need to be addressing urgently.

> Very few IPv6 only green fields to say the least.

Roughly 3% according to current statistics. So what?

> In fact, I would argue that by insisting on holding resources in the free pool for “possible future newcomers” you are, in effect, assigning them to organizations without any current proof of physical infrastructure in the AfriNIC service region to the disadvantage of organizations that do currently have proof of infrastructure and a documented need for the addresses within the region today.
> IMHO, your premise is flawed, in my experience (having worked for 3 SP startups and still do),  because we were all startups at some point when we involved ourselves in the business of connecting folk to the internet and every iron that we fired up then and today needed and still needs at the very least an IPv4 address to connect to the internet. 

You are ignoring reality because you don’t like the reality.

The reality is that some day there will not be any more available IPv4 addresses. The question isn’t how do we avoid running out of them. If it were, your arguments might make some sense. But the only answer to that question is “we can’t”.

Once you admit that to yourself, the question then becomes who should get priority for the few remaining addresses. Should we assign them to non-existent future newcomers to the detriment of real existing users trying to get connected today, or, should we focus on giving them to those entities that have real need today and recognize that anyone building a business model for a new business today that hasn’t looked at the 20+ year warning that IPv4 was finite and would run out is building on a severely flawed model and will have to face the consequences of that decision.

IMHO, stealing addresses from present need to hold them aside for potential future new entrants is poor resource management.

> In anycase, IPv6 internet is still developing sponteneously at almost 40k prefixes announced with a few case studies around the US, Europe, Asian and some parts of Africa and South America and could take another decade as long as telecoms  around the world still run CGN's.  

At the current rate of increase in IPv6 as a fraction of internet traffic, it has been consistently doubling every year for several years now. The current google statistic is 16% and Akamai shows somewhat higher. By this time 2018, we should be looking at 32% and 2019 the majority of traffic (64%) and by 2020, virtually ALL traffic. That’s not 10 years, that’s 3 years.

Let’s be generous and allow for 5 years.

> IMHO market forces and tech-dynamics (IoT) will push for IPv6 adoption and until then, the over a decade aggressive invetments in IPv4 internet will still stand even though most equipment and software today pretty much supports IPv6.

Certainly if we keep doing damage to existing networks in favor of encouraging people to deploy even more new networks using IPv4 because they can, this will be a self-fulfilling prophecy. However, this is exactly what I mean when I say that the only thing that can be accomplished this way is to prolong the increasingly severe pain that is represented by IPv4 today.


-------------- next part --------------
An HTML attachment was scrubbed...
URL: <>

More information about the RPD mailing list