Search RPD Archives
[rpd] Migrating quickly to IPv6
lee.howard at retevia.net
Wed Jun 6 15:09:57 UTC 2018
Several lists have been very active in the last 12 hours, and I have
much I want to say on this 6th anniversary of 6/6/12.
On 06/05/2018 05:15 PM, Owen DeLong wrote:
>> On Jun 5, 2018, at 12:44 , sm+afrinic at elandsys.com wrote:
>> Hi Lee,
>> [Subject line changed]
>> At 08:44 AM 05-06-2018, Lee Howard wrote:
>>> Done collaboratively, I can imagine ISPs, IXPs, and web hosting companies migrating quickly to IPv6 with just a small translator infrastructure between them. It could be similar to what we've seen in several parts of Africa where a small old-fashioned infrastructure (land lines, banks) have enabled Africa to leapfrog much of the world (mobile phones, phone payments).
>> There is an ICANN presentation about IPv6 deployment in several countries. It could be interpreted as IPv6 is not a replacement for IPv4. One of the points in the above is that a migration entails some cooperation between different entities. There are also issues, e.g. what to do about mobile phones which only have IPv4 support.
SM, can you find that presentation? I might need to have a word with a
colleague at ICANN.
What to do about old devices that only support IPv4? Several choices:
* Let them become obsolete, like devices that only support SNA or
DECNet: they are useful only on specific networks.
* Provide edge transition mechanisms, such at 464xlat, MAP, or DS-Lite.
* Provide them a public IPv4 address. Maybe charge them for it, since
they're costing you (the carrier) money.
> OK, there’s a difference in perspective about IPv6 being a replacement for IPv4 depending on the timeframe of your focus.
> Long term, there is no other viable candidate for IPv4 replacement available today and therefore IPv6 MUST replace IPv4 if the internet is to preserve any semblance of its current functionality for a growing population of internet users.
I agree. This is a migration. There's a transition period during the
I might disagree with your dates, but only quibbles.
I think it's also important for people to keep an eye on global IPv6
I suggest using a Logistic curve (S-curve), since that is typical of
Note that this particular case is just what Google reports; other
reporters have different numbers.
Those who say IPv6 deployment have slowed are using old data, and
ignoring the fact that it always slows in the first several months of
the year. There has been a significant rise in the past few weeks, which
puts us back on track for the majority of Internet hosts to have IPv6 in
Again: More than 50% of Google hits will use IPv6 in less than two years.
Time to get to work.
> Medium term, I believe that IPv6 will replace IPv4 as the lingua franca of the Internet. This does not mean that IPv4 will disappear from the world (or even the public Internet) in this time frame, but it does mean that more and more traffic will shift to IPv6 and at some point, IPv6 will represent a sufficient volume of traffic that a growing number of eyeball providers will consider IPv4 a “value added” service that can be provided only to customers willing to pay a premium for it (some providers will simply stop providing IPv4 support, or, provide it strictly via some translation mechanism).
> In fact, today, some mobile providers are already at that point (not providing native IPv4 support and only translation mechanisms).
> Here’s how I see things progressing from where we are today: Note, some countries (and even regions) will lag this timeline (Africa is likely one of the worst laggards based on experience to date) others may lead it slightly…
> Some mobile providers are IPv6-only native with NAT64 or other translation based solutions for IPv4.
> IPv4 is still the “default” ISP service for most providers.
> You can still get away with calling it “Internet Access” even if it doesn’t include the whole internet.
> T+2 years:
> In most places, you won’t be able to really claim you deliver “internet access” unless it includes IPv6.
> Most mobile carriers will be native IPv6 only with translation for IPv4.
> Most eyeball providers will still include IPv4 along side the IPv6 internet service(s) they sell.
> It will still be difficult to sell an IPv6-only “Internet Access” product unless it includes translation for IPv4.
> T+4 years:
> Many eyeball providers will start considering IPv4 a value-added service and turn it off for customers that
> don’t elect to pay a premium to keep it.
> IPv4 translation may still be necessary, but progressively less so.
> Carriers that want to support IPv4-only customers and/or customers with IPv4-only devices will have to start
> providing translation to reach IPv6-only web sites.
> More and more web sites will start turning off their IPv4 services depending on eyeball providers to do
> translation where necessary.
> T+10 years:
> IPv4 will start to disappear from peering sessions.
> T+15 years:
> Hardly anyone will use IPv4 outside of their local institution.
More information about the RPD