[AusNOG] My Predictions for the ISP Industry

[Geoff Huston]

On 15/03/2012, at 11:12 AM, Mark Andrews wrote:
>> 
> 
> This will be very much like removing A records when you have a MX
> record when the A record is only there for SMTP.  For years people
> put A records into the DNS so that old (not MX aware) MTA could
> reach a site.  At some point you say I don't think we need this A
> record anymore so you remove it and wait for complaints.
> 
> Similarly you bring a new service and just give a MX (AAAA) record
> and see what happens.  After a while you just don't configure legacy
> support as you know it is no longer a issue you need to deal with.
> 
> Site with really old IPv4 only clients will install hacks if they
> want those machines to connect to IPv6 only services or they will
> live with not being able to reach those services.  Remember most
> machines installed in the last 10 years are IPv6 capable, you only
> need to turn on IPv6 for them to use it.
> 

It's unclear, and this thread is a good illustration of that lack of clarity, whether the industry will simply do the necessary measures to support IPv6, particularly in the mobile sector which is the current highest sales volume and highest value sector of the industry, or whether it will head down a CGN path. Maybe some data about what is out there and which way its trending would help.

At APNIC we've been measuring IPv6 penetration for some months now, and with some generous assistance from Google, ISC and the RIPE NCC we've now managed to extend client-side testing of client's capabilities to fetch IPv6 web objects to a measurement system that now tests of the order of 500,000 clients per day across the Internet. 

The worldwide picture is that some currently 0.4% of clients out there will prefer to use IPv6 when given the choice, while some 5.8% of clients show themselves of being capable of end-to-end IPv6 when presented with a V6 only URL. (http://labs.apnic.net/ipv6-measurement/Regions/001%20World/)

Australia is a little better than the global average, with 0.83% of clients shows a preference for IPv6, and some 4.1% of Australian clients shows themselves as being capable of fetching an IPv6-only object. (http://labs.apnic.net/ipv6-measurement/Economies/AU/)

If you have your own ASN you can also use this site to show the total v6 capability of clients who are within that ASN - (for example, Internode's clients' IPv6 capability can be seen at http://labs.apnic.net/ipv6-measurement/AS/4/7/3/9/index-months.html)

The general picture is that the IPv6 numbers are really quite low numbers at present, but the next year or so will be fascinating to observe. 

I wonder how long IPv4 will hang around IF we start to push investment into IPv6 infrastructure - I hear these claims that IPv4 will persist forever, but frankly I have heard in the past the same claims being made about X.25 and Decnet, or FDDI for that matter, and I guess there are folk who would say the same about ATM. However my suspicion is that this industry is definitely not sentimental - as soon as interest and more importantly the money wanes the technology rapidly disappears! 

So the real question becomes one of trying to understand, preferably with reliable data, what everyone is doing. At this point opinions are fun, but its data that is really valuable. If you run a web site, you can help us to measure the IPv6 capability of your client base, even if your site is IPv4-only today. Details of the IPv6 measurement program are at http://labs.apnic.net/measureipv6/, and if you want to help us  and inform yourself about the IPv6 capabilities of folk who use your web site, a way to do this is at http://labs.apnic.net/tracker.shtml

cheers,

   Geoff