IPv6 websites accesibility guess. The GC Index.
The origin of this work was an informal proposal of Fernando Garcia Fernandez (TECNOCOM fella) in May 2006, in the middle of one of the social events at RIPE-52 meeting in Istambul, about doing some "real" and "useful" measurement that showed if IPv6 is really going to be deployed or not. Just guessing what our friends or relatives (specially those not technically skilled) would find in a world without IPv4 addresses, and only using IPv6 for everyday Intenet's life. And of course, present it at RIPE meetings.
For the first talk we did all the analysis by hand, and presented it at plenary session, October 3rd 2006, at RIPE-53 meeting in Amsterdam. Then updated the talk in Tallinn, RIPE-54, and after that we tried to automate the measurement process, using some scratch servers with decent connectivity at BT Spain Networking Labs, and refined the procedure and the script from time to time. We did some further talks at RIPE-55, RIPE-56, RIPE-58 and RIPE-59, and also at some ESNOG/GORE meetings in Spain, until we decided that showing day to day evolution of the measures through the web site would be enough.
The base URL list was formerly obtained from ALEXA.COM 500 TOP sites webpage,
that we arranged to get the first list of domains used to do the study, always prepending the domain name with a set of v6-ish prefixes
that the webmasters seemed to use to characterize the v6 flavour of the servers:
At March 13th 2008 we increased the base list merging with the not-repeated URLs from the top-100 ALEXA lists of 120 countries and also included the graph of the ratio "v6 hosts/total" to be independent of the number of domains explored.
At April 1st 2009, Alexa began to provide a global list of top 1,000,000 sites (from which we use the first 15000), so maybe a bump will show in the total graph, but (hope so) not in the Ratio graph.
Then at May 18th 2010, we started to resolve google.com and youtube.com domains directly in v6 (forwarding DNS traffic to Hurricane Electric DNS servers), so the number of hosts-per-domain have hugely increased, and a new bump in the graphs appeared, although the RATIO should should have been in the same range as previously. This bump lasted until July 9th 2010, and during this period Google seemed to be assigning one ipv6 address to each of the search pages in every ccTLD (google.ca, google.co.uk, google.va, etc). After that period, the ipv6 assignment changed to be done by blocks (e.g., 2a00:1450:4007:807::101f is the single ipv6 address google.de, google.uk, google.rw, google.com.na, google.mu, google.mg, etc. are resolved to) It would be an interesting exercise to guess the purpose of this distribution (although maybe this has been already explained in some blog/forum/list). And so until today...
The per-country results are available HERE
The complete list of URLs tested is HERE
The raw results file is HERE
The usual complete list of ipv6 sites detected, ratios and MX/DNS additional info, can be seen HERE
(This page is updated daily, the process finishing at (around) 07:30 CET)
Please, wait while the graphs are being built...
|Results: 60000 URLs scanned, 1372 Unique URLs with 1878 IPv6 addresses reachable|
1671 DNS servers and 1949 MX records with ipv6 detected
The Ratio of IPV6 hosts are: Found = 2.29 %, Reachable = 3.13 0
This process started the Wed Oct 22 01:59:02 CEST 2014 and finished on Wed Oct 22 08:27:48 CEST 2014
For contact, suggestions, or any info about this, just drop a mail to