In the Best Practice RFC 7772 – Reducing Energy Consumption of Router Advertisements for IPv6:
This issue is of concern IOT-class hardware that looks to be small, autonomous and battery powered. While this is just one of changes to our current technology that will adapt the networking fundamentals for IOT (look at the evolution happening around the 802.11ah WiFI standard for low powered devices) its a sign that even IPv6 will continue to go through a great deal of change which further undermines confidence in deployment.
Why ? Because IP is the unchanging, undying, rigid constant in networking.
This issue got me thinking about the design of IPv6. Last week I recorded a podcast with Geoff Huston who has a long history with various groups that are the custodians of the Internet such as APNIC, RIPE, IETF etc. In this podcast, he made the point that the design of IPv6 was never intended for the use cases of 2016. The original goals were simple – increase the size of addressing fields.
Then because adoption was slow people started fiddling with IPv6 to “fix” aspects of the IP protocol like endpoint address security, gateway auto-discovery, multi-gateway etc etc etc etc etc etc etc etc. In fact, the IETF working group is still mucking around with fundamental aspects of IPv6 today.
In the last two decades, adoption of IPv6 has been slow. Even the exhaustion of the IPv4 address space didn’t kickstart a revolution in adoption – Google charts show less 9% adoption at time of writing.
This number is questionable, “availability of IPv6” not actual IPv6 used. Real IPv6 use is likely less than 5%.
Fifteen years of doomsaying, calamitous predictions, beating of chests doesn’t hide the fact that IPv6 is unwanted. And Google’s chart may be on the generous side, real adoption seems much less.
I believe that the problem with IPv6 lies with the standards bodies, particularly the IETF and their endless tinkering without care for the consequences of endless change to the IP Protocol.
Change can be Bad
The concept of an infinitely changing IPv6 is a serious issue. Networking technology has been stagnant for since 2001 and left the market ill prepared for change much less the upheaval that we see today. The rate of change has accelerated since 2010 with the arrival of network virtualization & SDN there one specific area which has remained unchanged for decades – the IP Protocol.
I was considering what has changed in the last twenty years. We’ve seen changes in LAN where ATM/FDDI/SMDS/Frame Relay has been replaced with Ethernet. Protocols outside of TCP & UDP have died off (not that there were many anyway) and today.
Related: The Evolution of Layered Protocol Stacks Leads to an Hourglass-Shaped Architecture is a research paper that explains this clearly and in detail.
Ten years ago, application protocols like FTP, SIP and RTP were common on the Internet, today, just three protocols matter: HTTP/S, DNS and SSH.
The IPv4 protocol has been unchanged for more than twenty years and, in my opinion, this immutability has the key its success & longevity.
The Isolation of Networking Discipline
What can we read into the IPv6 resistance and the collapse of networking as a solution space ?
- Networking exists in isolation. Interaction and partnerships with operating systems have been poor.
- Developer Alienation: It has been far easier for developers to extend existing protocols than develop new ones. The development of HTTP2 that took years of IETF processes before a final standard was published.
The most likely situation is that networking is not a consideration for developers. In previous years,
Its Not About Addresses
IP Address Integrity is Zero. There is no integrity of IP addressing in the Internet. Long running attempts by the IETF to implement BGP security (BGPSec or RPKI) have been blocked by service providers who refuse to invest in their networks to provide secure route propagation and route validation. The result is that anyone can hijack IP routes with little fear of detection. IPv6 offers no more integrity (that is ZERO) than IPv4.
IP Source Integrity. The endemic use of address hiding technologies such as network address translation, forward proxy caches, CDNs (reverse proxy caches) means that using source IP addresses for identification has little value. Services that offer geolocation do not not put high value on the source IP Address but look for correlating data points like WiFi MAC addresses. Geolocation databases are rumoured to be less than 50% accurate.
Names matter more than addresses. Its a fact. One IP address:TCP Port can host hundreds of IP services. Software developers got tired of waiting for networking and just started using names instead of IP address.
Web servers allow one IP address to present virtual hosts with different domain names. Email servers use a single IP address for multiple email domains. The key factor is the name of the host that determines the SSL certificate to provide connection assurance.
Usability. IPv6 has serious usability problems. Its too long for the manual use like IPv4. The rules of representation for abbreviation are so arcane & stupid that it actually prevents day to day use.
The EtherealMind View
We could conclude that newer IPv6 addressing are not strictly required. Its true that we are faced with billions of new devices being connected to networks and the demand. Lets take predictions that 2 billion cars need IP addresses. This is true but cars might connect to an APN on 3G/4G network with private IPv4 addressing and pass through a gateway so that network security is maintained. Its not mandatory that every device in every car to have a public IP address. The same holds true for M2M/5G networks.
There are many things about the design of IPv6 that are broken and often the same things that were broken in IPv4.