The following video is an excerpt of a Google hosted discussion of the state of IPv6 on the Internet. I don’t recommend you see it even though the discussion is orchestrated by none other than Vint Cerf himself.
For something like IPv6 which is supposed to be “the future of the Internet” there seems to be a lot of religion out there. Let me try to summarize what I thought were the main reasons to go with IPv6.
- A Cisco representative in the panel suggested that ISPs were willing to add more customers to their network than allowed by their IPv4 address space budget. According to some other comments, the real battle is between the cost of acquiring more IPv4 addresses and the cost of IPv6 deployment. The way I see it… its an opportunity for Cisco to make money out of address scarcity. There seemed to be a difference in opinion between the Juniper representative and the Cisco representative about whether the current routers can handle IPv6 traffic. The Juniper representative in the panel said that recent routers should be fine… the Cisco representative thought that a complete network assessment was needed before such a decision could be made. This re-emphasizes my point about Cisco trying to make money out of this whole IPv4 -> IPv6 ordeal.
- The other point seemed “religious” for the lack of a better word. It was about end-to-end connectivity on the Internet. The IPv4 network currently has a lot of its edge clients behind NATs. Some speakers really thought that this was “bad” without giving any particular reasons. They stated that opening up home devices to the outside Internet would bring in “a lot of innovation”. The only thing I draw from this (which is the reason I posted this article in the first place) is that IPv6 opens up an opportunity for Google to crawl through the data behind NAT devices and promote the Free Culture without going through massive infrastructure investment like Youtube.com. What really bothers me is that nobody brought this point out as clearly as the security point. End-to-end connectivity is not just manifests a security stance but also a political stance and benefits companies which thrive on user-generated / free content.
Of course a number of people in the audience pointed out the difficulty of getting on IPv6 in the first place. Besides the address expansion, most of the benefits of IPv6 have been back-ported to IPv4 already. The chicken and egg problem was often sited:
- ISPs don’t want to transition to IPv6 because having no network destinations on IPv6 doesn’t justify putting in the infrastructure cost.
- Content sites don’t want to transition to IPv6 and incur infrastructure costs unless there is enough of an audience which is IPv6 only.
The funny point here is that (if we are to believe that network ISPs and content sites are close collaborators… which I think is true for cable Internet) then they have business incentive to keep the Internet in the content creator vs. consumer divide for as long as possible. It may not be a bad idea for the content sites to subsidize the IPv4 address scarcity to keep Google and entertainment startups further away from making available even more user generated content without investing heavily into infrastructure.
I understand that it may be impolite to directly discuss conflicting business interests publicly in a conference hall like setting. However, religion is hardly a substitute. From a consumer standpoint I don’t want to see my Internet bill go up by $5 or $10 for the IPv6 transition only to help some companies make more money. There should really be choice and transition costs must be paid only by the true beneficiaries of the IPv4 to IPv6 transition. Network connectivity at reasonable costs is more important than “elegant” design decisions in a practical network such as the Internet. May be the entire focus of the next generation IP should have been just laying out a killer transition strategy for expanding the address range. IPv4 is enough of a testimonial on how protocol specs can be improved upon incrementally without breaking the network or requiring the administrative overhead of a dual-stack.