IPv6 Transition to Cost US $75 Billion?

darthcamaro writes "There are alot of reasons why the US isn't moving as quickly as Japan and Europe in migrating to IPv6. One of those reasons is likely cost. An article on Internetnews.com cites an unreleased 'Dept. of Commerce report estimating it will take $25-$75 billion to pay for the transition.'"

Comments

Cost vs investment vs opportunity vs efficiency

[shanen]

Exactly the same kind of foolishness that keeps the US from going metric. If you prefer to see it as an opportunity to invest in new metric tools or IPv6 hardware and software, then it looks like an opportunity. The people who fight against such changes want to harp on the total costs, and generally refuse to consider rational transition strategies.

To me, it mostly comes down to efficiencies. The reason we measure things in the first place is so we can perform mathematical operations on the resulting numbers. Metric units are easier and more efficient for the mathematical operations, and therefore confer some competitive advantage on the people using them. It might be a large or small advantage, but it's always there.

IPv4 has some design limitations. IPv6 will address many of those problems, and the networks (and countries) that use that system will have competitive advantage.

What I find amusing is that many of the same people fighting against IPv6 on grounds of cost are the same people who want to argue the damage of Hurricane Katrina wasn't so bad. After all, it will give us the "opportunity" to invest billions of dollars in rebuilding things. Hey, why don't we destroy a major city every year? Look at all the "opportunities" we'd have. However, moving to IPv6 is NOT to be equated with random destruction, but is rather a rational form of evolution.

Re:What's needed?

[Michael Hunt]

You don't seem to know what you're talking about.

Any ISP with 100k customers (or even one with an order of magnitude less) is going to be assigned a /32 (or shorter) prefix, which is guaranteed to be globally portable.

The basic structure of an IPv6 address is:
0-31 Top-level network bits
32-47 16 bits for customer allocations (/48)
48-63 Customers' subnetworks
64-127 Local subnet addressing (EUI64)

If you've been allocated a /48 by your ISP, sure, you'll need to renumber every time you change ISPs. If you've been allocated a /32 or shorter prefix by a RIR, then you won't.

BGP4+ Routing tables will also be correspondingly smaller, because they'll only contain a number of /32 prefixes (a much smaller number than the current IPv4 soup, which includes prefixes as long as /24 for legacy reasons.)

I humbly submit that you do more research in future.

Custom Software Upgrades are Expensive

[billstewart]

Yes, lots of the equipment gets old and eventually needs replacing, but the government really does keep equipment around long after it'd be obsolete in the commercial world - after all, a desk grunt who's typing memos at 100wpm and sending a bit of email is only generating information at ~100 baud, and as long as you stay off the Upgrade-Microsoft-Office-Every-Year treadmill, the main reason not to be using a 386 PC is that too many web pages want newer memory-hogging browsers, and even upgrading to a 3GHz Pentium doesn't mean you need a bigger router for the office if he's not downloading a lot more material.

But upgrading custom software is a much different scale of project than simply upgrading boxes and reconfiguring some web servers. There's a huge amount of mission-critical big nasty badly-documented stuff out there running on mainframes, PCs, and Unix boxes of various sorts that knows about IPv4 and might or might not know about DNS and DHCP. Finding all of it isn't quite the same level of effort as finding Y2K bugs, but it's still a huge hard-to-estimate job.