Yahoo IPv6 Upgrade Could Shut Out 1M Users

alphadogg writes "Yahoo is forging ahead with a move to IPv6 on its main Web site by year-end despite worries that up to 1 million Internet users may be unable to access it initially. Yahoo's massive engineering effort to support IPv6 — the long-anticipated upgrade to the Internet's main communications protocol — could at first shut out potential www.yahoo.com users due to what the company and others call 'IPv6 brokenness.'"

Comments

Re:Real question is...

[longacre]

Not sure how or why, but they still get a shit-ton of traffic and Yahoo Mail has 3x as many users as Gmail.

Great logic there Lou

[OverlordQ]

From

IPv6 experts say some Internet users will experience slowdowns or have trouble connecting to IPv6-enabled Web sites because they have misconfigured or misbehaving network equipment

to

"IPv6 brokenness."

So I should blame the water company if I install my plumbing wrong?

Re:Great logic there Lou

[gstoddart]

So I should blame the water company if I install my plumbing wrong?

No, but if they changed their infrastructure to no longer be compatible with your existing (and working) plumbing and expected you to pay to upgrade, you'd be mad, right?

One of the problems with IPv6 is everybody already has networking equipment that they've paid for and that works ... I can't see much motivation for most people/organizations to switch to IPv6, especially if it means it breaks what they've already got. I can also see making everyday things like ping and telnet much more cumbersome.

All of the people with home routers and the like (and older operating systems) don't want to pay to upgrade for something which they don't understand what benefit it is supposed to give them. I must confess, except for a bigger address space, I'm not sure what benefit IPv6 has for *me* -- which is why IPv6 has been languishing in the "don't care pile" for seemingly forever.

If my ISP needs to change all of their cable-modems to support this, you can bet I'm gonna have to pay out of pocket.

Re:Great logic there Lou

[idontgno]

No, but if the water company switches to IPv6 water and your plumbing is incompatible, you might blame them. After all, water is water, and the plumbing worked just fine until they changed something.

Or, to put it in light of a bit of recent history, a lot of Americans are still grumbling at local broadcasters and the FCC because over-the-air TV was working JUST FINE until June 2009, when ALL OF A SUDDEN the rabbit ears weren't enough. And that was with a sustained, repetitive, annoyingly pervasive advertising campaign to raise awareness of the upcoming DTV transition, plus subsidies for converter gear. And a distinct minority still missed the transition. And that's just broadcast TV, which is stupid simple in terms of end-user infrastructure.

IPv6, implemented piecemeal, will simply black out parts of the net to many (most?) users until something like the DTV transition makes it (A) obvious to Joe Intarwebuser that the transition IS UNAVOIDABLY COMING, and (B) subsidizes replacements of incompatible key components of the users' and providers' network path. (I'm looking at you, manufacturers of residential gateway router devices and network ISPs.)

Re:Great logic there Lou

[klapaucjusz]

I think you're missing the point.

We're not speaking about switching to pure IPv6. We're speaking of making Yahoo accessible over both IPv4 and IPv6.

Pure IPv4 ("legacy") sites will have no problem, they'll just contact Yahoo over IPv4. Properly configured dual-stack sites will have no problem, they'll have a choice between IPv4 and IPv6. It's only mis-configured clients that might have problems.

The article claims that 0.05% of Yahoo's customers are mis-configured. These 0.05% will need to either disable IPv6, or fix their systems. --jch

Yahoo! is relying on old, incomplete data.

[Above]

Yahoo! has been talking about this at conferences for a while, but I'm not sure they are using good data. Here's a lighting talk from NANOG about it:

http://www.nanog.org/meetings/nanog46/presentations/Tuesday/Igor_ipv6_recursive_light_N46.pdf

Page 2 has the crux of the issue, Yahoo! claims if you add AAAA records that 0.078% of the user base "breaks", that is they understand a AAAA enough to try IPv6, but they lack IPv6 connectivity to the destination.

There was a time this made sense. A lot of early IPv6 deployments were islands without complete connectivity. Additionally, up until about 18 months ago there was a serious lack of IPv6 interconnectivity between ISPs, they were still figuring out how to turn up peering, filter, and so on.

However, times change. ISP's are now fairly well interconnected, and getting a lot better every day. Almost no one turns up IPv6 as an island anymore. Interestingly, some of the original islands still exist, on purpose, as they are test labs or other non-production deployments. The people use them expect them to be broken in some ways, in some cases to test what the user experience is when various things break. Indeed, I suspect the number of islands is small, and constant, and thus an ever decreasing percentage of the IPv6 user base.

Another large issue with the numbers is that they are only measuring the difference between the status quo and one of the four outcomes. A user could have:

A) Broken IPv4, Broken IPv6.
B) Broken IPv4, Working IPv6.
C) Working IPv4, Broken IPv6.
D) Working IPv4, Working IPv6.

What Yahoo has done is measure the status quo (IPv4 only) to bullet point C.

However, there will be some folks in bullet B. These are folks who can't get to Yahoo! today at all, but would be able to if Yahoo! had AAAA's. Granted, it's probably smaller, but still is an offset. Basically they are trying to scare folks that 470k folks might not be able to access Yahoo with IPv6. However, 470k folks may already be unable to access it via IPv4, they just can't measure that right now because they never see the requests!

There is also the looming issue. As a we run out of IPv4 addresses (likely in late 2011) ISP's will basically be forced to turn up IPv6 only users. Even if you take Yahoo!'s numbers as correct, that 0.078% are broken, then all you would need is a larger percentage than that of the user base to be IPv6 ONLY and it makes more sense to have AAAA's and exclude them. Basically 1% deployment of IPv6 completely flips their argument if the goal is serving the largest number of folks.

My take, some folks inside Yahoo! collected some rather raw data early on in IPv6's life cycle. Folks from Marketing and such read too much into it, and went into a panic that some large number of users wouldn't be abel to get to Yahoo! This created a huge issue for the engineers trying to deploy IPv6, which they have been fighting ever since.