Always better to be rich and healthy than poor and ill

If you're reading this, I have very little doubt you've seen the news about Apple announcing the requirement of IPv6 support, moreso, IPv6-only networks support, for all new apps. If you haven't seen that - the blog entry at Deploy360 does a great job summarizing it.

Also, you might have read the slightly disappointed reaction from Geoff Huston related to this subject. In this modest opus, I'll try to formulate (as much as my current #CLUS jetlag allows) my opinion about these news.

First of all - I completely agree with Geoff it would be great if we had dualstack tethering working, and a less russian roulette when it comes to their RFC6555 implementation, and... a lot of other things.

The reality though can be summarized by this phrase, that I translated from what my dad tends to say whenever there is a discussion about the tradeoffs - "It's always better to be rich and healthy than poor and ill". Isn't it obvious, you ask ? Of course, that's the point.

It would have been better to have all the unicorns we want, but the reality is that whereever was that magic wand that Apple has found, it was only one, and it had only one charge. One wish, with a simple elevator pitch.

And within this set of restrictions, I believe requiring the IPv6 support for all the new apps is the single most powerful thing that they could've done. Why ?

1. The developers started paying attention. IPv6 is no longer that optional thing that we do maybe, sometime, if we get someone asking for it. Or not at all. Whatever. No. IPv6 is something we have to do here and now. This puts the maturity date on the technical debt that the lack of IPv6 support is. With every fifth user in the US having IPv6, it's something that *has* to be treated this way.

2. This gives a massive business case for the IPv6 support in the backend infrastructure. AWS, Azure, Google Compute, etc. - today they do not support IPv6. Thus, have to be supplemented by an additional offering (For example, Cloudflare was quick to use the opportunity, but I've no doubt Akamai might follow suit - because they have similar functionality as well).

   Needless to say, this creates an opportunity for the existing customers to take a step back and reevaluate: is the ongoing cost of IPv6-to-IPv4 bandaid worth it, or maybe it's worth moving infra somewhere with IPv6 native support ? You don't need to have a PhD in Marketing to see that this is quite a tricky moment when your customer starts shopping around.

3. Apple itself. Every large company is an entity with complicated internal machinery. So, "they" in one part of the company is not the same as "they" in another. Having a powerful externally facing messaging will force the improvements internally. You can find an external competitor for most of the internal apps, especially within the very important collaboration/messaging area.

4. This opens a future possibility for migration options other than dualstack, namely, IPv6-only+NAT64. You will of course laugh me off, saying I am dreaming - no-one does it, right? Almost. At FOSDEM, for the past two years, the main SSID on the network did not have IPv4. Yes, of course we had fallback dualstack. But the default was IPv6-only. And it's not in the report, but this year ~40% of hosts hung on to the default IPv6-only SSID. Absolute numbers are in the multi-thousand range.

   And it was in January, half a year before Apple's announcement and just after Lollipop has shipped RDNSS support, allowing for IPv6-only operation as well.

   Not having dualstack nicely solves another problem, namely, having to second-guess the implementation of Happy Eyeballs by the operating system(s). That, and also not having to maintain/monitor/track two address families. With this announcement, I expect next year we should get a bigger number on the IPv6-only WiFi - and hopefully pave the way for a more mainstream deprecation of IPv4 at the access layer.

5. This event makes an IPv6-only discussion within *your* organization much more possible, depending on your fleet. (Shout-out here goes to Cameron Byrne, all the authors of RFC6877, who pioneered large-scale production IPv6-only).

In conclusion

Assuming a perfect world with rainbows and unicorns, could they have done more ? Certainly. Assuming the single-charge magic wand, could they have gotten any more impact from any other single action ? I don't think so.

P.S. Listening to session Your App and Next Generation Networks makes me even more convinced of the above. It's better than 464XLAT. It's your ticket to IPv4-free world.

Index of /blog/2015-06-12-Always-better-to-be-rich-and-healthy-than-poor/