[mikecb]

Just for everyone focusing on gimmicky features, this is an OnHub, which are intended to just work.

Under the hood, they run ChromiumOS, including things like trusted boot (yes, this home router has a tpm), and the A/B partitions so when you get an automatic update, (which happens during a time of low bandwidth usage), it reboots into the new version in about 6 seconds. The security team is awesome: they pushed an update to all onhubs within 48 hours of public disclosure of a critical RCE earlier this year.

There has been some slowness to expected features, like ipv6, but the PMs have been clear about their goal: they won't include a feature that's buggy. It's exactly what you want in a piece of infrastructure.

[modeless]

Wait, OnHub doesn't support IPv6?! That's inexcusable. There's no reason IPv6 couldn't "just work" too. Google Wifi had better support IPv6 at launch.

[givinguflac]

If you require IPv6 so crucially, you probably shouldn't be using consumer grade gear.

[mmagin]

I've been using IPv6 on a consumer-grade cable modem connection with a Cisco DPC3008 modem and a previous-generation Apple Airport extreme since 2012. There weren't even any difficulties getting it working.

[calinet6]

There aren't any difficulties if you don't get it working either (yet). The Internet still works with no issues.

[detaro]

There are quite a few providers in the world that run Dual-Stack Lite (native IPv6, IPv4 is tunneled to provider-run NAT gateways) and they all hand out consumer-grade gear.

Since the OnHub only comes with Ethernet you'd have to put something between it and the wire anyways of course, but it then has to be more than a dumb modem, and you'd be forcing all traffic through the tunnel.

IPv6 isn't without problems, but many people use it without noticing, through consumer devices way cheaper than this. The ones complaining the most are the enthusiasts which suddenly can't reach their home server via IPv4 anymore.

[toomuchtodo]

My ~$100 Linksys router supports IPv6. What's Google's excuse?

Yet another device Google will sell, and then discontinue in 24-36 months.

[brunoqc]

> Yet another device Google will sell, and then discontinue in 24-36 months.

You don't know that.

[toomuchtodo]

Besides acquired products (Nest) and Android phones, Google has a pretty poor track record of long term hardware support. I can still get firmware updates for my 4 year old Linksys router.

I've been burned, and I refuse to purchase yet another Google device for a market they're just testing the waters in. I'd rather spend my money with someone who needs it to support their product, and that product is their business.

[mikecb]

To be clear, you're saying they're just testing the waters with Chromium? Or with Fiber, which makes its own hardware STB/routers, or with this whole internet performance thing in general?

Chrome devices have 5 years of updates guaranteed. I have a cr-48 from 2010 which still gets updates (currently on -dev 56). How many times has there been a story on HN about RCEs unpatched for months or years in consumer routing gear? The fact that you purchased that hardware doesn't seem to be sufficient incentive based on the record. But when you have an incentive of a secure and performant internet, and also have the talent, infrastructure, and existing codebase to make something work and work well, you have economies of scope that make supporting it much easier and cheaper.

[p1mrx]

IPv6 is for everyone: consumers, businesses, whatever else exchanges packets.

[Jonnax]

That kind of thinking is 10 years out of date. Consumer broadband routers sent to subscriber homes for free support IPv6.

[kardashian007]

Gotta love easily NSA-dioded consumer gear which falls over under load, reboots randomly and doesn't play nice with standards.

[jbb555]

Really, it doesn't do IPv6?

How could they launch such a faulty product. If it was 3 years ago maybe, but it's not really optional now.

[ryanlol]

>That's inexcusable

Except it's not, we live in an IPv4 world.

Edit: How are you guys downvoting me on your v6 connections? HN only supports v4.

[modeless]

Maybe you haven't checked lately but we live in a world where IPv6 adoption is over 10% (over 25% in the US) after doubling each year for the past 6 years. https://www.google.com/intl/en/ipv6/statistics.html

[mikecb]

It's on the roadmap, but so far hasn't passed QA testing.

The engineering time is going into upstream too, so ath9k, hostap, and other big networking libs are benefiting from this.

[Shorel]

Projection:

Jan 2016 -> 10% Jan 2017 -> 20% Jan 2018 -> 40% Jan 2019 -> 80% Jan 2020 -> 100%

Seems quite good to me.

[ryanlol]

Why does IPv6 availability matter here? The amount of users that need, want or even directly benefit from IPv6 is vanishingly small.

[deathanatos]

I disagree. Directly, yes, the average user doesn't care. But application developers care; peer-to-peer protocols become a lot trickier with IPv4 due to the pervasive deployment of NAT; two machines ostensibly on the Internet can't connect to each other, requiring instead the use of STUN servers, which then requires infrastructure somewhere, or just doing it client-server, or some mix like having "supernodes" (like Skype, prior to MS tearing it out) that route traffic for NAT'd devices.

The ability to actually connect arbitrary devices, I hope, will be something that people will take advantage of. I know for many game servers I set up with siblings, the ability to not need to mess with a router's crappy "port forwarding" would be a welcome change. (Even if I had to mess w/ some local firewall, but that can perhaps be much more tightly integrated or at least, a better UX.)

[stock_toaster]

Yes, but until you have the percentage of overall IPv4 usage down to less than 5%, 10%, or even being generous say 15%, developers will still have to deal with those things (NATs, STUN, TURN, etc) anyway.

IPv6 has been around for almost 20 years now, and is only recently cracking 10% (and I wonder how much of that 10+% is also dual stack). IPv4 sure as hell isn't going away in my lifetime. Who knows, maybe the lifetime of my kids too. What a mess!

[tomschlick]

Because the quicker we adopt it the better.

The shittiest of routers support it so when you get one from a major internet company you should expect that it has support for an internet protocol which has been out for 18 years.

[ryanlol]

Doesn't qualify as "inexcusable" for me, sorry.

[jrockway]

IPv6 is important, but it's not like your ISP will lower your bill if you forego IPv4, and it's not like there are IPv6-only sites. If I recall correctly, I have to pay _extra_ for an IPv6 address.

[thwarted]

IPv4 addresses have been more expensive than IPv6 subnets on every dedicated uplink I've gotten prices for over the last two years. In fact, an IPv6 /56 is usually free, or cheap enough to be effectively free since the fee is mainly a NRC for the time to set up the route, if you ask for it and the provider supports IPv6. In contrast, IPv4 addresses often incur a MRC based on the number of usable addresses you request.

[lucb1e]

> If I recall correctly, I have to pay _extra_ for an IPv6 address.

That's weird, here we get a /48 v6 block by default and one /32 v4 address. You'll never need more v6 addresses, but each v4 address comes at a monthly fee.

[jbb555]

Most of my connections are over ipv6.

If this doesn't support ipv6 then it can only connect to legacy sites. Yes that's pretty much all sites today but this is inexcusable in a new product.

[madeofpalk]

Well you can never upgrade to a IPv6 world if you don't add it to routers...

[1_2__3]

Ah yes, deprioritize features then spin it as "we just want to make sure it's perfect", as if IPv6 is some kind of work of art or delicate baking.

[danielwang]

Just want to correct a little thing: TPM is used for measured boot, trusted boot is something else

[mikecb]

Ah, you're right. I was using the term loosely.

[hackuser]

What are the user privacy policies and technical implementations?