Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

GitHub seems to be enabling IPv6 for GitHub Pages

github.com

Post reply on HN
Print view View original

51–57 of 57 posts

Re: GitHub seems to be enabling IPv6 for GitHub Pages

#51

Post by jiggawatts »

aoetalks wrote: post #25
jiggawatts wrote: post #23

Earlier quoted context omitted.

It should be as simple as not being allowed to advertise a service as providing Internet unless it provides 100% native, full-featured IPv6. Imagine if in 2001 some company tried to sell an "Internet service" that didn't provide IPv4, or only provided RFC1918 addresses, or didn't allow something as trivial as non-NAT routing of IPv4, even as an option . They'd be sued for false advertising, and rightly so. Meanwhile,…

I run a full dual stack service in Azure [1]. I couldn’t get instance level IPv6 working, but everything else works in both v4 and v6. I will echo that many Azure services don’t support v6 yet, though. https://docs.microsoft.com/en-us/azure/virtual-network/ipv6-...

Microsoft voluntarily admits the following limitations on the page you linked:

• ExpressRoute gateways CAN be used for IPv4-only traffic in a VNET with IPv6 enabled. Support for IPv6 traffic is on our roadmap.

• VPN gateways CANNOT be used in a VNET with IPv6 enabled, either directly or peered with "UseRemoteGateway".

• The Azure platform (AKS, etc.) does not support IPv6 communication for Containers.

• IPv6-only Virtual Machines or Virtual Machines Scale Sets are not supported, each NIC must include at least one IPv4 IP configuration.

• When adding IPv6 to existing IPv4 deployments, IPv6 ranges can not be added to a VNET with existing resource navigation links.

• Forward DNS for IPv6 is supported for Azure public DNS today but Reverse DNS is not yet supported.

That doesn't even being to cover it! There's all sorts of other fun limitations scattered all over the place, you'd have to spend weeks scraping them together into a pile of shame.

Examples include:

• The maximum prefix IPv6 prefix size available is just 16 addresses, which is hilariously tiny and totally negates the core benefit of IPv6's huge address space: https://docs.microsoft.com/en-us/azure/virtual-network/publi...

• Forced NAT of IPv6, which entirely defeats the purpose and benefit of this protocol over IPv4.

• NAT doesn't interact with IPv6 address family. NAT can't be deployed on a subnet with an IPv6 prefix. https://docs.microsoft.com/en-us/azure/virtual-network/nat-g...

• If you add an IPv6 address to a rule, [Azure] firewall fails. https://docs.microsoft.com/en-us/azure/firewall/overview#kno...

• None of their PaaS services support IPv6 firewall rules. https://stackoverflow.com/questions/40461803/allow-ipv6-addr...

Etc, etc, etc...

Re: GitHub seems to be enabling IPv6 for GitHub Pages

#52

Post by weird-eye-issue »

patrakov wrote: post #49
weird-eye-issue wrote: post #48

Earlier quoted context omitted.

Sorry but this is just too anecdotal for me. There are so many factors that go into Google rankings and Google itself makes it pretty easy to discover any crawling or indexing issues on a website that it seems the root cause of the issue should have been discovered (if it was due to a crawl/index) and I doubt it was a properly working ipv6 setup (something broke somewhere on your company's end is what I'm getting at)…

Well, it was just a stock Hetzner dedicated server. All that was added/removed is a DNS AAAA record pointing to the server's IPv6 address. We used IPv6 before in order to access LXC containers hosted on that server. If THAT counts as a broken setup (which I doubt, see below), then, sorry, I don't see how we could do better. I have personally checked the result from my then-home connection (unfortunately, through he.n…

Having access to GSC is really a must for stuff like this. You can see exactly how Google sees your page and it's status (error fetching, not indexed, different canonical, etc etc)

Re: GitHub seems to be enabling IPv6 for GitHub Pages

#53

Post by Ericson2314 »

roenxi wrote: post #40
Ericson2314 wrote: post #19

Earlier quoted context omitted.

It would be nice to see this forced by law.

That is a bad idea. 1) Further entrenches the big players by making it harder to enter the market. 2) Making decisions for companies with no understanding of the costs is a bad strategy. If it was easy to support they'd just do it. 3) It is still arguable if IPv6 is even necessary. I think yes, but the situation isn't clear enough to legislate. Maybe some better solution will be found when the pressure is on.

I strongly disagree

> Further entrenches the big players by making it harder to enter the market.

On the contrary, I think dual stack from the ground up is cheap. Big corps are reluctant to switch because it means paying the "interest" in their tech debt.

> Making decisions for companies with no understanding of the costs is a bad strategy. If it was easy to support they'd just do it.

Who said we don't understand the costs? You are underestimating the stringiness of the bean counters that got us into this coordination failure. You are also assuming the aggregate demand for abstract public goods like IPv6 availability could ever be sufficient in already-developed countries for market-based solutions to work.

> It is still arguable if IPv6 is even necessary. I think yes, but the situation isn't clear enough to legislate. Maybe some better solution will be found when the pressure is on.

IPv6 is necessary for a bunch of decentralization to work better. Thks isn't good for profits (no, "crypto" isn't the decentralization I have in mind), but it is good for society. This is clear to me, but the exact sort of abstract thing which is hard to put political weight behind.

Re: GitHub seems to be enabling IPv6 for GitHub Pages

#54

Post by jrockway »

chrismorgan wrote: post #50
Gigachad wrote: post #33

Earlier quoted context omitted.

It's not that hard to say anyway. You read it as "quad a".

Fun fact: I mulled over whether to write “an AAAA record” or “a AAAA record”. I ended up deciding to go with “an AAAA record” even though I would read it “a quadruple A record” which is physically easier to say than “an A A A A record”.

Hah, I spell out the As -- "an A A A A record".

I also call it "w w w" and not "dub dub dub" as some people do. Though I feel neither come up in conversation all that often.

Re: GitHub seems to be enabling IPv6 for GitHub Pages

#55

Post by remram »

profmonocle wrote: post #31
jrockway wrote: post #27

Earlier quoted context omitted.

Maybe AWS, Azure, and GCP, but not places like Digital Ocean. IPv6 is still very niche, and it's suffering from that effect where it being very niche keeps it very niche. Nobody wants to be an early adopter (and momentum seems to be dying), and the result is that there's no adoption. I think we've been trying to make IPv6 happen since 1998, and ... it's not happening.

> it's not happening. It's absolutely glacial, but IPv6 adoption has been steadily rising over the last 8 or so years. Unfortunately the increase has been linear rather than exponential. That is, according to Google's stats of clients accessing their properties. Not a perfect snapshot of worldwide Internet users, but a pretty decent one. https://www.google.com/intl/en/ipv6/statistics.html

I mean, I don't have IPv6 at home (Verizon) and I don't have IPv6 at work (workstation or servers). It's frustrating.

Re: GitHub seems to be enabling IPv6 for GitHub Pages

#56

Post by remram »

est31 wrote: post #39
Gigachad wrote: post #34

Earlier quoted context omitted.

Even a linear growth is fine. By 2030 there will be pretty much 100% adoption at this pace. There may even be a tipping point when places like India are so stretched for v4 addresses and v6 is so widespread where local sites go v6 only which triggers a much faster adoption rate on the consumer side.

This website says that the 100% will be reached by 2030, but only if you use the quadratic model: https://www.vyncke.org/ipv6status/project.php?metric=p&timef... If you use the linear model, growth is slower and the 100% will be reached by 2038.

Why not go to degree 3, which predicts we'll be back at 0% by 2028.

Re: GitHub seems to be enabling IPv6 for GitHub Pages

#57

Post by chrismorgan »

jrockway wrote: post #54
chrismorgan wrote: post #50

Earlier quoted context omitted.

Fun fact: I mulled over whether to write “an AAAA record” or “a AAAA record”. I ended up deciding to go with “an AAAA record” even though I would read it “a quadruple A record” which is physically easier to say than “an A A A A record”.

Hah, I spell out the As -- "an A A A A record". I also call it "w w w" and not "dub dub dub" as some people do. Though I feel neither come up in conversation all that often.

In Australia I can’t recall hearing anything other than “double you double you double you”. (But if we could do a bit of language reform, shortening w, 7 and 0 each to a single syllable like all the other letters and digits would be the first thing I’d do.) But then, “double you” flows in sequence a bit better than “ay”.
Post reply on HN