The conversation

Lessons learnedExperience excerpt · English

I have three connections at home both as a way to increase available bandwidth and also to provide …

I have three connections at home both as a way to increase available bandwidth and also to provide some redundancy since I work from home and my ability to earn income is tied to my internet being up. I have two primary connections that I load balance between, each providing a /56. If either fails, all traffic goes to the other. I have a third connection that is used only as a backup in the event that both the other connections go down, providing a /64. So simple case, I assign an address from each connection out to each device. Except, whoops--the backup ISP only provides a /64 and despite that containing something like 18.5 quintillion addresses (a number so large my spell check is telling me it's not a real word), I can't actually split those up in any meaningful way so I can only allow _one_ subnet to access the internet via that connection. And then I'm pushing responsibility for load balancing between the two primary connections out to each client. You know, to the place with the _least_ useful information for making decisions in this situation. …

Original discussionIPv6 Is Hard438 source comments · at import
2

What was this responding to?

Parent context · original excerpt

CY

IPv6 is really an example of a second-system syndrome. It really _is_ hard because it makes a lot of IPv4 conventions inapplicable. For example, if you want to do multi-homing with IPv6 for an SMB, you need to either do NAT (not great), or have your router advertise/whithdraw prefixes dynamically. If you do the prefix withdrawal, you can end up in a situation where your devices have _no_ prefixes. So even your internal functionality like printing or VoIP starts breaking. Great. Then there's ULA that is mis-prioritized. You'd think that simply adding a stable ULA prefix to all devices would solve the issue above, but that's not the case. The devices will prioritize the global IPv6 source address, so stuff can _still_ break.

Full original

Community replies 0

No community replies yet. Add your perspective.

Replies from the source 2

Direct replies available in this import. Original-language excerpts.

CY

There is an RFC proposal to allow SLAAC on networks that are smaller than /64, so you might be able to utilize the third connection and still have an ability to do subnetting. Sometime around 2035.

Full original
NI

I think “and we’re gonna kill NAT too!” is precisely where IPv6 overstepped and started making things worse for itself. It didn’t need to be this way. NAT sounds like a bad idea at first glance, but it doesn’t exist just to alleviate address shortages (although it does.) It doesn’t just exist to provide security by default either (although it does that too.) NAT is also the only sane choice in residential setups. Because your public IP address prefix isn’t static. That was never going to happen. And just saying “never put an IP address in any configuration file” isn’t a satisfactory answer. When a reprefix happens, and your clients are all only speaking on GUA addresses, your network is going to be fucked until they expire the old addresses. Not in theory, in practice. None of my devices in my household did any sort of reasonable thing when Comcast changes my prefix. I had to just go aro…

Full original
Continue on Hacker News

Respect different perspectives. Share concrete experiences.

Sign in to post, reply, repost, and quote.