5 ms·

I’ve often wondered if ipv6s less human readable addresses is the primary cause. I dread using and interacting with ipv6, compared to ipv4. Sure ipv6 solves a l

I’ve often wondered if ipv6s less human readable addresses is the primary cause. I dread using and interacting with ipv6, compared to ipv4. Sure ipv6 solves a lot of scaling challenges, but it is so much harder to grock at first glance.

ashton314 5y ago
Address abbreviation helps a bunch. IPv6 (not an expert, but I’m in a class right now!) suffers from severe 2nd system syndrome: it overhauls DHCP in a totally convoluted way, overhauls so much of how the IP layer works, etc. It really is a mess.
- 1ncorrect 5y ago
  I think it would have been better to label everything that DHCPv6 covers with an entirely new name. Its functionality is fine for what it’s tasked to do, and much of the confusion I’d argue stems from those who (think they) know IPv4 trying to map their understanding to IPv6 unmodified.
ospray 5y ago
For all of it's faults Port Address Translation (NAT) sure has a way of making the address size more manageable.
- hansel_der 5y ago
  we managed to connect a few billion devices using about 200mio addresses and a protocol that was designed for a few thousand hosts. successfull technology indeed.
jml7c5 5y ago
It's a silly bikeshed, but I wish they'd gone for hyphens as separators rather than colons. There's something about [abcd-ef12-3456-7890-abcd-ef12-3456-7890] that's nicer to read than [abcd:ef12:3456:7890:abcd:ef12:3456:7890]
- nightfly 5y ago
  It's probably not like that because then you could have something be both a valid hostname and a valid ip address
  - usr1106 5y ago
    Not if you always use the square brackets. Which is of course slightly awkward to type every single time.
  - jml7c5 5y ago
    Yeah, I realized that just after I posted my reply. As usr1106 mentioned, the solution would be to always use square brackets. It would take up more space, but I think it would still be better for at least two reasons: • Even with the brackets, it always takes the same or fewer keypresses (no need to press shift). • It's easier to tell when the address starts and ends. This is helpful when you get into odd addresses like "xxxx::xxxx.192.168.123.234", or "::1", or "::". The versions [xxxx--xxxx.192.168.123.234], [--1], and [--] look nicer and more obvious, at least to me. But of course this is all silly to go over now, decades after the decision. We'll save these changes for IPv8! :-P
- hansel_der 5y ago
  it's absolutely not bikeshedding. anecdote: a few years ago there was a quiz at one of the ripe-conferences where participants were asked to pick the one invalid ipv6 address out of four. iirc the success rate was even below what would result from a random-pick
RedShift1 5y ago
Many of the "IPv6-isms" come from the fact that renumbering networks is a big PITA and so IPv6 was built with the pipe dream of "change the config on the router and the entire network renumbers itself", like the flick of a switch. It was optimised for the least used case. IMO they should have just increased the address bit size to 128 and be done with it. One of my own annoyances is the fact that local subnets have to be /64's (you can make smaller subnets but it breaks certain things) and ISP's are supposed to deliver a /48 or /56 to subscribers but some ISP's will only hand one /64. So if you want to create 2 separate networks in your home you're screwed.
- benttoothpaste 5y ago
  It’s even worse - many home routers will simply disable ipv6 if the ISP gives /64.
usr1106 5y ago
So did 64 bit CPU adoption get delayed because addresses are hard to read compared to 32 bit addresses? Maybe not so much... But the argument might still hold. If 80% of average network admins decide to hate IPv6 because they can no longer remember IP addresses of several machines and subnets by heart that might hinder adoption. Most of them are not daily affected by the fact that the world has run out of IPv4 addresses. And NAT works well enough for most of them, many are even convinced it gives them security.
- soneil 5y ago
  Remembering your own machines really isn't the problem people think it is. If I ask you what 192.168.1.1 is, you're probably going to guess it's a router/gateway. There is absolutely no reason for it to be, it's merely an accepted convention that the lowest-numbered host is the gateway. There's absolutely nothing stopping you doing the same in ipv6. If you need to be able to remember where your gateway is, put it on 2001:db8::1. The problem isn't that they're long, it's that for some reason we assume we have to remember dynamically assigned addresses. Any address you have to memorise shouldn't be dynamically assigned, whether it's v4 or v6.
  - magicalhippo 5y ago
    > put it on 2001:db8::1 Except my prefix is a hell of a lot longer than that... And it's not my choice that it changes every now and then (at least once a month). That said, I use DynDNS for IPv4 too, but it's easier since I only need to configure it for the router, but for IPv6 each machine I want to access needs it's own record.

ashton314 5y ago

Address abbreviation helps a bunch. IPv6 (not an expert, but I’m in a class right now!) suffers from severe 2nd system syndrome: it overhauls DHCP in a totally convoluted way, overhauls so much of how the IP layer works, etc. It really is a mess.

1ncorrect 5y ago
I think it would have been better to label everything that DHCPv6 covers with an entirely new name. Its functionality is fine for what it’s tasked to do, and much of the confusion I’d argue stems from those who (think they) know IPv4 trying to map their understanding to IPv6 unmodified.

ospray 5y ago

For all of it's faults Port Address Translation (NAT) sure has a way of making the address size more manageable.

hansel_der 5y ago
we managed to connect a few billion devices using about 200mio addresses and a protocol that was designed for a few thousand hosts. successfull technology indeed.

jml7c5 5y ago

It's a silly bikeshed, but I wish they'd gone for hyphens as separators rather than colons. There's something about [abcd-ef12-3456-7890-abcd-ef12-3456-7890] that's nicer to read than [abcd:ef12:3456:7890:abcd:ef12:3456:7890]

nightfly 5y ago
It's probably not like that because then you could have something be both a valid hostname and a valid ip address
- usr1106 5y ago
  Not if you always use the square brackets. Which is of course slightly awkward to type every single time.
- jml7c5 5y ago
  Yeah, I realized that just after I posted my reply. As usr1106 mentioned, the solution would be to always use square brackets. It would take up more space, but I think it would still be better for at least two reasons: • Even with the brackets, it always takes the same or fewer keypresses (no need to press shift). • It's easier to tell when the address starts and ends. This is helpful when you get into odd addresses like "xxxx::xxxx.192.168.123.234", or "::1", or "::". The versions [xxxx--xxxx.192.168.123.234], [--1], and [--] look nicer and more obvious, at least to me. But of course this is all silly to go over now, decades after the decision. We'll save these changes for IPv8! :-P
hansel_der 5y ago
it's absolutely not bikeshedding. anecdote: a few years ago there was a quiz at one of the ripe-conferences where participants were asked to pick the one invalid ipv6 address out of four. iirc the success rate was even below what would result from a random-pick

RedShift1 5y ago

Many of the "IPv6-isms" come from the fact that renumbering networks is a big PITA and so IPv6 was built with the pipe dream of "change the config on the router and the entire network renumbers itself", like the flick of a switch. It was optimised for the least used case. IMO they should have just increased the address bit size to 128 and be done with it. One of my own annoyances is the fact that local subnets have to be /64's (you can make smaller subnets but it breaks certain things) and ISP's are supposed to deliver a /48 or /56 to subscribers but some ISP's will only hand one /64. So if you want to create 2 separate networks in your home you're screwed.

benttoothpaste 5y ago
It’s even worse - many home routers will simply disable ipv6 if the ISP gives /64.

usr1106 5y ago

So did 64 bit CPU adoption get delayed because addresses are hard to read compared to 32 bit addresses? Maybe not so much... But the argument might still hold. If 80% of average network admins decide to hate IPv6 because they can no longer remember IP addresses of several machines and subnets by heart that might hinder adoption. Most of them are not daily affected by the fact that the world has run out of IPv4 addresses. And NAT works well enough for most of them, many are even convinced it gives them security.

soneil 5y ago
Remembering your own machines really isn't the problem people think it is. If I ask you what 192.168.1.1 is, you're probably going to guess it's a router/gateway. There is absolutely no reason for it to be, it's merely an accepted convention that the lowest-numbered host is the gateway. There's absolutely nothing stopping you doing the same in ipv6. If you need to be able to remember where your gateway is, put it on 2001:db8::1. The problem isn't that they're long, it's that for some reason we assume we have to remember dynamically assigned addresses. Any address you have to memorise shouldn't be dynamically assigned, whether it's v4 or v6.
- magicalhippo 5y ago
  > put it on 2001:db8::1 Except my prefix is a hell of a lot longer than that... And it's not my choice that it changes every now and then (at least once a month). That said, I use DynDNS for IPv4 too, but it's easier since I only need to configure it for the router, but for IPv6 each machine I want to access needs it's own record.