News for hackers. Stuff that matters.
Re:IPv4 pricing by rtutz on Wednesday July 28, 2021 @08:59AM (#27983170)
This whole problem could have been avoided if IPv6 would be easier to memorize. I feel like especially when setting up networks, the v6 part is not as natural as v4. It is simply additional overhead and causes a lot of "scratching my head" moments. Otherwise there would be no reason to not leave v4 behind and just move on.
Re:IPv4 pricing by oarsinsync on Wednesday July 28, 2021 @09:03AM (#27983199)
> This whole problem could have been avoided if IPv6 would be easier to memorize.

Thankfully, we have DNS. A lot of ISP issued consumer CPEs now automatically create lan-local DNS entries for clients based on hostname provided by the client at dhcp time, a lot of clients also natively support mDNS, and there are plentiful free DNS providers if none of the above applies to you, and you can't host your own.

Remembering IPs isn't something that people should need to do at this point in our networks maturity.

Re:IPv4 pricing by lowercased on Wednesday July 28, 2021 @09:44AM (#27983642)
Agreed. If, in 1997/98, the ipv6 spec had been "prefix 2 more 8 bit values at the beginning" - and all existing addresses moved in to 0.0.a.b.c.d - we could have had a much easier path for migration (imo). And yes, it wouldn't have been "128 bit!" but we still would have had 255 more address spaces of 4 billion each, which would have bought us some more time. I think we'd have been further along that migration path than where we are now, after 23 years.
Re:IPv4 pricing by mprovost on Wednesday July 28, 2021 @01:07PM (#27986171)
I mean we've managed to stretch v4 for 20 years longer than anyone thought possible. Adding one more bit to the address would have doubled the size of the v4 space, so another 8 bits would have been plenty.
Re:IPv4 pricing by lowercased on Wednesday July 28, 2021 @01:13PM (#27986250)
Yep. But... "now every star in our galaxy can have their own /16 block!". That's a paraphrased recollection I have from some networking colleague in '98 when this all was coming down. It seemed a strange goal, and I'm presuming he was just trying to illustrate how 'vast' IPv6 was.
Re:IPv4 pricing by yesco on Wednesday July 28, 2021 @09:44AM (#27983653)
IPv6 addresses theoretically should be easier to memorize & work with than IPv4 thanks to the double colon shorthand acting as a wildcard for zeros and due to it being hex grouped rather than octet grouped.

As an example 2001:0db8:0000:0000:0000:0000:0370:7334 could be written as 2001:db8::370:7334 instead (notice that leading zeros were also culled). This paired with the fact that hexadecimal tends to be easier to memorize and doesn't have the strange subnet masking logic like IPv4, gives it a lot of advantages over IPv4's address notation.

The problem is that it's almost like router firmware and ISPs go out of their way to make their addresses harder to work with by filling out all 8 hex groups in the addresses they grant. Considering the sheer amount of available IPv6 addresses, it's from my understanding, completely unnecessary and I'm really curious if they have any kind of justification or technical reasoning for doing this.

Re:IPv4 pricing by est31 on Wednesday July 28, 2021 @10:15AM (#27984011)
Two explanations come to mind:

1. easier routing tables if you can add meanings to specific bit ranges of your ipv6 address. In the tightly assigned ipv4 networks we have arrived at this is a bit annoying.

2. If the ipv6 conventions were that you set, say the highest 5 hex groups to 0, and use the lowest 3 hex groups for addresses, it would still be 65536 times as large as the ipv4 space and would suit most needs for the mid term future. You could even write ipv6 addresses nicely using e.g. ::ef13:2.1.7.100. This is a valid ipv6 notation! If this space ever got too tight one could open another one of the available hex groups and use two hex group prefixes. But I think when this happens, a lot of configurations would break because they'd assume that only 48 bits are used of the total 128. To prevent router,switch,firewall, etc. vendors from putting any such assumptions into their devices, using the full 128 bits from the start is a good option.

Re:IPv4 pricing by cortesoft on Wednesday July 28, 2021 @11:08AM (#27984707)
Even your shortened version is a lot harder to remember and type than an IPv4 address.
Re:IPv4 pricing by Symbiote on Wednesday July 28, 2021 @11:55AM (#27985296)
My aunt's phone number in Germany was 14 digits as dialled, compared to her brother's 6 digits; he lived in the same town as us.

Giving everyone, worldwide an internet address means they have to be longer than limiting it to the early adopters.

Re:IPv4 pricing by cortesoft on Wednesday July 28, 2021 @01:17PM (#27986310)
I understand that, but right now people are able to get by with IPv4 only, and aren’t going to switch until they have to. The long term reality isn’t going to make someone voluntarily switch.
Re:IPv4 pricing by yesco on Wednesday July 28, 2021 @11:46AM (#27985181)
That part is a bit more subjective I suppose. For me at least, I find hex far easier to remember than strings of numbers.
Re:IPv4 pricing by the8472 on Wednesday July 28, 2021 @10:03AM (#27983858)
Randomizing the prefix makes network scans more costly.
Re:IPv4 pricing by p1mrx on Wednesday July 28, 2021 @10:45AM (#27984402)
It's impossible to make an addressing scheme that's both memorizable, and abundant enough for the foreseeable future of the Internet. The human brain just isn't capable of dealing with numbers on that scale, which is why we invented computers in the first place.
Re:IPv4 pricing by elric on Wednesday July 28, 2021 @10:56AM (#27984567)
> It's impossible to make an addressing scheme that's both memorizable, and abundant

Not really. In fact, pretty much anything would have been easier to memorize than this colon-separated nonsense, which makes URL parsing more difficult, and which is so stupidly complex that it has a special syntax to ignore repeating zeros.

Re:IPv4 pricing by p1mrx on Wednesday July 28, 2021 @11:16AM (#27984804)
An IP address is fundamentally a 32-bit or 128-bit binary number, and hexadecimal is the most human-friendly base to represent those. Decimal gets pretty hairy once you introduce CIDR prefixes that aren't 8-bit aligned.

The [IPv6]:port syntax is unfortunate, but I'm not sure what they'd have done instead. Dotted hexadecimal would be ambiguous, because "1.2.3.4.5.6.beef.de" looks like a DNS hostname.

Zero compression exists because it's more convenient than writing all those zeroes, especially with CIDR prefixes like "2000::/3".

Re:IPv4 pricing by laurensr on Saturday July 31, 2021 @07:04AM (#28017040)
I think browser are advanced enough to parse [IPv6]:port. Note that they can even distinguish octal and decimal IPs, for example this is working in Chrome, Firefox and the Windows ping utility:

http://0100.0351.0251.0152/

As well as decimal:

http://1089055082/

Both lead to google.com

Re:IPv4 pricing by p1mrx on Saturday July 31, 2021 @10:11AM (#28018274)
> I think browser are advanced enough to parse [IPv6]:port.

Except for recent versions of Firefox Mobile: https://github.com/mozilla-mobile/fenix/issues/4343

Re:IPv4 pricing by ikiris on Wednesday July 28, 2021 @11:49AM (#27985220)
you're free to use the entire 128 bit number, or the older dotted decimal notation.
Re:IPv4 pricing by the8472 on Wednesday July 28, 2021 @09:34AM (#27983535)
You can roll for an ULA prefix once, note it down in some text files and then assign your pets to <prefix>::1, <prefix>::2, <prefix>::3, etc.

mDNS might also help, I haven't tried that approach.

Re:IPv4 pricing by pas on Wednesday July 28, 2021 @09:04AM (#27983207)
Um, who memorizes cloud IP addresses?
Re:IPv4 pricing by orev on Wednesday July 28, 2021 @09:33AM (#27983518)
Contrary to popular belief, the Cloud has not actually eaten all of IT.
Re:IPv4 pricing by pas on Wednesday July 28, 2021 @10:02AM (#27983841)
I mean, when you start a new VM on Hetzner (or AWS/GCP/Azure/DO/whatever) you don't memorize that address.

But cloud or not, if you setup a private network with v6 you can get a nice /48 prefix, and you give out /64 prefixes to VMs, so you'll have 48 unchanging bits to memorize (or put it into a .txt to have it near). And most of that will probably be zero anyway.

For example 2a00:1450:4001 is a /48, and 2a00:1450:4001:082b /64. Only change is "082b".

I know, it's not the same as just remembering 1.1.1.1, but most of the people working with v4 never had so simple addresses to work with. (And if we're talking about 10.0.0.0/8 and other private addresses, well, folks can continue to use them, if they want to endlessly debug NAT and static routing hacks.)

Re:IPv4 pricing by rtutz on Wednesday July 28, 2021 @09:14AM (#27983309)
Not necessarily remembering cloud adresses, but it is fairly easy to design v4 networks. Subnet masks for example are short and understood with a brief glance at them. If v6 would be simpler, it would also be the first choice for more local networks, hence more widespread.
Re:IPv4 pricing by detaro on Wednesday July 28, 2021 @10:03AM (#27983850)
How are IPv6 subnet masks more complicated?
Re:IPv4 pricing by est31 on Wednesday July 28, 2021 @09:22AM (#27983381)
I use addresses I memorize to debug broken networks, to check whether it's a DNS or a general network issue.
Re:IPv4 pricing by taf2 on Wednesday July 28, 2021 @09:09AM (#27983260)
Me - I have far too many pets
Re:IPv4 pricing by pantalaimon on Wednesday July 28, 2021 @10:05AM (#27983889)
2600:: is a neat one
Re:IPv4 pricing by pas on Wednesday July 28, 2021 @10:10AM (#27983949)
Oh, and it pings and even serves HTTP too. Pretty neat indeed!
Re:IPv4 pricing by sswaner on Wednesday July 28, 2021 @09:20AM (#27983361)
Just 8.8.8.8
Re:IPv4 pricing by DrBenCarson on Wednesday July 28, 2021 @09:23AM (#27983394)
1.1.1.1 and 1.0.0.1 for me :)
Re:IPv4 pricing by throaway46546 on Wednesday July 28, 2021 @10:03AM (#27983851)
1.1.1.1 and 1.1
Re:IPv4 pricing by mnd999 on Wednesday July 28, 2021 @10:43AM (#27984384)
Not this one again, at this point it’s an “I don’t like it, it’s different!” whine.
Re:IPv4 pricing by knuthsat on Wednesday July 28, 2021 @09:00AM (#27983183)
Any reason why having server infrastructure in only IPv6 is an issue?
Re:IPv4 pricing by gzer0 on Wednesday July 28, 2021 @09:06AM (#27983229)
It would be a massive problem. IPV6 adoption and implementation was at a mere 33% (at least among Google users). [1]

[1] https://www.google.com/intl/en/ipv6/statistics.html

Re:IPv4 pricing by yjftsjthsd-h on Wednesday July 28, 2021 @09:14AM (#27983308)
If you're behind a CDN, your origin can be pure IPv6
Re:IPv4 pricing by est31 on Wednesday July 28, 2021 @09:20AM (#27983362)
Then it doesn't matter whether you use ipv6 or ipv4 with the private 10.0.0.0/8 space either.
Re:IPv4 pricing by yjftsjthsd-h on Wednesday July 28, 2021 @12:40PM (#27985832)
Yes! If you're behind ex. cloudflare, you should 100% look at running pure IPv6 with no listening ports, just their service locally.
Re:IPv4 pricing by selfhoster11 on Wednesday July 28, 2021 @09:21AM (#27983370)
Plenty. If you expect to access it from IPv4-only networks, you'll have to provide a gateway. Additionally, things like Docker interoperate very poorly with IPv6.
Re:IPv4 pricing by Hamuko on Wednesday July 28, 2021 @09:06AM (#27983230)
You want an IPv4 address if you want to be reachable by people.
Re:IPv4 pricing by api on Wednesday July 28, 2021 @09:30AM (#27983483)
I have been saying this for years. Nobody gets it because nerds don’t get the critical importance of ergonomics and usability.

If we had added 16 bits to v4 we would have 100% adoption by now.

Re:IPv4 pricing by kaliszad on Wednesday July 28, 2021 @10:14AM (#27983991)
The issue would be more or less the same. You'd have to buy new hardware and check all software anyway but would drop many of the benefits of the IPv6 we have. E.g. in enterprise networks, it is very nice you don't have to think about the size of a subnet for a VLAN anymore, you just give every VLAN /64 and it will suffice. The extra address space is also nice for autoconfiguration and much more we don't even think about yet. I think, IPv6 is ok as it is. A practical protocol is never perfect and will not please everybody but IPv6 stood the test of time, there is considerable traffic over IPv6 and we are slowly, but surely getting there.
Re:IPv4 pricing by api on Wednesday July 28, 2021 @11:10AM (#27984730)
Adding IPv6 support has never really been the issue. It's in every single piece of hardware or software I have. The problem is that people don't want to use it, as evidenced by the fact that people avoid it on overlay or virtual networks and use IPv4 if possible.

The very slight convenience you mention is far outweighed by 32+ digit IP addresses.

Also please don't bring up DNS. Anyone arguing that DNS is a solution to this problem has never done devops or IT.

Re:IPv4 pricing by kaliszad on Wednesday July 28, 2021 @02:59PM (#27988122)
Actually, IPv6 addresses cannot be longer than 32 digits. Some practical ones can be rather short, usually just slightly longer than a comparable IPv4 address. Such addresses would be used where remembering/ recognizing the exact IPv4 or IPv6 is relevant, such as the DNS servers or the network hand-off IP/ floating-IP on a firewall cluster or something like that that are used for the bring-up of other services. I have done my fare share of devops/ IT/ administration and engineering of largish enterprise and campus networks.

You would be surprised how much hardware and software doesn't support IPv6 properly. Sometimes it is the basic things, sometimes the more advanced stuff but that just means it takes a second or multiple days to find out. The problem is, it just is a similar but different protocol so you have to be quite diligent and check everything you need for the device/ service to work.

People do all kinds of stuff on underlay and overlay networks. E.g. some Dell VxRail hyper-converged appliances use IPv6 for the management network https://i.dell.com/sites/csdocuments/Shared-Content_data-She.... This is basically just link-local addresses for L2 reachability if I remember correctly but they could've gone with IPv4 there as well. It certainly would be more common for enterprise appliances to not rely on IPv6 for anything even when it shouldn't make a difference whether you do.

Re:IPv4 pricing by pas on Wednesday July 28, 2021 @10:04AM (#27983881)
Well, accidentally we added 96 instead of 16, oops.