insert inflammatory v6 thread

serveradministrator

I dont care about you all but I love IPv6. All my domains are dual stack. Even my home is dual stack and I only allow inbound IPv6 on my VPN to connect to my home. Even my SSH server on my domain only listens on IPv6, never had an attack on it.

Maounique

@TimboJones said: If IPv4, without changes, can't talk to the new IP address scheme, then it's no different than replacing with another version that can be rewritten from ground up instead of bloating into a bigger IPv4 mess. So all the drawbacks and no benefits.

Nope, the new technology must support the old, not the other way around, i.e. IPv6 should be written as such, the problem is that it is not.

@TimboJones said: You can buy a PCIe 3.0 x16 card and put it into a PCIe 2.0 slot, but it'll only ever work as good as a PCIe 2.0 slot.

IPv6 is not really a technically better standard. Yeah, it has some minor extra things but it is not faster, nor capable of transporting more data or anything. It is more or less a way to address the IPv4 numbering limitation.

jsg

@ahnlak said:

@jsg said:
Try again when you actually have arguments beyond "No! Because ahnlak says so!" and "you are evil!".

If that's all you've got I've clearly touched a nerve, so I'll just assume I won and we can leave it at that, ok?

Nice - but stupid - try.

Here you go, 128 bit version:

int main(int argc, char **argv)
{
    volatile unsigned __int128 a, last_a;

    for(int i =0; i < 2000000000; i++) {
        last_a = (unsigned __int128) (123456789 * i);
        a = last_a ^ (unsigned __int128) 0x55aa55aa55aa55aa;
    }
    printf("%X\n", a & 0xFFFFFFFF);
    return 0;
}

and here the exact same but 64 bit:

int main(int argc, char **argv)
{
    volatile uint64_t a, last_a;

    for(int i =0; i < 2000000000; i++) {
        last_a = (uint64_t) (123456789 * i);
        a = last_a ^ (uint64_t) 0x55aa55aa55aa55aa;
    }
    printf("%X\n", a & 0xFFFFFFFF);
    return 0;
}

Run time 64 bit version: 968 ms
Run time 128 bit version: 1919 ms

Oopsie, about double the runtime. So much for "64 bit vs 128 bit makes no difference". And I was very generous and used no array (as routers need to).

Second (out of more) point:

So you think that 1000 IPs per person might not be enough and the world population is about to grow by a factor of 2 million times 4 billion times 4 billion. Accordingly you seem to think that getting a say /64 IPv6 allotment is good and desirable.

Quick question: How are the people to route to/from their couple of million or even couple of billion devices? With the cheap plastic box they get for $50 or $100? Good luck with that ...

In other words: the wise and forward looking IPv6 proponents who tell us how great it is to get billions (or even trillions or quadrillions) of IP addresses for personal use ... but they conveniently ignore that all those addresses must be routed (and switched and ...) too.
"Well, they do not really need trillions of private IPv6 addresses" you say? Well, if they can live with say 1 mio private addresses a 128 bit address space simply isn't needed.

(As he seems to not have got it - like so many other things: I do not read what "Mr. F_ck" writes) - just mentioning it for politeness sake

SirFoxy

I really just made this thread on a conversation tip because I know it'd get a discussion started, but:

Love you:
@yoursunny
@Harambe
@Francisco
@Maounique
@raindog308
@TimboJones
@jsg
@ezeth
etc

The conversation is beautiful, even from different perspectives.

adef_1

Ipv6 has been out for a long time, if it is really useful, it has be widely used.

SirFoxy

@adef_1 said:
Ipv6 has been out for a long time, if it is really useful, it has be widely used.

very cool, thank u.

Maounique

@adef_1 said: Ipv6 has been out for a long time, if it is really useful, it has be widely used.

It is helpful but not yet a necessary condition to get online. As long as the old way works with minimal overhead cost, it will not change, I am curious how much is the cost people are able to tolerate before thinking of change.

dahartigan

IPv6 is just another scam created by some nerd's wife to fuck with him, but it ended up becoming a virus in itself.

Now nerds all over the world are scrambling to adapt this imaginary IPv6 to avoid an imaginary IPv4 shortage.

The last time a man's wife told a lie to fuck with him, almost 2022 years ago, look what happened.

It's all scams by females to oppress male nerds further. Jack Ma disappeared not by the hand of a man..

Am I serious? Am I trolling? Depends how you observe it..

lebuser

@jsg said:

Quick question: How are the people to route to/from their couple of million or even couple of billion devices? With the cheap plastic box they get for $50 or $100? Good luck with that ...

Actually using IPv6 instead of IPv4 NAT may increase your routing throughput on such devices. I don't think that's unexpected. NAT comes with performance penalities, and with IPv4 you need to calculate header checksums at each hop (also when not uaing NAT).

See for example the performance comparison made with D-Link DIR-857.

https://foxnetlab.com/index.php/16-reviews/85-d-link-dir-857-or-hd-media-router-3000

mcgree

@serveradministrator said:
I dont care about you all but I love IPv6. All my domains are dual stack. Even my home is dual stack and I only allow inbound IPv6 on my VPN to connect to my home. Even my SSH server on my domain only listens on IPv6, never had an attack on it.

Scanning IPv6 seems too difficult,~

If there is no special detection disadvantage, the time required to scan /96 is the total time to scan the whole IPv4 Internet, it is said that masscan can scan the whole IPv4 exchange network (1 port) in 3 minutes, so it takes 3 minutes to scan /96, how long is it to scan /64? What about the entire IPv6 Internet(2000::/3)? What about all ports?

jerry_me

Personally, I think NAT can alleviate IPv4 problems and last forever, and there is nothing that requires IPv6, so it is hard to popularize it.

Are there any large websites that are IPv6 Only?

mcgree

@yoursunny said:

@ezeth said:
We still got Class E with 300M IPs literally "reserved for future use"

yoursunny summer host Inc has a secret deal with Starfleet Headquarters that we will start providing experimental IPv4 service out of Class E addresses someday.
All of Class E ranges are mine.

You just need to make 999.999.999.999 available to have a lot of IPv4 Plus addresses.

Then let me occupy 256.256.256.256.

mcgree

@Maounique said:

@TimboJones said: If IPv4, without changes, can't talk to the new IP address scheme, then it's no different than replacing with another version that can be rewritten from ground up instead of bloating into a bigger IPv4 mess. So all the drawbacks and no benefits.

Nope, the new technology must support the old, not the other way around, i.e. IPv6 should be written as such, the problem is that it is not.

@TimboJones said: You can buy a PCIe 3.0 x16 card and put it into a PCIe 2.0 slot, but it'll only ever work as good as a PCIe 2.0 slot.

IPv6 is not really a technically better standard. Yeah, it has some minor extra things but it is not faster, nor capable of transporting more data or anything. It is more or less a way to address the IPv4 numbering limitation.

Fixing a broken thing is much more difficult than developing a new thing, but I don't know if IPv4 is considered broken, for example, IPv6 removes ARP (using ICMP instead), because ARP is broken?

mcgree

Many ultra-conservative users allocate /128 IPs to their clients, but with a minimum BGP size of /48, can their routers really accommodate 2^80 devices? Do they really have that many users/devices/endpoint?

Maounique

@mcgree said: Many ultra-conservative users allocate /128 IPs

Another place where ultra-conservative means braindead.

@mcgree said: You just need to make 999.999.999.999 available to have a lot of IPv4 Plus addresses.

That is not really possible because it has to have a mask, and masks are binary, not decimal, but 255.255.255.255.255.255.255.0 masks are feasible and will give us enough IPv4+.

@mcgree said: Fixing a broken thing is much more difficult than developing a new thing,

This is not broken, IPv4 was not broken (well, it is, but IPv6 doesnt fix the inherent flaws with fragmentation, flooding, UDP, many flaws which are exploited for attacks and which plague routing) and adding new features as well as expanding the numbering is not really a fix, just an actualization/expansion. It would also be possible to be backwards compatible. But it is too late now, we are stuck with IPv6 and we have to live with it or pay for the right to procrastinate forever.

rcy026

@jsg said:

Here you go, 128 bit version:

int main(int argc, char **argv)
{
volatile unsigned __int128 a, last_a;

for(int i =0; i < 2000000000; i++) {
last_a = (unsigned __int128) (123456789 * i);
a = last_a ^ (unsigned __int128) 0x55aa55aa55aa55aa;
}
printf("%X\n", a & 0xFFFFFFFF);
return 0;
}

and here the exact same but 64 bit:

int main(int argc, char **argv)
{
  volatile uint64_t a, last_a;
  
  for(int i =0; i < 2000000000; i++) {
      last_a = (uint64_t) (123456789 * i);
      a = last_a ^ (uint64_t) 0x55aa55aa55aa55aa;
  }
  printf("%X\n", a & 0xFFFFFFFF);
  return 0;
}

Run time 64 bit version: 968 ms
Run time 128 bit version: 1919 ms

Oopsie, about double the runtime. So much for "64 bit vs 128 bit makes no difference". And I was very generous and used no array (as routers need to).

I am no programmer so I will not and can not argue with above, however, I am a network engineer and all I see is that the routers that run pure IPv6 has much better performance than the ones that do IP4, even if the hardware spec is the same. And that happens all the time, everywhere. We stopped doing dualstack and went with separate routers for IPv6 and IP4 due to the awesome performance we get from the pure IPv6 networks.
I'm sure that in theory, jsg has a point somewhere. He usually does. Unfortunately, reality often disagrees with jsg.

jmgcaguicla

@rcy026 said:
and all I see is that the routers that run pure IPv6 has much better performance than the ones that do IP4, even if the hardware spec is the same.

You're gonna have to back those claims with some numbers son, we're not running a cult here.

jsg

@lebuser said:

@jsg said:

Quick question: How are the people to route to/from their couple of million or even couple of billion devices? With the cheap plastic box they get for $50 or $100? Good luck with that ...

Actually using IPv6 instead of IPv4 NAT may increase your routing throughput on such devices. I don't think that's unexpected. NAT comes with performance penalities, and with IPv4 you need to calculate header checksums at each hop (also when not uaing NAT).

See for example the performance comparison made with D-Link DIR-857.

https://foxnetlab.com/index.php/16-reviews/85-d-link-dir-857-or-hd-media-router-3000

You mean to say it may (slightly) increase throughput on some plastic boxes in a favourable context that btw is nothing to do with my argument, which was about having to route billions of private IPv6.

@rcy026 said:

@jsg said:

[some code]

Run time 64 bit version: 968 ms
Run time 128 bit version: 1919 ms

Oopsie, about double the runtime. So much for "64 bit vs 128 bit makes no difference". And I was very generous and used no array (as routers need to).

I am no programmer so I will not and can not argue with above, however, I am a network engineer and all I see is that the routers that run pure IPv6 has much better performance than the ones that do IP4, even if the hardware spec is the same. And that happens all the time, everywhere. We stopped doing dualstack and went with separate routers for IPv6 and IP4 due to the awesome performance we get from the pure IPv6 networks.

Even if that were true, and frankly I doubt it, so what? The whole point was "we need a larger address space!" and not "let's make routers faster!".

And again, current world population is about 2^33 and we have every reason to assume that this planet can not sustain even a population of 2^34. So reasonably assuming a max population of 2^34, or no, let's go insane and assume a world population of 2^36 - that is 8 times today's population - and let's give 4000 IPs average to each and every one ... and we're at 2^48.
IPx with an address space of 64 bits would be 65000 times that; a very generous reserve, space to grow, ... that would allow us to colonize thousands of planets and still have plenty free address space (not that we'll actually discover even just 100 planets within reach that support human life but hey, that's the argument of some IPv6 zealots, so let's include that imaginary "need").

The question - to which I still didn't get a realistic and sensible answer - is why the IPv6 zealots say we need a 128 bit address space.

Maounique

@jsg said: why the IPv6 zealots say we need a 128 bit address space.

I am an IPv6 zealot and do not say we need an 128 bit space.
I only say that we run out of IPv4 and, while atm the crisis is a bit artificial for profiteering, there will come a time the IPv6 will have to be everywhere.
It is a bad solution, I was underwhelmed from the get-go, but it is too late now, we will roll it out volens-nolens. The faster, the better.

jtk

@jsg said:
The question - to which I still didn't get a realistic and sensible answer - is why the IPv6 zealots say we need a 128 bit address space.

One simple answer is stated at the beginning of the standards doc if anyone cares to take a look. From the latest version of IETF RFC 8200:

IPv6 increases the IP address size from 32 bits to 128 bits, to support more levels of addressing hierarchy, a much greater number of addressable nodes, and simpler autoconfiguration of addresses.

Many were not satisfied with how things turned out and wonder what might have happened if alternative proposals were selected. Nonetheless, this is what we have to replace IPv4 with for better or worse. It is not going away and it is highly unlikely we will see another IPng.

Stryp

@jsg said:
I think @Francisco nailed it. And I'd add my own major criticism, the insane address size.
Which is much to do with people too btw, because people tend to dislike or even reject anything with (significantly) more than 7 of anything. IP4 is 4 address elements and I'd even go as far as strongly suggesting "IP5a" (64-bit) to be written with 4 elements too rather than 8, like so "abcd.1234.5679.abcd".
People need the feeling they can remember their IP(s).

IPv6 feels impractical, is impractical, and is totally oversized. Something very much IP4-like but with 64-bit addresses is what we need - and I've yet to hear a sensible and realistic reason why we allegedly need more IP addresses than molecules in our solar system.

IPv6 is like telling someone whose 4 cores pre-Ryzen PC feels a bit slow that he absolutely needs IBM's largest mainframe plus a 1 Tb fiber to his house.

Accordingly, the result of 2 decades of "explaining", marketing, trumpeting, herding, pushing IPv6 is ... not even 10% real acceptance and use.

This is the heart of the problem for me. I remember my home IP address, and two addresses for my VPS servers. With ipv6, it just isn't feasible for me to remember 2a00:1110:206:dhd4:bc09:c6n7:116b:84g3, which similar to my real home IPv6 address.

If they designed an IP address format that had [0-9a-f] at still the same max length (4 blocks of 3 characters) which I will call IPvX (for example: 1a5.75d.bb1.12f)
IPvX: 281,474,976,710,656
IPv4: 000,004,294,967,296

65 thousand times as many IP addresses.
35,700 IP addresses for every living human.
From 0.545 IP addresses for every living human.

End users wouldn't see that much change, only that their IPs can now contains letters between A-F. Cool AF.

So while IPv6 is likely the future, I would still like to keep using my IPv4 addresses for my servers and such, it is just easier.

stratagem

Is that not what DNS is for?

jsg

@Maounique said:

@jsg said: why the IPv6 zealots say we need a 128 bit address space.

I am an IPv6 zealot and do not say we need an 128 bit space.
I only say that we run out of IPv4 and, while atm the crisis is a bit artificial for profiteering, there will come a time the IPv6 will have to be everywhere.
It is a bad solution, I was underwhelmed from the get-go, but it is too late now, we will roll it out volens-nolens. The faster, the better.

No, IMO there is still enough "bubbles" in IP4 we can cut open, provided we are willing to, to buy us the time to design a reasonably sized and well engineered IP4 successor.
Btw, you seem to be the kind of IPv6 zealot I can take seriously.

@jtk said:
One simple answer is stated at the beginning of the standards doc if anyone cares to take a look. From the latest version of IETF RFC 8200:

IPv6 increases the IP address size from 32 bits to 128 bits, to support more levels of addressing hierarchy, a much greater number of addressable nodes, and simpler autoconfiguration of addresses.

That is plain BS as is to be expected from a mental asylum.

@Stryp said:
This is the heart of the problem for me. I remember my home IP address, and two addresses for my VPS servers. With ipv6, it just isn't feasible for me to remember 2a00:1110:206:dhd4:bc09:c6n7:116b:84g3, which similar to my real home IPv6 address.

If they designed an IP address format that had [0-9a-f] at still the same max length (4 blocks of 3 characters) which I will call IPvX (for example: 1a5.75d.bb1.12f)
IPvX: 281,474,976,710,656
IPv4: 000,004,294,967,296

65 thousand times as many IP addresses.
35,700 IP addresses for every living human.
From 0.545 IP addresses for every living human.

End users wouldn't see that much change, only that their IPs can now contains letters between A-F. Cool AF.

So while IPv6 is likely the future, I would still like to keep using my IPv4 addresses for my servers and such, it is just easier.

Yes. But I have bad news: Humans, we, the billions of users, are utterly irrelevant to them. What's relevant (to the deranged IPv6 "engineers") is, e.g. "more levels of addressing hierarchy, a much greater number of addressable nodes, and simpler autoconfiguration of addresses."

Sorry.

ahnlak

@jsg said:

I am no programmer so I will not and can not argue with above, however, I am a network engineer and all I see is that the routers that run pure IPv6 has much better performance than the ones that do IP4, even if the hardware spec is the same. And that happens all the time, everywhere. We stopped doing dualstack and went with separate routers for IPv6 and IP4 due to the awesome performance we get from the pure IPv6 networks.

Even if that were true, and frankly I doubt it, so what? The whole point was "we need a larger address space!" and not "let's make routers faster!".

Err, nice try but one of your central arguments here has been how much more processing power and energy all those extra bits will absorb.

When someone actually comes along with real world experience to counter your synthetic and largely meaningless "benchmark", you just suddenly decide that performance is no longer an issue?

Shot2

@Stryp said:

If they designed an IP address format that had [0-9a-f] at still the same max length (4 blocks of 3 characters) which I will call IPvX (for example: 1a5.75d.bb1.12f)
End users wouldn't see that much change, only that their IPs can now contains letters between A-F. Cool AF.

Actually, that could be more or less the case (...if/when done properly).

To try and build upon your analogy, the first 4 nibbles (2001:db8:babe:cafe: ) represent your home address - as impossible to remember as "135 Downing Road, Building E Apartment 2, Lundun 3C56E, United Kingdom of America" (that's why phonebooks and notepads exist). Streetnames postcodes etc. are not up to you, it was decided by urban planners and city officials.

Anything that comes after that (::1, or 3d58:4af7:b5:952a) refers to various places within your home - as decided by you/your dad/your architect/interior designer (e.g. "the bathroom" or "the red shoebox in the wooden cabinet in Emily's bedroom"). Theoretically, it does not have to be memorable (e.g. temporary addresses) or complicated (just use ::1, ::2 etc. for each machine or device)

jsg

@ahnlak said:

@jsg said:

I am no programmer so I will not and can not argue with above, however, I am a network engineer and all I see is that the routers that run pure IPv6 has much better performance than the ones that do IP4, even if the hardware spec is the same. And that happens all the time, everywhere. We stopped doing dualstack and went with separate routers for IPv6 and IP4 due to the awesome performance we get from the pure IPv6 networks.

Even if that were true, and frankly I doubt it, so what? The whole point was "we need a larger address space!" and not "let's make routers faster!".

Err, nice try but one of your central arguments here has been how much more processing power and energy all those extra bits will absorb.

When someone actually comes along with real world experience to counter your synthetic and largely meaningless "benchmark", you just suddenly decide that performance is no longer an issue?

I proved your "128 bit vs 64 bit makes no difference" wrong, plain and simple. And that doesn't change by some anecdotal "evidence". Sad enough btw that you needed a proof for something that everyone in IT worth his salt knows.
Similarly your getting increasingly personal doesn't change the facts.

Maounique

Okay, let's talk about performance.
Yes, the routing tables WILL be larger as space taken, the computing more complex, but we will not do that for EVERY IPv6 address, we do this for at most a /64 chunk at a time so there will be way less lookup than now. The idiots using /80 /112 /128 will do the "distribution" in house, kind of their NAT ("distributing" a unique /64 to multiple interfaces), if you like, will not affect global routing, so we actually have a 64 bit system, more or less. Of course it is not that simple, but in some aspects, it is.
So, we will trade the computing for the highly fragmented routing tables atm (people want to break it further to /25 /26...) to a higher bit count but much less fragmentation. Who gets a /48 will likely not need another for most regular usage scenarios.
A /64 for end users, maybe /56 for a crazy IT enthusiast like me. A /24 for the provider should be enough for said provider forever, even in China.
Overall, I don't think it will be more computing expensive and we will completely eliminate NAT.

rcy026

@jsg said:

The question - to which I still didn't get a realistic and sensible answer - is why the IPv6 zealots say we need a 128 bit address space.

I have never heard anyone say that we need a 128-bit address space.
But then again, it does not hurt, so why not? Your claim that it would hurt performance does not seem to reflect in real world figures.

@jsg said:
Similarly your getting increasingly personal doesn't change the facts.

Said the man that constantly and repeatedly refers to anyone not agreeing with him as zealots, mental asylum, deranged and similar.

I've said this before and I say it again. If you show me a better solution than IPv6 I will gladly try it out. As of now, I use networking equipment from all major vendors you could possibly name, and they all support IPv6. And it works extremely well, I have never run into any of the problems you try to point out despite running IPv6 in huge production networks for over a decade. This makes it extremely hard to take your objections seriously. Theoretically I might agree to some extent with some of your points, but in reality, they do not exist.
Show me an alternative, any kind of alternative, and I will try it out. It could even be unsupported by vendors and involve running custom software on routers, I would still test it out and give it an honest try. If you can not, I will conclude that IPv6 is the only viable alternative that exists and continue to use and advocate it.

This whole discussion is like the one of whether or not a bumblebee can fly. There are scientists that claim that bumblebees can not possibly fly and they do have some arguments to back it up. But still, bumblebees do fly, and it should be pretty obvious to anyone with half a brain that it is not the bumblebee that is wrong.

rcy026

@jsg said:
You mean to say it may (slightly) increase throughput on some plastic boxes in a favourable context that btw is nothing to do with my argument, which was about having to route billions of private IPv6.

That is not how routing works, especially not with IPv6.
Most, if not all, IPv6 routers actually have smaller routingtables then their IP4 counterparts due to more efficient and hierarchical routing in IPv6.

ahnlak

@jsg said:
I proved your "128 bit vs 64 bit makes no difference" wrong, plain and simple. And that doesn't change by some anecdotal "evidence". Sad enough btw that you needed a proof for something that everyone in IT worth his salt knows.

I'll take real world experience over a synthetic benchmark that has little to do with the problem space any day - as would anyone "in IT worth his salt".

FWIW I never said it "makes no difference" - I said modern hardware wouldn't have the slightest problem with it, and questioned your repeated assertion (that you've never lowered yourself to respond to) that it will be slower and "waste lots of energy".

I'll tell you what else wastes a lot of energy - this thread.

@jsg said:
Similarly your getting increasingly personal doesn't change the facts.

Sure. I'm the one getting personal