I recently heard a talk about the demise of the internet as a result of the exhaustion of ip addresses with ipv4.
I always figured, ‘aahhhh, what a load of crap, just NAT the shit out of everything.’ But the speaker pointed out you run into the problem of port exhaustion on the internet-facing machine. Okay, point taken, I concede, the NAT forever thing won’t work. Although it certainly could last a long long time if they bothered to organize a little better, but I’ll let that go, we really are running out of addresses.
Still, the sky is far from falling, I have one really simple thought that made all of ipv6 really pointless and a terribly complicated exercise in wasting everybody’s time.
ipv4 has its share of problems, but the biggest one is that we’re running out of addresses, or rather in February, the IANA actually handed out the last batch. That’s it no more.
IPV6 was designed starting 15 years ago or so, and nobody lifted a finger to fix something that wasn’t broken. But in all that time, like c++ and everything else, they had grand plans, and they added features. IPV6 was going to streamline all sorts of byte wasting excessive packet size, it was going to enable ipsec at the ip layer (or something like that I forget the details) and they were going to add this useful feature, and that useful feature and so on and so forth for all 15 years that everybody was ignoring them and not implementing it.
But fast forward to now, and it turns out the only problem we ACTUALLY have to solve is that we’re running out of addresses.
ipv6 offers a 128 bit source and destination address, and the current rollout of ipv6 as it is being adopted is pretty much doing absolutely nothing other than solving the problem of running out of addresses. All that ipsec and all that other grand vision feature stuff is all gone. People are implementing ipv6 because they need more addresses and that’s it.
ipv6 was supposed to be many things to many people, but as it turned out, we only really needed the bigger address space.
Well if you look at the ipv4 header there’s got to be 3-5 bytes of shit that nobody ever uses for anything (like the fragment stuff), that just go to waste and could have been repurposed for an extra byte or two of source and dest addresses. It may not get you 128 bits of address but it would push out the address exhaustion problem a few centuries. It would have taken 1 guy maybe 2 days to hack it into the linux kernel (and you could even swipe a bit from the version to say whether or not this is a new-address-style packet so it could be backward compatible.) Microsoft would wait 2 years, then add support and say they invented it and are responsible for saving the world from the collapse of the internet.
But no. Instead everybody and their mother had to implement ipv6 which does nothing but add address space.
You almost can’t blame all those fucking morons. If they had just set out to solve the problem that needed solving, they could have implemented the hack ipv4 solution YEARS ago and there would never have been a problem, people would have had plenty of time to implement it before we started using the ‘extra’ address space.
But no, they had to design the next great thing which was going to solve all the problems of networking in one fell swoop. And because they’re fucking morons, they’re too dim to see that every other fucking process in the world falls apart the exact same way, and therefore could not have predicted what actually happened that ipv6 would be pared down to its one useful feature.
No, you can’t blame them, because they’re too fucking dumb.
Absolutly…
IPv6 has so many unbeliveable complicators that i doubt it will ever come reality.
I think the biggest one is handling in term of enable technicans to easy handle – using hex instead of real numbers make things less real – harder to understand to learn – which is an issue imagine how many people we need to make the hole transistion
also handle in term of managing any network – way harder with numbers you dont understand and “shortcuts” you easy overlook a :::
second one is infrastructure – just thinking about adding ipv6 to dns – think about all the domains and no its not so easy to script that shit belive me many headaches for many many hosters programmed
or if i think on my virtual hosts – i drive a mix of nat and bridging on my hypervisors (because of need of ips some ips have to split for 3 or 4 vhosts) some need max speed with min latency so they are bridged – really i dont wanna imagine to setup parallel now a real ipv6 support yet
and dont even ask about the firewalling then – transparent firewall gone to hell welcome gigabytewise firebuilder scripts lol
also the current bandaide nat solution isnt a bad thing always
i remeber a time where companys really had one real ip each pc
now they have 1-5 ips for the hole network making them natting the nessesary and hiding the rest in private ip areas
thats a good thing – now they NEED a vpn to access from outside or connect 2 networks over the internet
but ipv6 wanna take that away from us – plans no more nat, so everything has to be firewalled seperatly – no clear outside inside structures (yea i know you can still do kinda nat but this isnt whats gonna happen – we all know they will go the easy way plugin and let dhcp us the way to hell)
and thesere are just a few of upcomming problems many of em psychologicall some technicall but as you already mentioned to much revolution for one step beside unessesary specs never testet out i also see a big proivacy issue upcomming too
btw hacking current ipv4 is not the only solution, one help (not solution) would be lifting up current protocolls enable em using dns requests instead of port/ip combo like http already can (like virtual hosting)
proxy support for any protocoll (proxy/nat solutions for example)
would shrink the need of adresses within datacenters to a minimum
using virtual swichtes instead of traditional broadcast (i know means routing in the inner net instead of fast ethernet/broadcast but really most datacenters have way less innet net datatransfer so better route than broadcast)
also i see no point for mobiledevices using official adresses – if needed tunnels can be used (for that 1% of people need outside access to their phone lol)
also not much point for private households having official ipv4 addresses
beside bittorrent or similar 🙂 but here big upnp gateways for ips could to the job easy one ip for 60k+ user (beside all big ips already running virtual switches/loadbalancer and fitlering gateways possible to implement such things on that level) or or or
so many ways – if we use some of those technics and only people who need official adresses gets some – im pretty shure we do not need more addresses for a very very very long time
the argument of smart devices need access (lige fridges and stuff) is bullshit – first i dont wanna have a hacking script kiddy turning off my frdige only because if forgot to update my kitchen
second easy to have gateways where smart dev connect to it and access is over that gateway (way more secure and no difference)
so my questions is really who really need that much adresses ipv4 can provide – the point is in smarter usage in first place
I agree 100% .. Call it ipv4 + and add a 4 byte suffix to standard address. 00 to ff. Wow a trillion new addresses. And you can nat your way to backwards compat. No need to teach the world a new paradigm.
I had a different idea about how to manage the transition. Just begin rolling out a new version of IP where all addresses are variable length. Then you can just keep subnetting and subnetting and subnetting forever. Every address in the world could then be either a host or a network. But here’s the magic part: until everyone has their network stacks upgraded, you just put out a moratorium on using variable length addresses that happen to be more than 32 bits long. That’s how you eliminate the need to have a dual stack period. One day you wake up and realize all the routers and computers and applications have long since upgraded to support variable length addresses, so you start deploying 40-bit addresses. And 48-bit. Etc.
But all the conjecture in the world won’t change the fact that IPv6 with its fixed length 128-bit addresses is here to stay as the ordained solution to the IPv4 address exhaustion problem. As a network operator, I have already upgraded my network backbone (four big routers handling the full BGP table) to dual stack. As a software developer, I have already upgraded my server software to handle dual stack. It’s a nuisance but it really isn’t that hard.
Everything u said was right, and yes they are fucking morons, but u missed one key point:
THEY ACTUALLY DIDNT SOLVE THE ADDRESS SPACE PROBLEM. god damn it, how the fuck could they have missed that.
in between their mental masturbations for all the features they wanted to add in IPv6 they failed to
come up with a decent expansion of the address space.
Now I’ve heard a million times the analogies these loons use to describe how
big their address space is. Oh if we assigned 10 IPs to every electron,
we would have more ips than electrons in the galaxy and that baloney.
but none of that matters. it doesnt matter for a nanosecond, how many different numbers u could fit
in 128 bit address space. what matters is IPv6. not some theoretical discussion about 128 bits.
and ipv6 is so fucking retarded they wasted the first 64 bits with their stupid “auto magic ip based on mac address feature”
because of how retarded these ppl are,
half the address space is gone right there.
every1 is supposed to be assigned a /64 subnet.
even a p2p connection which logically by definition only has 2 nodes in it gets a /64 subnet…..
there are 18 million trillion addresses in there and we are only gonna use 2 ips.
And u know what dont even try to not assign less than a /64 for a p2p connection
because you will break everything, including neighbour discovery features and what not ….
so the engieeners who tell to drop the ipv4 conservatism and assign a /64 are right. the blame is in the design of ipv6.
the comparison between ipv4 and ipv6 is not individual ips. thats an apples to oranges comparison. the apples to apples comparison between ipv4 and ipv6 is a single IP from v4 to a /64 subnet from v6.
but it gets even worse, because even that is not quite fair. that works if starting NATing the shit out of everything in ipv6, but if my isp today started to hand me v6 ips, my modem shouldn’t get a /64. it should get a /48 so that it can assign /64s to different devices/vms/appliances internet of things and god knows what.
some ppl we could get by with a /56 per house/condo/apartment …. but no its not enough. because of internet of things its pretty easy to see how every house could have more than 256 nodes in it. not likely today but also not unlikely in my house i have easily over 10 physical machines and 50+ VMs not to mention all my Arduinos that are getting 3$ wifi connectivity thanks to ebay….
so ISPs should really hand out /48 for every subscriber ….. now they are doing a single v4 IP
so now the comparison has collapsed down to a /48 subnet from v6 for every v4 ip address.
add to that the first 3 bits that are reserved and we are left with only 45 bits …
to remind you ipv4 had 32 bits so the expansion is actually only 13 bits more.
thats right these fucking morons gave us a address space that has only 8192 times more usable space in it than the current ipv4
thats 8000 times (== 8k) (8 with three zeros in front of it.) thats it. thats fucking it. the world is gonna spend tens of billions of dollars moving to ipv6 to get 8000 times more space in it than what we have now ….
now these numbers i used, are not ass pulled… actually
https://www.ietf.org/rfc/rfc3177.txt
recommends a /48 at minimum to be assigned to every household ….
i used to think the ford mustang was the biggest insult in the history of engineering, oh i was wrong. oh i was wrong. these losers had designed the #1 to #1000 biggest insults to engineering 20 years ago and no1 paid any attention….
these animals can not understand that ipv6 is going to cost tens of billions of dollars to adopt. we were supposed to not have to change to another IP protocol except if the universe as we know it changed. except if we started to colonize other galaxies, except we discovered dark matter civilization, but what these losers have done, we are gonna have the address space too short conversation in 40 years time. oops i’ll take that back we are having that conversation now… actually we were having it
in 2005:
check this:
http://meetings.ripe.net/ripe-50/presentations/ripe50-plenary-wed-ipv6-roundtable-report.pdf
ideas have consequences. Stupid ideas have painful consequences. 10s of billions of losses means
a ton of misery to be absorbed by society. we have wasted these billions due some ppl’s remarkable arrogance and self-adorance who designed ipv6.
we are wasting these billions at a time were people are literally dying over 1000s of dollars.
people die all the time because of lack of funds…. people die because there are not enough
ambulances. people die because they have to (not want to) work in coal mines.
people die of disease, because 10 years ago a dollar did not buy enough computing power
to simulate protein structures and thus create a cure drug….
we will have to move the world into ipv7 when ipv6 is exhausted
or implement NAT everywhere
or come up with new software that works with less than /64 subnets
all of these options cost billions and thus cost lives.
but i agree u cant blame them cause its not their fault, they are just dumb-fucking-retarded.
u can blame them no more than u can blame a goat for failing to discover newtonian physics and improve the lives of the goat species.
I think I like this guy.
I remember when I first heard that they wanted to assign a /64 to every household that it seemed to me a bit overkill. I find it easier to imagine godzilla than the idea that a single household would ever have use for an entire current-sized-internet worth of addresses in their house.
Perhaps the plan is to assign an ip to each unique surface. I’m sure that has value somehow.
Maybe it’s all just a big marketing joke.
@sam, as much as I hate IPv6 for the reasons you’ve mentioned above (but hey, they could have integrated MPEG, bitcoin and Invigilation Header), I think you’ve mistaken the numbers. Delegating /48 to the end node shrinks global address space to 80 bits (128-48).
What IPv6 didn’t solve is:
1. routing – IPv4 BGP holds 0.5M+ routes, what happens when IPv6 comes to live (for real) is frightening – while there should be no more prefixes than active AS numbers, longer prefixes would still be announced unaggregated for traffic engineering purposes,
2. security – IPSec is so complicated that is not widely used. Except when in corporate environment with IT department. IPv6 doesn’t change anything – you still need SSL/TLS implemented in every client and server application.
3. mobility – connection is still established between IP addresses, just like in IPv4, so it’s not transferrable into different network. There are approaches deriving IP addressess from end-node identifiers (like Host Identity Protocol), so the live session can change underlying service providers, which is plain and simple, but that would be too easy – let some HomeAgent do the job (hahaha, who’s going to maintain these?)
My favorite nonsense that some spew about IPv6 is how wonderful each device having its own Internet facing IP will be. While NAT wasn’t implemented as a security layer, it has become one–one of many layers that should be in place.
I read recently some professional Admin’s post about how not having NAT in IPv6 by default is a wonderful thing, as NAT is NOT any form of security. He had to be getting paid from someone for that post, as that is like saying your underwear isn’t part of keeping you warm in the winter because it isn’t a sweater or coat.
My personal view on IPv6 is that they made a future situation untenable. All they had to do was extended IPv4 in some way. But then I think that the REAL purpose for IPv6 was the reported subversion of IPSec and the forcing of all devices to be Internet facing. Or put another way, anything Micro$oft is gung-ho about is something whose skirt I want to look up and check for a noodle. But that is another topic for another day.
I don’t know crap, but I know a lot of crap.
2017 still not much progress
however isps in my city now handing out ipv6 only products (going over an ipv4 gateway).
aaaaaand its horrible.
you get an dhcp you cant change, you get adresses widly assigned no way to change a bit
the firewall is by default to absolute no security and all devices are exposed to the internet.
and that got be thinking. one thing nobody ever mentioned about nat is that above all (additional security, additional little work) its one thing. independence.
no matter if you run multiwan or change your isp, all you need is one IP for most normaly networks and your firewall. multiwan? no issue, you nat the shit out of your gateways for all needed incommings, do little dynamic dns or another failover provedure and done.
but what with all internet facing ipv6? you cant change provider or you loose your ipv6 range. so you need to be dam shure to have an dam good internal dns up and running to be able to switch quickly without interruption.
but multiwan – well then you need your providers assistance or – forget it
even if you buy and pay your own adress space, its a PIA to convince your providers to configure it for multiwan
no man nat isnt evil at all, it decomplicate things by the order of a magnitude.
and who the fuck need internet adresses within their own network?
nobody. not at all.
I hadn’t considered the ipv4 owning-your-own-range via nat thing. I like that idea. You’ve sold me.
IPV4 FOREVER!
I’m reading this in 2017 because I was wondering how far IPv6 adoption has come by now… surely it must’ve been close to finished? Ehhhh no, all I see all the time anywhere is IPv4 addresses. Something’s not working there 😉
2017 is more than 50% but IPv6 is still waiting for the great day. I think IPv6 is died, but nobody want to see that.
I’m not a fan of IPv6 either. A group of us came up with an IPv4 address extension scheme (similar in spirit to above) and implemented it in Linux. Go to the web site http://www.enhancedip.org/learnmore.html to get the cliff notes for how it works. We’d be interested in positive (this is cool I would use it if Linux and other OSs incorporated it) and negative feedback (this will never work and here’s why).
Nicely done. 🙂
This is an interesting article and comments. Although this was started many years ago, the last couple comments are recent enough and referred to a piece of work that I reviewed during our study. I like to share what we did.
1) A few years ago, we accidentally ventured into studying the IPv4 address pool exhaustion challenge, perhaps due to the curiosity from our telephony background. We now have submitted a proposal called EzIP (phonetic for Easy IPv4) to IETF:
https://tools.ietf.org/html/draft-chen-ati-adaptive-ipv4-address-space-03
Basically, the EzIP approach will not only resolve IPv4 address shortage issues, but also largely mitigate the root cause to cyber security vulnerabilities, plus open up new possibilities for the Internet, all within the confines of the IPv4 domain by utilizing the original IPv4 standard RFC791 and the long-reserved yer hardly used 240/4 address block to expand each IPv4 public address by 256M fold. These should relieve the urgency to deploy the IPv6 for an appreciable length of time. However, since this work is not aligned with any of IETF’s active Working Groups, it has not been granted the permission to publish.
2) Although many already regarded IPv4 as a bygone topic sometime ago, our efforts were inspired and guided by two regularly updated world statistics:
https://ams-ix.net/technical/statistics/sflow-stats/ether-type
https://stats.labs.apnic.net/ipv6
So, we thought that the initial application for EzIP would be emerging regions and rural areas of developed countries where IPv4 addresses are in short supply.
3) A recent article about IPv6 adoption status stated that the US Federal Agencies have been proceeding at “a glacial pace”. This is a big surprise because it seemed to debunk the theories behind why IPv6 roll-out has been so slow, such as limited financial resources, no peer pressure, etc. The issue must be more fundamental than skin deep.
https://dyn.com/blog/ipv6-adoption-still-lags-in-federal-agencies/
4) Then, we found a recent web article reporting about debates between UN’s ITU-T and Internet’s IETF:
https://www.theregister.co.uk/2018/05/30/internet_engineers_united_nations_ipv6/
5) Apparently, ITU-T is drafting a Recommendation on IPv6RefModel that is opening up a series of reviews of the IPv6 designs, perhaps starting from the basic philosophy of not backward compatible to IPv4. These made IETF rather uptight. But, I believe these are would lead to a robust Internet that will be benefit the general public.
https://www.ripe.net/participate/ripe/wg/ipv6/documents/itu-ipv6refmodel
I would appreciate very much of any thoughts and comments.
Abe (2018-07-22 20:36)
Hahaha this post is just made my day.
It was in one of those moments when it didn’t matter how hard I try to disable the god damn ipv6 in x11 vnc it just don’t want to do it
regardless that I have disabled it on kernel level with sysctl 😀
23/08/2018 15:04:29 Listening for VNC connections on TCP port 5901
23/08/2018 15:04:29 Listening for VNC connections on TCP6 port 5900
So what the hell I just shoot an “I hate ipv6″ google search what landed me at your post and I agree with 100 % in it.
IPV6 was a c00l h4xx back in my age 18 when hanging out on irc with bouncers like im.the.king.of.irc.org wohoo but years went buy working in the industry I say otherwise.
I moved from small to midsized companies and none of them used ipv6 nowhere, not on lan, not on dmz, not on public internet. There are not even talks of it in 2018.
IPv6 is an ultimate fucking nightmare for sysadmins and netadmins. I can remember ipv4 addresses for a lot of internal servers and devices but no way on earth that I could do the same with v6.
They talk about the big security things v6 solves, on the contrary if you run public servers with ipv6 enabled, there is a good chance a lot of your daemons listening on v6 already and your normal ipv4 iptables rules
will let all those connections to happen but that just brings us to the next nightmare configuring ipv6 based firewalls where you have to buy a god damn 48” monitor to be able to see 1 rule per line.
I hate ipv6 as much as systemd, noone wanted it, noone really needed it, some pepole are just forced it down on our throats.
If the IPV4 addresses would have been justly allocated (in proportion to a country/organizations size) then we were nowhere near to running out.
But what happened is that the early adopters such as: Xerox, IBM, MIT, Microsoft and just to say US Corporate and Govt entities suckered up whole class A /8 IP ranges.
That is also 17 million ip for a shitty university which could be put behind 2-3 NAT gateways.
Although IPV6 is still at least 30 years away from wide spread use there might be a chance that will never happen. Until that just put:
net.ipv6.conf.all.disable_ipv6=1
net.ipv6.conf.default.disable_ipv6=1
net.ipv6.conf.lo.disable_ipv6=1
to your /etc/sysctl.conf and good buy iptardV6!
This is one of the best sites and comments I’ve read for some time.
I love you “fuck-ipv6-people” !!!
I hate IPv6 because I can memorize an IPv4 address but can’t an IPv6 address, so I refuse to use it.
and here we go again, still hating ipv6, still no adoption.
my datacenters giving me ipv6 for free in the trillions yet id rather pay for each ipv4 address
i think the biggest issues, of them all, is readability. whoever created ipv6 did never manage networks, thats the only explanation why it isnt so obivious why ipv6 is evil.
see every IP adress or at least range has to be handled manually at least once at rollout. now the issue is, those ranges can never be really that big. every network is segmented, so no matter how much you automate in that process
you need to put it somewhere at least for once.
then when something is wrong, you need to handle some adresses again manually. if smokepinping, tracing, mtring,..
everytime you do that youre already usually in a complicated places full of routes and firewall rules, latencys and packetlosses…
you cant have a gazillion hex digits
it get worse if you only have a spice terminal and cant copy paste that shit, then you need to type it. its annoying enough with ipv4, but with 6 you gonna search for a gun and anti depressents.
you look torugh your firewall protocoll to see why certain traffic is unwantingly blocked, you need to differenciate between allowed and not allowed addresses
no matter where you look, having the correct ips is often relevant.
for security and or function and or finding an error. often jsut to identify who is comming here and when (you know the secret life of your vms is often revelaed at the firewall)
this is plain and simple not to be handled by ipv6. its hard enough on ipv4 sometimes. but with 6 the error rate will skyrocket.
the costs of maintining things will rise. we will have more issues, longer timeframe to solve em, and therefore higher costs.
if we look at it globally the decision to make unreadable adresses, will costs billions anually.
ipv6 shortcuts are the worst as they are now automatic and at output mandatory. did you knwo that ipv4 has shortcuts too ? no? well it does, just nobody ever used em. for good reason.
nothing worst than having logs with the same data but diffeerent ip lenghts.
Dear Colleagues:
1) Please have a look at the following discussion thread on the “state of IPv6”. The cause may be within.
http://www.circleid.com/posts/20190529_digging_into_ipv6_traffic_to_google_is_28_percent_deployment_limit/
2) Then, you may like to have a look at the feasibility demonstration report below about our proposal for expanding IPv4 address pool, etc.:
https://www.avinta.com/phoenix-1/home/RegionalAreaNetworkArchitecture.pdf
These should provide some material for furthering the dialog.
Abe (2020-08-29 23:22 EDT)