Related: the name for your Tailnet is randomly-generated from a list of 'tails', and a list of 'scales' in the format `<a tail>-<a scale>.ts.net`.
Many puns and easter-eggs in both! Recommend giving it a quick glance, some are pretty fun. Some examples include 'fable' as a 'tail' (IE 'tale'), and 'universe' as a 'scale'.
Words can be present in both lists when the entity is applicable to both a tail and a scale, so a tailnet named `koi-koi.ts.net` is theoretically possible too.
Found out about it after I randomly lucked into rolling 'miku' as my 'tail' when first creating my account :)
I think you can only pick a named network once, and it’s an active choice in some setting. There’s some impact to TLS and DNS as well. I haven’t done it yet either, so mine is hexadecimal as well.
This makes me think about something. Why don't we support this as a standard?
It would be nice if local ipv6 spelled out "local"
imagine this: local::homelab in your url-bar just working
edit: Yes I know DNS solves this problem, but it would be very neat non the less, making 128bits of packet overhead actually serve a purpose without any software glue.
The `.local` TLD exists as a zeroconf domain name for this purpose on local networks. Your homelab can advertise as `homelab.local` without setting up custom DNS servers.
Also, link local is almost always `fe80::` [0]. Today for privacy (and simplicity) reasons most devices pick a random address under that like `fe80::6923:83ab:201:273e`. But most operating systems allow configuring a static link address so if you really wanted you could use `fe80::403e:7ab` for something close to "homelab" (and `fe80::cafe` for your coffee maker) you absolutely can.
[0] The allocated range is fe80::/10 so fe(8|9|A|B)[0-F], but in practice it seems rare for devices today to use anything but fe80::.
mDNS my goat. I have a lot of spare machines and I do a lot of various experiments on said machines, some of which could brick the system even on NixOS. I now have a custom usb installer over mDNS and reflashing any compute I have from scratch for the express purpose of destroying it in a few days hasn't been easier
In a lot of countries that were not lucky enough to get online early, the ISPs were not able to get IPv4 for every CPE WAN interface. If you go to the by-country tab, you'll find a number of countries are >70%:
The fact that the US has a lot of IPv4, and a lot of software companies are in the US, may mean that a lot of software is IPv4-centric just because the US-centric outlook can ignore IPv6 with little personal impact. If more devs that to dogfood being behind a CG-NAT I think things would change.
Mobile networks and consumer home use. "Happy eyeballs" protocols runs on most consumer devices at this point (preferring IPv6 over v4 at all times unless v6 is intolerably slower).
This has forever been and will forever be the dumbest anti-IPv6 complaint.
How often do you really need to communicate a complete IP address manually, human to human, via voice or handwriting where you can't just copy/paste?
I mean REALLY NEED. Not just "I'm too lazy to set up DNS", not some hypothetical nonsense where you've found yourself on a network that has functional IPv6 but all autoconfiguration beyond SLAAC is disabled, etc. It just doesn't happen at a rate sufficient to justify how much people like you whine about it.
DNS works. You set it up correctly once when you provision the system, it stays working, you never need to think about a raw IP address again until it's time to deploy a new system or something weird happens like changing a NIC. If your DNS is that broken that you can't depend on it, fix that first and stop worrying about trying to remember complete addresses. If you need to tell a colleague what system you're looking at and DNS is not properly configured for that system, "the one ending in 44a1" is more than sufficient 99% of the time in the same way "the one ending in 127" was for v4 and the rest of the time you have computers and copy/paste works.
Actually yesterday I was at a client doing a demonstration. I am a software developer, but not an IT/network guy, so I just knew to connect to their network, look at ipconfig and tell them the IP of my computer so they could connect to my machine to test the software. I just look at it and said "i'm at zero point fifty four" (it was 192.168.0.54). How would this look in IPv6? (It's a genuine question, because from what I see it would be something like "i'm at f e seven four colon colon colon zero one a", it honestly adds friction.
mDNS autodiscovery by name would be my preferred solution, which should work by default on major OSes as long as the network operator or device administrator hasn't gone out of their way to break it. <yourhostname>.local
Like I said in my previous post, for the most part if you need to actually communicate the raw IP that probably means something has gone wrong and you're now troubleshooting with someone who should have access to a higher level view of the network where "the one ending in 3 charlie echo 5" should be more than sufficient just like it would be if you were communicating a MAC address while troubleshooting DHCPv4 issues.
In the 12 years I've had a dual-stack network I've never had to communicate a full IPv6 address verbally, but if I did I'd pretty much do it the same way I would an IPv4 address. One group at a time, with a pause implying the separator. If there was one or more zero subnets involved that could be shortened with double-colon I'd say that.
For example the record I just got back for Google of 2607:f8b0:4023:141b::71 I'd read off as:
two six zero seven (or maybe twenty-six oh seven)
foxtrot eight bravo zero
four zero two three (or maybe forty twenty-three)
one four one bravo
double colon
seven one
edit: HN makes it way harder than it should be to have multiple lines.
If you have a Mac or Windows laptop, since they configure mDNS by default, you could say <my-machine>.local. The client will do a mDNS query and get your IPv4 or IPv6 address and connect.
That was the whole point of mDNS.
This will work even on a pure ipv6 network (with no hosts having ipv4 addresses).
I have a completely default single router setup at home. I know .50 is my desktop and .1 is my router. How do I reach that via mDNS instead? Pretty sure this is broken.
~ $ arp -a
? (192.168.50.1) at e8:9c:25:52:83:10 on en0 ifscope [ethernet]
? (192.168.50.9) at 64:9a:63:eb:ed:f1 on en0 ifscope [ethernet]
? (192.168.50.50) at 14:98:77:70:17:a7 on en0 ifscope [ethernet]
On a Mac, run `dns-sd -B _ssh._tcp` to find all machines advertising SSH on your local network. From there, you can either use the advertised `.local` hostname directly, or you can pass the advertised friendly name to `dns-sd -G v4` to resolve it to an IPv4 address.
AFAIK you must always specify a service like `_ssh._tcp` or `_http._tcp`; you can’t ask dns-sd to just dump all the advertised mDNS services on your network. That might be possible with the underlying `nw_browser_t` APIs, though.
Ok so this one works for my Mac (.50) in particular but not my RPi on the same network with an SSH server, because turns out it doesn't advertise that by default.
"you can’t ask dns-sd to just dump all the advertised mDNS services on your network": I ran into this too. So besides the defaults and maybe network itself standing in the way, the tooling around mDNS is user-unfriendly.
Only if you know the name already and no two hosts took the same name, assuming it works at all. Shouldn't those question marks be the hostnames on arp -a? On some networks it shows them, on others it doesn't.
I haven’t used Linux in over 10 years on a laptop, but even back then this was very well supported there as well. `ssh foobar.local` just worked, for example.
Just as LAN addresses (link local) are usually 192.168.$x.$y and you are just passing $x.$y, with IPv6 they are usually `fe80::$z` and in theory you could just pass $z and the person on the other end can assume `fe80::`. Because of how double colon works $z could be just 54 in a similar situation (especially if link local addresses are statically assigned like most workplaces try to statically assign IPv4 addresses).
Same with ULA addresses like 10.$x.$y.$z for private networks. Most of those you can assume `fd$x::$z` and just need to pass whatever is on the other side of the double colons. (Where $x is optional and likely network-wide.)
They're long though. Just checked and my desktop has fe80::1068:f88b:8b1b:4e45 for en0. They can't be short because of possible collisions, unless you're using DHCP or manually allocating static ones, and DHCP6 isn't broadly supported on hosts or routers.
Sure, most systems default to SLAAC which creates random long addresses. You can manually allocate them just fine and short addresses aren't going to cause collisions (and if they did, SLAAC handles it). Some of the networking people most (ab)using 192.168.x.y heavily are manually allocating them.
DHCP6 is pretty well supported on a lot of Enterprise routers, so the fd:: ULA space, which is what DHCP6 is most good for, has no required reason to be long addresses (outside of silly things like this link's mentioned Tailscale fingerprinting/advertising IP).
When, every single network I've ever used has had a broken mDNS. And there's no way I'm managing a DNS server just to have memorable addrs, especially when it's not my own network.
If you want a real example, I was at a nonprofit event helping with their video livestream. They have a Zowiebox that displays its IP address on a digital display. I reached its web management portal that way on their Windows PC that ofc had no nmap etc installed. Oh and that box is IPv4-only.
How can a network have “a broken mDNS”? mDNS is a distributed protocol spoken by the devices on it, not something centrally/administratively provided like DHCP.
Not saying it's centrally administered, but there are ways it can be centrally broken. I honestly don't even know why mDNS doesn't work on my home network, I didn't change anything.
> How often do you really need to communicate a complete IP address manually, human to human, via voice or handwriting where you can't just copy/paste?
Uhh, that one time in 2003 over Teamspeek, but it was really cool and really important, I promise! Definitely important enough to justify the horror that is NAT.
Related: the name for your Tailnet is randomly-generated from a list of 'tails', and a list of 'scales' in the format `<a tail>-<a scale>.ts.net`.
Many puns and easter-eggs in both! Recommend giving it a quick glance, some are pretty fun. Some examples include 'fable' as a 'tail' (IE 'tale'), and 'universe' as a 'scale'.
Words can be present in both lists when the entity is applicable to both a tail and a scale, so a tailnet named `koi-koi.ts.net` is theoretically possible too.
Found out about it after I randomly lucked into rolling 'miku' as my 'tail' when first creating my account :)
https://github.com/tailscale/tailscale/tree/main/words
an easter egg in the easter egg: https://github.com/tailscale/tailscale/commit/f8ad538a3ba0b9...
I added "wyrm" into both lists a while back to make wyrm-wyrm possible for some lucky individual.
Hmm, did they always have this? Mine is `tail<a few random letters + numbers>`.ts.net.
I think you can only pick a named network once, and it’s an active choice in some setting. There’s some impact to TLS and DNS as well. I haven’t done it yet either, so mine is hexadecimal as well.
Are we counting the long hair as a "tail" or something?
Facebook does this too:
They might be the first to do this. Google could have done it easily with 6006:1e, but they never took the opportunity.
This makes me think about something. Why don't we support this as a standard?
It would be nice if local ipv6 spelled out "local"
imagine this: local::homelab in your url-bar just working
edit: Yes I know DNS solves this problem, but it would be very neat non the less, making 128bits of packet overhead actually serve a purpose without any software glue.
The `.local` TLD exists as a zeroconf domain name for this purpose on local networks. Your homelab can advertise as `homelab.local` without setting up custom DNS servers.
Or .internal if you do setup your own DNS servers.
.internal is reserved by ICANN, but there is no IETF RFC ('just' a draft (for now)):
* https://en.wikipedia.org/wiki/Top-level_domain#Reserved_doma...
* https://en.wikipedia.org/wiki/Special-use_domain_name
* https://en.wikipedia.org/wiki/.internal
* https://datatracker.ietf.org/doc/html/draft-davies-internal-...
Not sure how important that is.
We want IPs to be distinctive from domains, having real words in both would be highly confusing
dots versus ::
Maybe because IPv6 digits are hexadecimal? and we also have DNS service for that.
> Maybe because IPv6 digits are hexadecimal?
Mostly for us humans of course.
IPv4 addresses are generally in base-10, but binary and octal can also technically be used (on the CLI).
My favorite IPv4 schema is a uint32. Lots of squinting and head scratching every time it pops up
mDNS already serves homelab.local style addresses.
Also, link local is almost always `fe80::` [0]. Today for privacy (and simplicity) reasons most devices pick a random address under that like `fe80::6923:83ab:201:273e`. But most operating systems allow configuring a static link address so if you really wanted you could use `fe80::403e:7ab` for something close to "homelab" (and `fe80::cafe` for your coffee maker) you absolutely can.
[0] The allocated range is fe80::/10 so fe(8|9|A|B)[0-F], but in practice it seems rare for devices today to use anything but fe80::.
The 128 bits are serving a purpose alright: Globally unique host addressing.
Seems like you want to repurpose them to something completely different (what exactly)?
And what is “interpret this part of an IPv6 address completely differently than in all other contexts” if not software glue?
mDNS does that. Your devices are already accessible on your network as eg “himatas-iphone.local” with no configuration required.
mDNS my goat. I have a lot of spare machines and I do a lot of various experiments on said machines, some of which could brick the system even on NixOS. I now have a custom usb installer over mDNS and reflashing any compute I have from scratch for the express purpose of destroying it in a few days hasn't been easier
mDNS is unreliable
I don't have to imagine, in fact its even easier. I can just go to http://brother in by browser and it goes to my brother printer. Incredible!
Is there a technical reason it starts at index 2 instead of 0?
The `fd` part indicates it's a ULA, a local private ip address. (like 192.168.0.x)
Sound it out with your mouth: "What DNS do you use?"
"One hundred one hundred one hundred one hundred"
or
"one hundred, 4 times"
vs.
"f d seven a colon one one five c colon a one e zero colon colon five three"
This is just a rant but IPv6 will never be widely adopted.
Almost half of traffic to Google is ipv6: https://www.google.com/intl/en/ipv6/statistics.html
That’s because of mobile networks predominately using IPv6, correct.
Not because of home or B2B use.
> Not because of home or B2B use.
Depends on your home. In the US Comcast has had IPv6 for a while:
* https://corporate.comcast.com/comcast-voices/ipv6-deployment...
In a lot of countries that were not lucky enough to get online early, the ISPs were not able to get IPv4 for every CPE WAN interface. If you go to the by-country tab, you'll find a number of countries are >70%:
* https://www.google.com/intl/en/ipv6/statistics.html#tab=per-...
The fact that the US has a lot of IPv4, and a lot of software companies are in the US, may mean that a lot of software is IPv4-centric just because the US-centric outlook can ignore IPv6 with little personal impact. If more devs that to dogfood being behind a CG-NAT I think things would change.
Mobile networks and consumer home use. "Happy eyeballs" protocols runs on most consumer devices at this point (preferring IPv6 over v4 at all times unless v6 is intolerably slower).
It's mainly just B2B stacks lagging behind today.
This has forever been and will forever be the dumbest anti-IPv6 complaint.
How often do you really need to communicate a complete IP address manually, human to human, via voice or handwriting where you can't just copy/paste?
I mean REALLY NEED. Not just "I'm too lazy to set up DNS", not some hypothetical nonsense where you've found yourself on a network that has functional IPv6 but all autoconfiguration beyond SLAAC is disabled, etc. It just doesn't happen at a rate sufficient to justify how much people like you whine about it.
DNS works. You set it up correctly once when you provision the system, it stays working, you never need to think about a raw IP address again until it's time to deploy a new system or something weird happens like changing a NIC. If your DNS is that broken that you can't depend on it, fix that first and stop worrying about trying to remember complete addresses. If you need to tell a colleague what system you're looking at and DNS is not properly configured for that system, "the one ending in 44a1" is more than sufficient 99% of the time in the same way "the one ending in 127" was for v4 and the rest of the time you have computers and copy/paste works.
Actually yesterday I was at a client doing a demonstration. I am a software developer, but not an IT/network guy, so I just knew to connect to their network, look at ipconfig and tell them the IP of my computer so they could connect to my machine to test the software. I just look at it and said "i'm at zero point fifty four" (it was 192.168.0.54). How would this look in IPv6? (It's a genuine question, because from what I see it would be something like "i'm at f e seven four colon colon colon zero one a", it honestly adds friction.
mDNS autodiscovery by name would be my preferred solution, which should work by default on major OSes as long as the network operator or device administrator hasn't gone out of their way to break it. <yourhostname>.local
Like I said in my previous post, for the most part if you need to actually communicate the raw IP that probably means something has gone wrong and you're now troubleshooting with someone who should have access to a higher level view of the network where "the one ending in 3 charlie echo 5" should be more than sufficient just like it would be if you were communicating a MAC address while troubleshooting DHCPv4 issues.
In the 12 years I've had a dual-stack network I've never had to communicate a full IPv6 address verbally, but if I did I'd pretty much do it the same way I would an IPv4 address. One group at a time, with a pause implying the separator. If there was one or more zero subnets involved that could be shortened with double-colon I'd say that.
For example the record I just got back for Google of 2607:f8b0:4023:141b::71 I'd read off as:
two six zero seven (or maybe twenty-six oh seven)
foxtrot eight bravo zero
four zero two three (or maybe forty twenty-three)
one four one bravo
double colon
seven one
edit: HN makes it way harder than it should be to have multiple lines.
If you have a Mac or Windows laptop, since they configure mDNS by default, you could say <my-machine>.local. The client will do a mDNS query and get your IPv4 or IPv6 address and connect.
That was the whole point of mDNS.
This will work even on a pure ipv6 network (with no hosts having ipv4 addresses).
I have a completely default single router setup at home. I know .50 is my desktop and .1 is my router. How do I reach that via mDNS instead? Pretty sure this is broken.
On a Mac, run `dns-sd -B _ssh._tcp` to find all machines advertising SSH on your local network. From there, you can either use the advertised `.local` hostname directly, or you can pass the advertised friendly name to `dns-sd -G v4` to resolve it to an IPv4 address.
AFAIK you must always specify a service like `_ssh._tcp` or `_http._tcp`; you can’t ask dns-sd to just dump all the advertised mDNS services on your network. That might be possible with the underlying `nw_browser_t` APIs, though.
Ok so this one works for my Mac (.50) in particular but not my RPi on the same network with an SSH server, because turns out it doesn't advertise that by default.
"you can’t ask dns-sd to just dump all the advertised mDNS services on your network": I ran into this too. So besides the defaults and maybe network itself standing in the way, the tooling around mDNS is user-unfriendly.
This is to discover services, which you really shouldn't need to do by hand outside of troubleshooting, your apps should be doing it themselves.
For address discovery which is what we're really talking about here it's just "<hostname>.local"
I was answering the question of “how do I discover another host’s mDNS hostname?” It turns out that is a per-service question.
<hostname>.local
Here's me pinging my laptop (macOS 27) from my file server (Ubuntu 24.04)
Only if you know the name already and no two hosts took the same name, assuming it works at all. Shouldn't those question marks be the hostnames on arp -a? On some networks it shows them, on others it doesn't.
I haven’t used Linux in over 10 years on a laptop, but even back then this was very well supported there as well. `ssh foobar.local` just worked, for example.
Just as LAN addresses (link local) are usually 192.168.$x.$y and you are just passing $x.$y, with IPv6 they are usually `fe80::$z` and in theory you could just pass $z and the person on the other end can assume `fe80::`. Because of how double colon works $z could be just 54 in a similar situation (especially if link local addresses are statically assigned like most workplaces try to statically assign IPv4 addresses).
Same with ULA addresses like 10.$x.$y.$z for private networks. Most of those you can assume `fd$x::$z` and just need to pass whatever is on the other side of the double colons. (Where $x is optional and likely network-wide.)
They're long though. Just checked and my desktop has fe80::1068:f88b:8b1b:4e45 for en0. They can't be short because of possible collisions, unless you're using DHCP or manually allocating static ones, and DHCP6 isn't broadly supported on hosts or routers.
Sure, most systems default to SLAAC which creates random long addresses. You can manually allocate them just fine and short addresses aren't going to cause collisions (and if they did, SLAAC handles it). Some of the networking people most (ab)using 192.168.x.y heavily are manually allocating them.
DHCP6 is pretty well supported on a lot of Enterprise routers, so the fd:: ULA space, which is what DHCP6 is most good for, has no required reason to be long addresses (outside of silly things like this link's mentioned Tailscale fingerprinting/advertising IP).
When, every single network I've ever used has had a broken mDNS. And there's no way I'm managing a DNS server just to have memorable addrs, especially when it's not my own network.
If you want a real example, I was at a nonprofit event helping with their video livestream. They have a Zowiebox that displays its IP address on a digital display. I reached its web management portal that way on their Windows PC that ofc had no nmap etc installed. Oh and that box is IPv4-only.
How can a network have “a broken mDNS”? mDNS is a distributed protocol spoken by the devices on it, not something centrally/administratively provided like DHCP.
Not saying it's centrally administered, but there are ways it can be centrally broken. I honestly don't even know why mDNS doesn't work on my home network, I didn't change anything.
That’s really strange, it’s definitely something your networking hardware needs to actively do, e.g. by filtering/not forwarding multicast traffic.
> How often do you really need to communicate a complete IP address manually, human to human, via voice or handwriting where you can't just copy/paste?
Uhh, that one time in 2003 over Teamspeek, but it was really cool and really important, I promise! Definitely important enough to justify the horror that is NAT.
It could have easily just been "f d zero zero colon colon 1", They're just trying to be fancy with their network addresses.
99.999999% of human population would never even know what DNS is. You are yelling at cloud.
"Nobody needs human-memorable IP addresses"