A new generation of Tailscale access controls(tailscale.com)
tailscale.com
A new generation of Tailscale access controls
https://tailscale.com/blog/grants-ga
68 comments
I reckon this (and that it can be used in conjunction with grants) implies they're translating the old config to the new format if it exists on the fly. Otherwise, if it involved running two things one in maintenance mode (and somehow ensuring they both take effect and don't conflict) it'd be ballsy to promise 'forever'.
Although not sure why that's preferable to converting it once-off, so it's still not something a user has to worry about, but then the old way is done with.
Although not sure why that's preferable to converting it once-off, so it's still not something a user has to worry about, but then the old way is done with.
Stripe has been doing it for god knows how long now and it seems to be working for them. With a robust test suite, I can’t think of too much that can/would go wrong. Honestly, I wish more software would approach things this way instead of introducing a major api version every 5 years. Just continuously ship the latest thing and translate everything that came before. Hopefully if people are also using “official” client libraries, they will eventually upgrade their dependencies and their API version will also get bumped. If there is a particularly heavy usage client jumping forward 100 versions and using too many of your resources, hopefully your org can dedicate some resources to help them upgrade which theoretically benefits both of you.
Now that I think of it Adyen did this to us a few years back. We’ll probably stay in the v5 branch until they politely poke us to upgrade again in another 5 years…
Now that I think of it Adyen did this to us a few years back. We’ll probably stay in the v5 branch until they politely poke us to upgrade again in another 5 years…
> Although not sure why that's preferable to converting it once-off, so it's still not something a user has to worry about, but then the old way is done with.
This way also supports users who have scripts or workflows that produce the old form of ACLs. If you had a single cut-off date, those scripts would presumably need to be updated.
This way also supports users who have scripts or workflows that produce the old form of ACLs. If you had a single cut-off date, those scripts would presumably need to be updated.
Post author here!
We are indeed converting the old syntax to grants behind the scenes. In fact, we have been quietly doing this for months :)
As tczMUFlmoNk mentioned further down, we wanted to allow both syntaxes in perpetuity to account for scripts, Terraform configs, version-controlled policy files, and other automation that may rely on the original syntax.
We will provide a button on the JSON editor in our admin console to do a one-click conversion for anyone who would like to make a hard switch.
We are indeed converting the old syntax to grants behind the scenes. In fact, we have been quietly doing this for months :)
As tczMUFlmoNk mentioned further down, we wanted to allow both syntaxes in perpetuity to account for scripts, Terraform configs, version-controlled policy files, and other automation that may rely on the original syntax.
We will provide a button on the JSON editor in our admin console to do a one-click conversion for anyone who would like to make a hard switch.
How did they react to Headscale? I must’ve missed the announcement.
I'm not sure what they meant, but Tailscale links to Headscale and says this:
> Headscale is an open source alternative to the Tailscale coordination server and can be self-hosted for a single tailnet. Headscale is a re-implemented version of the Tailscale coordination server, developed independently and completely separate from Tailscale. Headscale is a project that complements Tailscale — with its own independent community of users and developers. Tailscale does not set Headscale’s product direction or manage the community, and neither prohibits nor requires employees from contributing to Headscale.
> Our opinion is that Headscale provides a valuable complement to Tailscale: It helps personal users better understand both how Tailscale works and how to run a coordination server at home. As such, Tailscale works with Headscale maintainers when making changes to Tailscale clients that might affect how the Headscale coordination server works, to ensure ongoing compatibility.
https://tailscale.com/opensource#encouraging-headscale
> Headscale is an open source alternative to the Tailscale coordination server and can be self-hosted for a single tailnet. Headscale is a re-implemented version of the Tailscale coordination server, developed independently and completely separate from Tailscale. Headscale is a project that complements Tailscale — with its own independent community of users and developers. Tailscale does not set Headscale’s product direction or manage the community, and neither prohibits nor requires employees from contributing to Headscale.
> Our opinion is that Headscale provides a valuable complement to Tailscale: It helps personal users better understand both how Tailscale works and how to run a coordination server at home. As such, Tailscale works with Headscale maintainers when making changes to Tailscale clients that might affect how the Headscale coordination server works, to ensure ongoing compatibility.
https://tailscale.com/opensource#encouraging-headscale
The distilled version of this is that they let Headscale, a FOSS, use their proprietary client to connect to the Headscale server rather than Tailscale’s. They don’t need to allow this, they could easily disable that feature and have every right to, but they don’t. This makes the development of Headscale a lot more straightforward since the entire client codebase is just removed from the equation.
The tailscale client is not proprietary.
The GUIs for Windows, macOS, and iOS/iPadOS are proprietary.
GP phrased it a confusing way around imo - proprietary or not, they don't need to give the client the capability to target a non-Tailscale server!
An excellent way to position yourself towards an open source alternative. Everything they do seems to be grounded in an engineering mindset and with respect to users.
What are some other companies like that?
The ones that have an MBA-driven anti-user mindset IMO:
The ones that have an MBA-driven anti-user mindset IMO:
Apple, Google, Microsoft, Meta, Oracle, Adobe, Nvidia, AMDI think they've figured out a way to align their interests with what's truly good for their users. I'm sure a lot of their sales funnel is just nerds like us using Tailscale at home, and then bringing it to their employers.
We're really happy Tailscale users at home. No bullshit, just works.
We're really happy Tailscale users at home. No bullshit, just works.
> I think they've figured out a way to align their interests with what's truly good for their users. I'm sure a lot of their sales funnel is just nerds like us using Tailscale at home, and then bringing it to their employers.
I can't say for certain (I do not work for TailScale, nor have I ever) but I suspect you are right.
TailScale was _immediately_ surfaced in our team's "what could we replace open-vpn with?" conversation recently... precisely because most of us _are_ using it at home and have nothing but praise.
I can't say for certain (I do not work for TailScale, nor have I ever) but I suspect you are right.
TailScale was _immediately_ surfaced in our team's "what could we replace open-vpn with?" conversation recently... precisely because most of us _are_ using it at home and have nothing but praise.
> I'm sure a lot of their sales funnel is just nerds like us using Tailscale at home, and then bringing it to their employers.
Honestly, I don't know why this stopped being the dominant strategy. Matlab, Adobe, Solidworks, and many other tools used to be free for students. It's not hard to see how people working with certain software while in school translates to using that same software outside. The same is true for at home! You get frustrated with whatever BS thing your work has and then ask your boss if you can use X instead.
I think it stopped because they got too big. Becoming the de facto tool they figured they could charge students and hobbiests too. I'm pretty sure this just lead to more piracy and open source solutions. I mean Coca-Cola still advertises... they sure aren't doing it for brand recognition... everyone knows who they are (same with McDonalds!). They do it to associate feelings. These companies should still do those student and hobbiest versions for the same reasons. It prevents others from displacing you. Companies aren't going to complain about the price and pirate your stuff instead, they have the money. Only hobbiests and students do that stuff...
There is one problem... I've seen employees at certain places use personal accounts for business work. Worse, I've seen them do this despite the business already having seats! I don't know why this is going on, but maybe we should do a bit of peer pressure if we want to ensure we keep those free hobbiest licenses. Plus, it puts your business in a legally dubious area. Not likely to get sued, but it sure isn't legal...
I think it's still a dominant strategy. Docker functions on the same ideology. VMWare switched to the same strategy with their Workstation Pro being free for non-commercial use, etc.
Agreed on the licensing concerns. I hate seeing members of commercial entities discuss using personal/non-commercial products like that. I moved to a new company and wanted to use Obsidian for some individual note-taking purposes, then realized I could only use it for 30 days before I needed a commercial license if I did anything work related with it. I know Obsidian proper wouldn't even know if I used it this way, but I avoided doing so until I got word that Obsidian was changing their commercial licensing policy.
Agreed on the licensing concerns. I hate seeing members of commercial entities discuss using personal/non-commercial products like that. I moved to a new company and wanted to use Obsidian for some individual note-taking purposes, then realized I could only use it for 30 days before I needed a commercial license if I did anything work related with it. I know Obsidian proper wouldn't even know if I used it this way, but I avoided doing so until I got word that Obsidian was changing their commercial licensing policy.
I think that you would've lose those users anyway, so at least leave a good impression on them as they can become advocate when they're in another context where Tailscale is the better solution.
They let it be. The decision is many-fold correct:
1) hobbiests that are obsessed with self-hosting can use it -- tailscale will never see a penny from them anyways
2) some businesses will consider headscale as an escape hatch de-risking if tailscale were to be purchased out by an extortionist (like broadcom's acquisition of VMware)
3) It signals they are doing business in a position of strength: businesses don't buy software, they buy solutions to problems
1) hobbiests that are obsessed with self-hosting can use it -- tailscale will never see a penny from them anyways
2) some businesses will consider headscale as an escape hatch de-risking if tailscale were to be purchased out by an extortionist (like broadcom's acquisition of VMware)
3) It signals they are doing business in a position of strength: businesses don't buy software, they buy solutions to problems
4) Tailscale can benefit from innovations that Headscale finds.
Competition can be beneficial for all parties involved. If the Headscale devs make novel contributions and solve problems Tailscale didn't think of or figure out themselves, Tailscale can copy Headscale just as Headscale copies Tailscale. Competition has this benefit of distributing research expenditures.
I think it is easy to forget that even monopolies are bad for monopolies. They can benefit by being fat and lazy, but this is at the cost of larger future profits. It's "safer", but there's plenty of competitive markets where there is little risk of failure. At least compared to the risk that naturally exists.
5) Headscale users may convert to Tailscale users.
That can happen if Headscale was implemented by a enthusiastic employee who later leaves. Company learned to like the benefits but no longer has the support. This makes a natural conversion. The same thing happens if Headscale fizzles out.
I really just don't see how Tailscale really has much, if anything, to lose from Headscale. It definitely has things to gain! The only reason to go after Headscale would be if they are greedy and irrationally fearful. Which the action itself would scare a lot of its users. Tailscale really is in touch with its users and honestly, I'm not sure there's anything negative I have to say about them. They're not creating enemies, they're paying attention to users, and they're just focusing on making a great product. Seems like the ideal strategy that we'd want all businesses to follow.
Competition can be beneficial for all parties involved. If the Headscale devs make novel contributions and solve problems Tailscale didn't think of or figure out themselves, Tailscale can copy Headscale just as Headscale copies Tailscale. Competition has this benefit of distributing research expenditures.
I think it is easy to forget that even monopolies are bad for monopolies. They can benefit by being fat and lazy, but this is at the cost of larger future profits. It's "safer", but there's plenty of competitive markets where there is little risk of failure. At least compared to the risk that naturally exists.
5) Headscale users may convert to Tailscale users.
That can happen if Headscale was implemented by a enthusiastic employee who later leaves. Company learned to like the benefits but no longer has the support. This makes a natural conversion. The same thing happens if Headscale fizzles out.
I really just don't see how Tailscale really has much, if anything, to lose from Headscale. It definitely has things to gain! The only reason to go after Headscale would be if they are greedy and irrationally fearful. Which the action itself would scare a lot of its users. Tailscale really is in touch with its users and honestly, I'm not sure there's anything negative I have to say about them. They're not creating enemies, they're paying attention to users, and they're just focusing on making a great product. Seems like the ideal strategy that we'd want all businesses to follow.
As I understand it, headscale was implemented by an enthusiast user who became an enthusiast employee, who is allowed to spend work cycles on maintaining it.
My guess is they probably don't care because their pricing is really low. If you want to self-host, for any serious company, you need a very resilient setup (multiple servers) because you can't afford to have your "authz" down. If your headscale is down, so is everything in your organization. $6/month/user is then relatively cheap. The savings on self-hosting will be minimal.
> How did they react to Headscale? I must’ve missed the announcement.
They hired the dev: https://tailscale.com/blog/opensource#the-open-source-coordi...
Fun fact: Tailscale have an almost-assuredly-accurate count of Headscale users due to how few people disable the Tailscale client's telemetry which by default sends them real-time events about everything you do on a Headscale network. See KB1011: https://tailscale.com/kb/1011/log-mesh-traffic
“Each Tailscale agent in your distributed network streams its logs to a central log server (at log.tailscale.io). This includes real-time events for open and close events for every inter-machine connection (TCP or UDP) on your network.”
It's possible to opt out of this spying on Unix/Windows/Mac by starting Tailscale with `--no-logs-no-support` or `TS_NO_LOGS_NO_SUPPORT=true` environment variable (see https://tailscale.com/kb/1011/log-mesh-traffic#opting-out-of...), but it is not currently possible to opt out in the Android/iOS clients: https://github.com/tailscale/tailscale/issues/13174
This is why they had the data telling them they should hire the Headscale dev to nip that competition in the bud before it had any chance to usurp their business by overtaking them in mindshare using their own client. Definitely a sound business decision, but it really really irked me when I realized I was feeding Tailscale detailed usage data for everything I did on my supposedly “““private””” Headscale network despite having no business relationship with them.
For an example of how invasive this is for the average user, this person discovered Tailscale trying to collect ~18000 data points per week about their network usage based on the number of blocked DNS requests for `log.tailscale.com`: https://github.com/tailscale/tailscale/issues/15326
The privacy-adversarial behavior made me switch from Headscale to NetBird for my personal network now that it's available in FreeBSD Ports: https://www.freshports.org/security/netbird/
They hired the dev: https://tailscale.com/blog/opensource#the-open-source-coordi...
Fun fact: Tailscale have an almost-assuredly-accurate count of Headscale users due to how few people disable the Tailscale client's telemetry which by default sends them real-time events about everything you do on a Headscale network. See KB1011: https://tailscale.com/kb/1011/log-mesh-traffic
“Each Tailscale agent in your distributed network streams its logs to a central log server (at log.tailscale.io). This includes real-time events for open and close events for every inter-machine connection (TCP or UDP) on your network.”
It's possible to opt out of this spying on Unix/Windows/Mac by starting Tailscale with `--no-logs-no-support` or `TS_NO_LOGS_NO_SUPPORT=true` environment variable (see https://tailscale.com/kb/1011/log-mesh-traffic#opting-out-of...), but it is not currently possible to opt out in the Android/iOS clients: https://github.com/tailscale/tailscale/issues/13174
This is why they had the data telling them they should hire the Headscale dev to nip that competition in the bud before it had any chance to usurp their business by overtaking them in mindshare using their own client. Definitely a sound business decision, but it really really irked me when I realized I was feeding Tailscale detailed usage data for everything I did on my supposedly “““private””” Headscale network despite having no business relationship with them.
For an example of how invasive this is for the average user, this person discovered Tailscale trying to collect ~18000 data points per week about their network usage based on the number of blocked DNS requests for `log.tailscale.com`: https://github.com/tailscale/tailscale/issues/15326
The privacy-adversarial behavior made me switch from Headscale to NetBird for my personal network now that it's available in FreeBSD Ports: https://www.freshports.org/security/netbird/
> For an example of how invasive this is for the average user, this person discovered Tailscale trying to collect ~18000 data points per week about their network usage based on the number of blocked DNS requests for `log.tailscale.com`: https://github.com/tailscale/tailscale/issues/15326
18000 data points per week seems IMHO pretty low (only 1.7 requests per minute for a whole network? Unlikely, even my Android phone does way more that 1.7 requests per minute just to ad networks, nevermind everything else on the network summed), it's probably way more.
18000 data points is also a lie, that's a different issue - the issue of UDP DNS sucking in general so your average application/OS keeps reissuing requests if they think the UDP packet got dropped.
18000 data points per week seems IMHO pretty low (only 1.7 requests per minute for a whole network? Unlikely, even my Android phone does way more that 1.7 requests per minute just to ad networks, nevermind everything else on the network summed), it's probably way more.
18000 data points is also a lie, that's a different issue - the issue of UDP DNS sucking in general so your average application/OS keeps reissuing requests if they think the UDP packet got dropped.
> only 1.7 requests per minute for a whole network?
Connections != Requests
Telemetry consisting of an open/close event pair would be, like, one entire SSH session, or one RDP session. Even HTTP has supported Keep-Alive since 1997.
Connections != Requests
Telemetry consisting of an open/close event pair would be, like, one entire SSH session, or one RDP session. Even HTTP has supported Keep-Alive since 1997.
> Connections != Requests
DNS lookups != Requests too...
DNS lookups != Requests too...
Headscale creator here.
> Fun fact: Tailscale have an almost-assuredly-accurate count of Headscale users due to how few people disable the Tailscale client's telemetry which by default sends them real-time events about everything you do on a Headscale network. See KB1011: https://tailscale.com/kb/1011/log-mesh-traffic
That is not true. We actually instruct all the clients to send no logs - since at least three years:
https://github.com/juanfont/headscale/blame/main/hscontrol/t....
And also in the config example:
https://github.com/juanfont/headscale/blob/main/config-examp...
That said, I would have loved to know how many nodes use Headscale... :)
> Fun fact: Tailscale have an almost-assuredly-accurate count of Headscale users due to how few people disable the Tailscale client's telemetry which by default sends them real-time events about everything you do on a Headscale network. See KB1011: https://tailscale.com/kb/1011/log-mesh-traffic
That is not true. We actually instruct all the clients to send no logs - since at least three years:
https://github.com/juanfont/headscale/blame/main/hscontrol/t....
And also in the config example:
https://github.com/juanfont/headscale/blob/main/config-examp...
That said, I would have loved to know how many nodes use Headscale... :)
That’s appalling. I really feel like a fool — I had thought the world of Tailscale but no longer.
This needs to be more widely known.
This needs to be more widely known.
I always appreciate that Alpine Linux maintainers disable[1] these things by default.
[1]: https://gitlab.alpinelinux.org/alpine/aports/-/blob/master/c...
[1]: https://gitlab.alpinelinux.org/alpine/aports/-/blob/master/c...
> It's possible to opt out of this spying on Unix/Windows/Mac by starting Tailscale with `--no-logs-no-support` or `TS_NO_LOGS_NO_SUPPORT=true` environment variable
Network flow logs are available to help you understand which devices are connecting to one another over time, that is, the flow of traffic across your tailnet.
These logs strictly do not contain any information about client operations or contents of network traffic.
Network flow logs must be enabled[1].
[0] https://tailscale.com/kb/1011/log-mesh-traffic
[1] https://tailscale.com/kb/1219/network-flow-logs#enable-netwo...
> Am I misreading the docs?
Yes, but don't feel bad, because this seems designed to be confusing and easy to overlook. See the privacy policy: https://tailscale.com/privacy-policy#information-we-collect-...
“When you use the Tailscale Solution, we collect limited metadata regarding your device used to access the Tailscale Solution, such as: the device name; relevant operating system type; host name; IP address; cryptographic public key; user agent (where applicable); language settings; date and time of access to the Tailscale Solution; logs describing connections and containing statistics about data sent to and from other devices (“Inter-Node Traffic Logs”); and version of the Tailscale Solution installed.” (emphasis mine)
Here's a Tailscale client startup with telemetry enabled, which is the silent default:
Yes, but don't feel bad, because this seems designed to be confusing and easy to overlook. See the privacy policy: https://tailscale.com/privacy-policy#information-we-collect-...
“When you use the Tailscale Solution, we collect limited metadata regarding your device used to access the Tailscale Solution, such as: the device name; relevant operating system type; host name; IP address; cryptographic public key; user agent (where applicable); language settings; date and time of access to the Tailscale Solution; logs describing connections and containing statistics about data sent to and from other devices (“Inter-Node Traffic Logs”); and version of the Tailscale Solution installed.” (emphasis mine)
Here's a Tailscale client startup with telemetry enabled, which is the silent default:
2025/04/13 17:02:07 logtail started
2025/04/13 17:02:07 Program starting: v1.82.0, Go 1.24.2: []string{"/usr/local/bin/tailscaled", "-port", "41641", "-tun", "tailscale0", "-statedir", "/var/db/tailscale"}
2025/04/13 17:02:07 LogID: b933623d2f5012930c72af6593b1a94f36749b5495076c937bde51dd620d21e5
2025/04/13 17:02:07 logpolicy: using system state directory "/var/db/tailscale"
2025/04/13 17:02:07 dns: using dns.openresolvManager
2025/04/13 17:02:07 ifconfig destroy: exit status 1
ifconfig: interface tailscale0 does not exist
2025/04/13 17:02:07 wgengine.NewUserspaceEngine(tun "tailscale0") ...
2025/04/13 16:29:20 You have disabled logging. Tailscale will not be able to provide support.
2025/04/13 16:29:20 logtail started
2025/04/13 16:29:20 Program starting: v1.82.0, Go 1.24.2: []string{"/usr/local/bin/tailscaled", "-port", "41641", "-tun", "tailscale0", "-statedir", "/var/db/tailscale", "--no-logs-no-support"}
2025/04/13 16:29:20 LogID: b933623d2f5012930c72af6593b1a94f36749b5495076c937bde51dd620d21e5
2025/04/13 16:29:20 logpolicy: using system state directory "/var/db/tailscale"
2025/04/13 16:29:20 dns: using dns.openresolvManager
2025/04/13 16:29:20 ifconfig destroy: exit status 1
ifconfig: interface tailscale0 does not exist
2025/04/13 16:29:20 wgengine.NewUserspaceEngine(tun "tailscale0") ...Ah thanks!
For anyone wondering, Lammy provided the head of journalctl for tailscaled.service. I got similar results.
I think my confusion comes down to there being two different types of logging. It is the network flow logs that are on the enterprise.
Per the privacy policy statement, this does not set off any alarm bells for me. Reason being that the capacity to create logs would necessitate that language. But I definitely don't like that they are collecting that much information. Thanks, I'll change /etc/defaults/tailscaled
For anyone wondering, Lammy provided the head of journalctl for tailscaled.service. I got similar results.
I think my confusion comes down to there being two different types of logging. It is the network flow logs that are on the enterprise.
Per the privacy policy statement, this does not set off any alarm bells for me. Reason being that the capacity to create logs would necessitate that language. But I definitely don't like that they are collecting that much information. Thanks, I'll change /etc/defaults/tailscaled
Randomly stumbled across this last week when I upgraded my DNS blocklist and saw all the tailscale logs. Definitely unexpected
I believe they also hired the developer of headscale!
Speaking of ACLs and grants, I wish Tailnet Lock[0] extended to those as well, i.e. allowed me to sign my tailnet policy file, so that if the coordination server and one of my nodes ever get compromised, the attacker can't just modify the policy file to access arbitrary services in my tailnet from the compromised node, including – in the worst case – SSH[1].
[0]: https://tailscale.com/kb/1226/tailnet-lock
[1]: https://tailscale.com/kb/1193/tailscale-ssh
[0]: https://tailscale.com/kb/1226/tailnet-lock
[1]: https://tailscale.com/kb/1193/tailscale-ssh
I would love for Tailscale to work on a "paranoid mode". At the moment, you can be sure that Tailscale can't add machines to your network with Tailnet Lock where you have to authorize a new machine from a trusted node and not from the UI (which Tailscale could also do without me clicking it). That's great feature compared to their competitors which basically have a backdoor into your network.
But changing ACLs and many other changes can be done from the WebUI without needed authorization from a host. So I have to authorize the change in the UI - but Tailscale could really also do it without my consent. Would be great to have a mode that _ANY_ change to the Tailscale network must be authorized from a trusted node with a detailled changeset. So I could be sure that Tailscale can't tamper with my network.
Sure, I could run Headscale and completely work without the Tailscale servers. But I prefer to pay them the small monthly fee to not have to manage this central service of my fleet.
But changing ACLs and many other changes can be done from the WebUI without needed authorization from a host. So I have to authorize the change in the UI - but Tailscale could really also do it without my consent. Would be great to have a mode that _ANY_ change to the Tailscale network must be authorized from a trusted node with a detailled changeset. So I could be sure that Tailscale can't tamper with my network.
Sure, I could run Headscale and completely work without the Tailscale servers. But I prefer to pay them the small monthly fee to not have to manage this central service of my fleet.
We were reviewing our SaaS spending recently and I noticed that Tailscale is now our biggest spend. Their approach is apparently working.
Tailscale at least on Linux is a mess of surprise configuration and netfilter rules.
It's not for everyone, but if you're into control of your netfilter tables, it's certainly not for you.
It's not for everyone, but if you're into control of your netfilter tables, it's certainly not for you.
I recently discovered this is how Calico works and it can indeed be really surprising when it goes wrong. At least in a k8s context you have the option of just burning the node and starting again rather than trying to debug it, though.
Vaguely related - there’s still no easy way to get at the email/user name of a user using tailscale ssh right? This is one of the things I really liked about teleport, you could use it to properly attribute git commits on shared machines, without any special setup on the user side.
Post author here!
Are you saying you need an easy way to match up a Tailscale user’s email to a user on the host? (Ie, [email protected] matches to the username sam)
If that’s the case, take a look at the local-part section here: https://tailscale.com/kb/1193/tailscale-ssh#users
If that’s not what you meant, could you clarify a bit further? I’d love to understand how to close this gap for you.
Are you saying you need an easy way to match up a Tailscale user’s email to a user on the host? (Ie, [email protected] matches to the username sam)
If that’s the case, take a look at the local-part section here: https://tailscale.com/kb/1193/tailscale-ssh#users
If that’s not what you meant, could you clarify a bit further? I’d love to understand how to close this gap for you.
No, for disambiguating between multiple users logging into one shared account. Basically, I'd need for tailscale to inject an env variable into the session. Teleport does this with something like SSH_TELEPORT_USER (I don't have access to a teleport instance right now to check the exact name, sorry). It's not really for security purposes (anyone can spoof any env variable they like, anyhow), just for convenience for helper scripts. In robotics it's super common to have just one or two users per machine, and give shared logins to those machines.
While I have you here, it would also be neat if there were a way to supplement regular SSH's auth with tailscale SSH's keys. Specifically for situations where the target machine has lost internet connectivity but you can still get onto the same network as it. More robotics problems. :) Can understand if this would never "work" for tailscale's auth model.
While I have you here, it would also be neat if there were a way to supplement regular SSH's auth with tailscale SSH's keys. Specifically for situations where the target machine has lost internet connectivity but you can still get onto the same network as it. More robotics problems. :) Can understand if this would never "work" for tailscale's auth model.
Yeah this is all great feedback, thanks. Makes sense about the env var!
On the second item- this is interesting. I definitely see the reason for asking :) We actually don’t use SSH keys for Tailscale SSH, since the auth is handled within Tailscale (more info at https://tailscale.com/kb/1193/tailscale-ssh#authentication-a...). I’ll think more about this. Do you have a break-glass setup with a normal SSH key for this today?
On the second item- this is interesting. I definitely see the reason for asking :) We actually don’t use SSH keys for Tailscale SSH, since the auth is handled within Tailscale (more info at https://tailscale.com/kb/1193/tailscale-ssh#authentication-a...). I’ll think more about this. Do you have a break-glass setup with a normal SSH key for this today?
We don't use Tailscale SSH, we do our own management, which I'd like to move away from.
There seems to be a missing comma in the first grant example. If I am not beffudled.
At the end of the dst line.
In fact, comma usage seems inconsistent for the last element of some of the nested objects.
To be fair, I don't remember what's the json rule either and would need a quick lookup.
In fact, comma usage seems inconsistent for the last element of some of the nested objects.
To be fair, I don't remember what's the json rule either and would need a quick lookup.
They use hujson: https://github.com/tailscale/hujson
But it doesn't accept things like the example that omits an intermediate comma:
But it doesn't accept things like the example that omits an intermediate comma:
{
"foo": 42
"bar": 123
}
> hujson: line 3, column 5: invalid character '"' after object value (expecting ',' or '}')They should've gone with Relaxed JSON, which is more relaxed than JSON5, which is more relaxed than HuJSON/JWCC.
Yeah, the example that's missing the commas is wrong, if I'm not mistaken about their json parser. The optional comma at the end is fine with JSON5, which I think they're using, as that allows optional commas, comments, things like that.
Okay but when can we be connected to more than one tailnet at once and have tail scale auto reconnect directly after resume from sleep? That's the quality of life stuff users really care about
You talking about macOS I presume, as I’ve noticed this as well. After every sleep I have to manually toggle tailscale, yet it happily reports being online before I do this…
When would I want this over a more generic solution like linking to an idp using something like keycloak? The upside is that its network level blocking, but it still requires modifying the application itself same as connecting an idp, correct? The downside of this approach is that it only works for internal access where your end users are on the tailnet
> When would I want this over a more generic solution like linking to an idp using something like keycloak
I assume everyone using Tailscale are opting for "simple code and adminstration" instead of "let me pierce NAT and handle Wireguard tunelling myself". On that point of view, "Identity management" is part of "stuff" that Tailscale helps you in the first place. So why not use it as the "Identity management" solution as well?
When you use `tailscale serve`, you get the headers Tailscale-User-Login, Tailscale-User-Name and Tailscale-User-Profile-Pic that identifies who connected thru the tailscale endpoint. https://tailscale.com/kb/1312/serve#identity-headers
On a separete side note, if you had an exposed OpenID Connect service in the first place, you can use it as the Tailscale tailnet auth SSO: https://tailscale.com/kb/1240/sso-custom-oidc
I assume everyone using Tailscale are opting for "simple code and adminstration" instead of "let me pierce NAT and handle Wireguard tunelling myself". On that point of view, "Identity management" is part of "stuff" that Tailscale helps you in the first place. So why not use it as the "Identity management" solution as well?
When you use `tailscale serve`, you get the headers Tailscale-User-Login, Tailscale-User-Name and Tailscale-User-Profile-Pic that identifies who connected thru the tailscale endpoint. https://tailscale.com/kb/1312/serve#identity-headers
On a separete side note, if you had an exposed OpenID Connect service in the first place, you can use it as the Tailscale tailnet auth SSO: https://tailscale.com/kb/1240/sso-custom-oidc
Doesn‘t tailscale force you to bring your own SSO / IdP already?
So it‘s not part of what Tailscale brings, it just adds another layer of indirection between the SSO / IdP you have already and the app, plus requires some custom library integration work, further enhancing lock-in.
So it‘s not part of what Tailscale brings, it just adds another layer of indirection between the SSO / IdP you have already and the app, plus requires some custom library integration work, further enhancing lock-in.
If you write internal apps, this is a lot easier to support.
You don't have to think about redirects, sessions, cookies, syncing your users and their roles in TS and Keycloak, maintaining two separate policy files and all that.
You don't have to think about redirects, sessions, cookies, syncing your users and their roles in TS and Keycloak, maintaining two separate policy files and all that.
The tradeoff you get is lock in to tailscale and no portability. Though I assume that is true about some of the networking features tailscale offers as well.
It’s at the network level. It will block clients from connecting to the destinations on the network. You don’t need to modify any applications at the destination.
Am I misinterpreting like, the entire body of the article?
> For private services inside your tailnet, like internal tools, implementing auth on every app is overkill. That’s why we developed tsnet, a Go library that embeds Tailscale directly in your applications. This lets you see and react to the identity of every user who makes a request to your app. So, authentication is handled…what about authorization?
The article then proceeds to show how to modify an application using tsnet to embed in the application. Which sounds a whole lot like modifying the application. The point might be that it’s more straightforward to do this then gate behind something like keycloak, but fundamentally it still requires modifying the application
> For private services inside your tailnet, like internal tools, implementing auth on every app is overkill. That’s why we developed tsnet, a Go library that embeds Tailscale directly in your applications. This lets you see and react to the identity of every user who makes a request to your app. So, authentication is handled…what about authorization?
The article then proceeds to show how to modify an application using tsnet to embed in the application. Which sounds a whole lot like modifying the application. The point might be that it’s more straightforward to do this then gate behind something like keycloak, but fundamentally it still requires modifying the application
Ah yes. This seems to be something new and entirely optional. They’re saying that it might make sense to some to keep both network and application level authorization defined in one place. Personally, I’m with you in just using a normal SSO solution.
Will there eventually be a UI to manage grants so we don't have to write them by hand?
Yes.
Nice business decision. Just like their reaction to Headscale, I'm reminded Tailscale 'gets it' when interacting with the dev community and why their popularity will continue to grow.