Blog

Tailscale is split into a control plane and a data plane. The data plane is built out of direct WireGuard links that provides end-to-end encryption between any two machines on the network. The control plane is responsible for verifying the identity of users, validating machine keys, and delivering the public keys of peers to each machine in the network. This document focuses on the management of keys in the control plane. For a broader overview of Tailscale, see “How Tailscale Works.”

Lately, I get people asking me when microservices are a good idea. In systems design explains the world, I talked about big-picture issues like second system effect, innovator’s dilemmas, and more. Can systems design answer the microservices question?

Yes, but you might not like the answers. First, we'll need some history.

If you’re like most people, your answer to this is… “What? Why?”

When ssh was introduced back in the 1990s, its appeal was simple. Passwords are too short, too guessable, too phishable, too often stored incorrectly, too MITM-able, too brute-forceable. Also its primary competition was rsh’s classic “no authentication,” but we don’t talk about that.

The team has been hard at work making Tailscale more Tailscale-y. Today we’re announcing v1.2 is stable and ready for teams and hobbyists alike. Most notably, this release includes MagicDNS for everyone and major improvements for our Windows client.

How to update:

• Linux: update instructions (apt update, install, etc.)
• Windows: update instructions
• macOS: update via the Mac App Store*
• iOS: update via the App Store*
• Android: update via the Play Store

*For macOS and iOS, you may need to quit Tailscale first; the App Store doesn’t seem to update running VPN apps.

We covered a lot of ground in our post about How Tailscale Works. However, we glossed over how we can get through NATs (Network Address Translators) and connect your devices directly to each other, no matter what’s standing between them. Let’s talk about that now!

We’re once again happy to announce a new version of Tailscale.

What comes after 0.99? 0.100, of course!

This is a pretty notable release, containing a major rewrite of our “magicsock” connection code that sits between WireGuard and the network, finding the best path between peers and getting through NATs.

If you’ve had any connection woes previously, definitely give this a try.

One catch, though: the new 0.100 connectivity code only kicks in if two peers trying to connect to each other are both running 0.100 or later. So make sure you update all your devices.

How to update:

• Linux: see https://pkgs.tailscale.com/stable/ (apt-get update, upgrade, etc)
• Windows: from that same page, download tailscale-ipn-setup-0.100.0-107.exe
• macOS: update from the Mac App Store (you’ll likely need to stop Tailscale first; the App Store doesn’t seem to update VPN apps that are running)
• iOS: we’re giving it a few days until we mark 0.100 as our stable build on iOS, but you can join our TestFlight beta program to get it today
• Android: the latest Tailscale Android beta builds use the new 0.100 connectivity code

In addition to the connectivity improvements, there are a number of other fixes and cleanups:

• The Linux client now respects DNS settings set in the Tailscale admin console.
• The Windows client now has “About” and “Exit” menu options. The “About” dialog will show the current stable version. (No auto-update option yet, but it’s a start.) Windows service start-up errors are now also surfaced in the UI, which is still a sad experience if it happens but should make for better Windows bug reports at least. We’re working on those. Long tail is long.
• The macOS client now stays off when you turn it off via the OS network settings.
• The tailscale status subcommand (only currently included on Linux) now consistently shows asterisks around a peer endpoint address only when that path is active, and also now shows asterisks around DERP relays if that’s what’s being used.

Enjoy!

And as always, email us or tweet us (@tailscale) if you have any problems and we’ll try to help.

At the beginning of May we welcomed our first ever batch of interns to the Tailscale team! They’ve all been hard at work the past few weeks, and we want to formally introduce them.

Joining us from the University of Waterloo are Zijie, Wendi, and Dmytro:

Zijie Lu (@lzjluzijie) is a Mathematics student at Waterloo. Originally from Beijing, Zijie has experience writing Go, React, and Vue, and is most known for his websocks project, a secure WebSocket-based HTTP proxy. (As soon as we saw that project, we knew he’d be a great fit.)

Zijie has never used, let alone owned an Apple device, instead preferring to run a dual-boot Fedora / Windows machine. In his spare time, he plays DOTA2, and is currently ranked in the top 2000 players in the Americas!

This term, Zijie will improving our network admin console, to make managing devices and auth settings easier for teams.

Wendi Yu (@wendi-yu) is studying Software Engineering. She’s a member of Waterloo’s rocketry design team, building tools to model tank fill and P&ID systems for rocket launches. (Tailscale’s own rocketry fans are excited to have another member join.) And if that wasn’t enough, she’s also a sousaphonist for Waterloo’s concert band. As she puts it, “There’s something immensely liberating about being able to honk back at the geese who attack me when I walk through campus.”

Currently based in Edmonton, Wendi is working on real-time auditing and visualization of networks to help teams secure and monitor their devices.

Dmytro Shynkevych (@dshynkev) is pursuing his Computer Science degree, and has already completed internships at SideFX, Cognite, and Kik Interactive, working on machine learning and 3D rendering projects.

Originally from Ukraine, in his spare time, Dmytro pseudonymously translates online content, mostly songs:

Doing so well, which is to say, localizing instead of merely translating, and preserving the rhythm (if not the rhyme) of poems and songs is a fun challenge!

He’s also excited about cybersecurity, and regularly participates in CTF competitions on weekends. He’s particularly proud of his 2018 team’s work in the CSAW Quals: “we were motivated, efficient, and worked in perfect synchrony.”

While at Tailscale, Dmytro is developing our MagicDNS feature, letting teams access network devices with memorable names, in addition to Tailscale IP addresses.

We’re happy to have Wendi, Zijie, and Dmytro with us! You’ll likely see their contributions on our public repositories over the next few months.

We’re happy to announce a new version of Tailscale. This minor release fixes various connectivity issues and squashes some annoying platform-specific bugs. Thanks to everyone who wrote in to report these issues!

A few highlights from the complete changelog on GitHub:

• We now prefer IPv6 over IPv4 when sending encrypted packets between nodes. Note: this does not yet make IPv6 available inside the Tailscale network.
• Switching between different networks is now smoother than ever, particularly between Wi-Fi and LTE, or when moving a sleeping laptop between different networks.
• Windows no longer resets active connections when new nodes get added to the network.
• We’ve adjusted MTU settings to avoid packet loss for users on Google Cloud or DSL.

This release only contains connectivity and stability fixes, so we recommend everyone update to the latest version. You can see out-of-date nodes on the machines page of your admin console.

You can find update instructions for your platform.

An article by Michael Tremer titled Why not WireGuard is sometimes shared in VPN discussions. Unfortunately that article contains several misconceptions and some out-of-date information that deserves to be addressed.

Let’s go through his arguments section by section:

Just over a year ago, we founded Tailscale with a common sense of nostalgia for the “good old days” of LANs. In our collective opinion (then and now) networking and cloud infrastructure has become too complicated. Attempts to increase team connectivity and migrate towards remote work results in a corresponding burden of security. This reduces productivity. Systems and approaches don’t scale without significant time and effort. Everyone wins.

That’s why we are happy to announce that we’ve raised a $3M seed round, led by Heavybit with participation from Uncork Capital and others. This investment sets the expectation on what we’re aiming to achieve: a return to simple computer networking for everyone that works anywhere you can access the Internet.

People often ask us for an overview of how Tailscale works. We’ve been putting off answering that, because we kept changing it! But now things have started to settle down.

Let’s go through the entire Tailscale system from bottom to top, the same way we built it (but skipping some zigzags we took along the way). With this information, you should be able to build your own Tailscale replacement… except you don’t have to, since our node software is open source and we have a flexible free plan.

We have some catching up to do. Tailscale opened our waitlist for signups in April 2019, almost a year ago, but we haven’t shared much news! It’s time to rectify that.

Over the past 11 months we’ve grown the team and narrowed our focus to just one core product: a company-wide mesh overlay network based on the WireGuard® VPN.

As a “fully remote work” company, we had to make some choices about the technologies we use to work together and stay in touch.

We decided early on — about the time we realized all three cofounders live in different cities — that we were going to go all-in on remote work, at least for engineering, which for now is almost all our work. As several people have pointed out before, fully remote is generally more stable than partly remote. In a partially remote team, the remote workers seem to always end up treated as an underclass, overlooked in meetings, bypassed for promotions, fired when they eventually refuse to relocate because the remote work policy inevitably changes (hi, Yahoo!), etc.

The good news with our plan is the founders could “dogfood” a few different remote work ideas ourselves before we ever hired anyone. So we decided to try some stuff. Here’s what we discovered.

Some news — we have deb and rpm package repositories up!

pkgs.tailscale.com

Currently serving unstable-track packages for tailscaled, a replacement for our current linux relaynode.

If you’re brave, give it a try! Stable release with docs coming soon.

We just made the first bits of the Tailscale code public, starting with the Linux client and its dependent/common code.

https://github.com/tailscale/tailscale

Still lots of rough edges & TODOs everywhere so temper expectations accordingly. We want to hack in open and not wait until it’s perfect.

I used to tolerate and expect complexity. Working on Go the past 10 years has changed my perspective, though. I now value simplicity above almost all else and tolerate complexity only when it’s well isolated, well documented, well tested, and necessary to make things simpler overall at other layers for most people. For example, the Go runtime is relatively complex internally but it permits simple APIs and programming models for users who then don’t need to worry about memory management, thread management, blocking, the color of their functions, etc. A small number of people need to understand the runtime’s complexity, but millions of people can read & write simple Go code as a result. More importantly, Go users then have that much more complexity budget to work with to build their actual application. I would’ve never built Perkeep had I needed to fight both its internal complexity and the complexity imposed on me by other contender languages/environments at the time.

Long ago, I wrote git-subtree to work around some of my annoyances with git submodules. I’ve learned a lot since then, and the development ecosystem has improved a lot (shell scripts are no longer the best way to manipulate git repos? Whoa!). Thus, I bring you: git-subtrac. It’s a bit like git-subtree, except it uses real git submodules. The difference from plain submodules is that, like git-subtree, it encourages you to put all the contents from all your submodules into your superproject repo, rather than scattering it around across multiple repositories (which might be owned by multiple people, randomly disappear or get rebased, etc). As a result, it’s easy to push, pull, fork, merge, and rebase your entire project no matter how many submodules you like to use. When someone does a ‘fetch’ of your repo, they get all the submodule repos as well. I wrote a longer git-subtrac README describing how to use it and its internal workings.

Identity on the internet is messy. The result is some things that should be easy are hard.

I am leery of jargon. I am as guilty of using it as the next engineer, but there comes a point where there are just too many precise, narrowly-understood terms polluting your vocabulary. The circle of people you can talk to shrinks until going to the store to buy milk feels like an exercise in speaking a foreign language you took one intro course to in college. Less jargon is better.

Thus the first few times I heard the terms zero trust network and microsegments I ignored them. The conversation went on even though I was a bit confused. Eventually I heard these enough that I had to figure out what these words mean. Turns out they are useful!

So what are they?