with Thomas Ptacek
Security - a devtools discussion with Thomas Ptacek (Fly.io). Episode 2 (Season 2) of the Console Devtools Podcast.
In this episode we speak to Thomas Ptacek, currently a software engineer at Fly.io and previously a co-founder at security firms Latacora and Matasano Security. We discuss the state of software security in sectors like energy and healthcare, how software developers should think about supply chain risk, and what they should do about securing their dependencies. We also explore how security threats have changed over the years, and what developers working on open source should do to improve their own security.
- Google cloud authentication
- Server-side request forgery
- Burp Suite
- Black Hat
- Emacs Tramp Mode
About Thomas Ptacek
Thomas Ptacek is a leading security researcher. Best known as one of the co-founders of Matasano Security, which was prior to its acquisition by NCC Group one of the largest software security firms in the US. Working in software security since 1995, Thomas was a member of the industry’s first commercial vulnerability research lab - Secure Networks. Thomas is currently a software engineer at Fly.io
David: Welcome to the Console Podcast. I'm David Mytton, co-founder of Console.dev, a free weekly newsletter highlighting the best and most interesting tools for developers. In this episode, I speak with Thomas Ptacek , currently a software engineer at Fly.io and previously a co-founder at security firms Latacora and Matasano Security. We discuss the state of software security in sectors like energy and healthcare. How software developers should think about supply chain risk and securing their dependencies. How security threats have changed over the years. And what developers working on open source should do to improve their own security. We're keeping this to 30 minutes, so let's get started. I'm here with Thomas Ptacek: . Thomas, thanks for joining the Console Podcast.
Thomas: Thanks for having me on the Console Podcast.
David: Let's start with a brief background. Tell us a little bit about what you're currently doing and how you got here.
Thomas: I'm an engineer. I'm a developer at a company called Fly.io. We're a fun way of running applications in the cloud, close to users. I can talk more about what Fly does. I get the sense that when people want to hear from me, they're generally interested in security stuff. I've been doing software security since the mid 1990's. So, as a vulnerability researcher and as a security consultant and doing lots of security work for startups as well. So, my background is kind of heavily in finding vulnerabilities and working with especially small tech companies to kind of navigate that whole space and figure out what to do about that whole landscape of things. Everything's being broken everywhere and all software being garbage. That's kind of my portfolio.
**David:**So, when we think of tech companies, we generally think of the big names like Apple, Google, Microsoft, and Amazon. These are companies that build software as part of their core business. And even though they invest billions of dollars into security every year, there are regular major security vulnerabilities in their products. I wanted to start a little bit outside of this and talking about companies that use technology and software, but to help them deliver on some of their core business like energy, transports, or healthcares. Sectors where they might not have billions of dollars to spend on security. What's your take on the current state of security in these types of industries?
Thomas: I guess it's a bit of good news and a bit of bad news. Right? The bad news is kind of the same bad news no matter who we're talking about, which is, again, computers are garbage. Especially software. All software is trash. So, the basic problem that everyone faces is no matter kind of what industry you're working in, even if it's a heavily regulated industry, even if it's kind of inconceivable that you could have terrible vulnerabilities and the sky would not be falling or whatever, they're all built on the same basic couple of software stacks. And those stacks were all generally kind of conceived of in the 1990's and early 2000's. And they're riddled with security vulnerabilities. We've barely kind of come to grips with how to build software with even a modicum of just basic security promises. Like if I can talk to it, I can't run remote code on it. Right?
Thomas: That's a problem we've barely figured out how to deal with. The bad news is that if you're working on infrastructure or a regulated business, or especially in places like energy or utilities or things like that, you're facing a lot of the same problems, basically all of the same problems, that everyone else does. But the kind of the subtext I get there is those are places where technology is kind of a cost center. It's not the core thing that people are building. It's something that people are taking advantage of to reduce costs and things like that. And what that tends to mean is that those companies, don't staff. Less so security people and more the kind of developer that keeps kind of abreast of changes in how we're building software to make it more resilient. So, in my previous life as a security consultant, I did kind of a fair bit of work with large regulated companies, financial exchanges, energy companies, and things like that.
And they're not less exposed to vulnerabilities than smaller tech companies. The good news kind of is those companies tend to have more resources than a typical startup does. And so there's a whole cottage industry of delivering kind of consulting and security work, especially for people in the energy and the transport sectors. There are whole well known security practices that just exist to do security audits of trains and things like that. I don't know. Is there a particular kind company in that space that you're thinking of? What's the model of the small company that doesn't have ten million dollars to spend on security consulting this year?
David: Right. So, I suppose we're hearing all the time about state sponsored cybersecurity on critical infrastructure and that often centers around the idea of shutting down the electricity grid is the big catastrophe that's potentially out there. Do you think that is the right thing to be focusing on?
Thomas: I'll say two things. Right? The second thing I'll say is something that everyone's heard already before, and I don't have any real unique insight to offer for it. The little bit of insight I can say here is that when we think about things like shutting the electricity grid down, that's not an overblown theory of what could happen. Right? There's good reason to believe that attacks like that are totally viable. They're probably more viable than we think they are. They're probably more viable today than they were 10 years ago. So, like 10 years ago, when we were thinking about shutting electricity grids down, we might have been thinking about somehow infiltrating the network of a power grid operator and then blowing up a transformer by getting a remote code execution on a controller somewhere. Or hitting directly a bunch of smart meters and doing a remote shutoff on every smart meter in the area.
Those attacks were probably viable. They were certainly viable 10 years ago. Maybe they're less viable now, but what we see now is back office attacks. Right? I sound like a salesperson for a security product company. And I promise I'm not a nihilist or a salesperson on any of this stuff. I'm really jaded about all of it. But also I think it's becoming clear to us that we can shut down companies like private companies that are offering these services just by shutting their back offices down. Just taking down the computers that administrative assistants are using to do data entry and things like that. If you shut people's business operations down, you can shut those things down as well. So, part of me thinks that we're not paying enough attention to it. And then the obvious thing to say here, though, is that there's a reason this doesn't happen all the time and it's not because the attacks are prohibitively hard to do. It's because if they're state sponsored attacks and you shut down all of public transportation in an area, or you shut down the electric grid or something like that, the US response to that, or the UK response to that, won't be a retaliatory cyber attack.
Right? It'll involve missiles. So, yeah, in the same sense that I think we're all kind of at peace with the fact that you could also shut down the power grid by planting bombs in a couple of strategic places or something like that. Right? We live with that risk. And we just assume that if it's going to happen, it's a prelude to a much bigger conflict. It's kind of out of our hands. So, it's worth paying attention to. I'm thankful that I don't work in kind of a corporate IT space where I'm kind of held hostage to the software that other people build. I've spent my career mostly working, at least when I'm building my own stuff, working in spaces where we're responsible for the software that we build and the whole business is paying attention to how we build that software. It's much harder if you're like a corporate IT person or an IT security person. And the best you can do is get a consultant to kick the tires on something before you buy it and operationalize it.
David: Right. So, it's more likely that it's going to be a ransomware attack that's led into someone who's non-technical on the network rather than some highly elaborate zero day and some control plane somewhere.
Thomas: I think both are possible, but I think the ransomware style attack is more straightforward. It's more lucrative. I think it's more likely to be the way we see it going down.
David: But how should the broader tech industry think about this systemic risk, for example, will running our code in data centers that require electricity?
Thomas: Yeah. Your guess is as good as mine. If the question is about whether we should use memory safe programming languages, I'm right there with an answer. If the question is what we do about our enormous dependence and our interdependence on all these different actors doing different levels of security assurance for what they're delivering, I have no no freaking idea. I'm totally lost. Someday the power is going to go down and I'm just going to live with it until it comes back on.
David: Well, that's actually a good link into discussing open source and our dependencies on a huge supply chain and all of those libraries that we're all importing to our code that we can, in theory, look at the source code for but very few people are doing. How should software developers think about that supply chain risk with all the dependencies and the dependencies, dependencies, and so on?
Thomas: It's a real problem. So, from, I think, around 2016 to 2020, I was a principle at a company called Lata Quora. And that company still exists, I'm just at Fly now. We'd set up security teams for startups that were mostly building on top of typical open source software stacks. Jango or a lot of Node.js, A lot of NPM ecosystem stuff. Right? And we went in thinking that it was going to be kind of straightforward for us to just vet packages and follow dependency chains and keep up with security alerts. And it's a really difficult problem. In particular, I think the test case for this is kind of Node.js where literally every possible piece of software has been factored out of every other possible piece of software and into its own package. And those packages have a weird web of dependencies and maintainers, and I think it's basically hopeless to keep up with are changes being introduced maliciously?
Thomas: My odds of detecting that are really minimal. Right? There's so much code in an NPM and a node setup. There's so much code that you're pulling in through dependencies and in each of those dependencies you're using so little of it and keeping a good map of which parts of which packages you're working with is very difficult. I think the best advice I can kind of give there is to be conscientious. Again, I think this is kind of straightforward. I think people probably get this already, but to be conscientious about what depths you're pulling in. I think if you're serious about being resilient against supply chain attacks, for instance, if that's a real tenant of how you're running your engineering team, you're going to make sacrifices in order to kind of live up to that tenant. Right?
So, you might pick different dependencies. You might choose to build on different and less shiny possible stacks just to avoid your exposure to left pad or whatever the thing is now. I think it's worth being cognizant of the fact that when you opt into kind of shiny new things, you're also opting into all the dependencies that those things have. For a while, I'm sure it's still the case. GitHub will give you alerts for all of the vulnerabilities detected in a node package or really any package. Right? And those alerts are pretty close to worthless because every single package you use has a dependency somewhere that has some random vulnerabilities that's being detected every day. But the annoying thing about it is that most of those vulnerabilities aren't relevant to what you're building. And so you're getting a constant stream of alerts, but those alerts aren't operationalizable. If it was the case that when you got an alert from GitHub about something in your software stack, and it was a real vulnerability, you'd be very attentive to that. Right?
You'd be quick to respond to it. And it'd be really easy to justify resourcing. Constantly triaging and updating packages and things like that. But the reality is that most of what's actually discovered in these packages are low tests, low grade vulnerabilities. And they're also in corner cases for that software. Right? It's difficult. For me, I would mostly opt into more mainstream packages. So, for instance, I'd probably build on React and I'd probably keep it to vanilla React as much as I could. Things like that. Right? Staying within the golden path of the ecosystem. I really like Deno is doing here. Deno is that Rust rewrite of Node that has its own kind of package ecosystem. They're kind of doing things right there. I guess I'm contradicting myself as Deno is pretty shiny as well, but I like where Deno is going with this.
I don't know that I have a clean answer for what to do about the supply chain problem with things like Node.js. If you're not in Node.js, I'm not sure what your excuse is. Right? If you're running lots of random stuff. A bit of advice I've given and people always look at me weirdly when I give this advice is I like the idea of being prepared to rebuild your own dependencies from scratch. At least the big ones. You might remember, I think it was back in 2012, I'm trying to remember the dates for Heartbleed. Right around the time of the Heartbleed vulnerability in open SSL, the consulting company I was at at the time, Monsanto Security, we were on an engagement and we found almost identical vulnerability in nginx. So, it wasn't in open SSL. It was in the web server itself.
You could send a request to nginx and if it had a null character somewhere and a header, I'm remembering this weirdly, but you could send a malformed header and the response would come back with random crud pulled out of the memory of the nginx process, which is the exact same pattern of the vulnerability in Heartbleed. And it worked that way. If you just repeatedly did that, you'd gradually be able to gravel through the memory of the server process and you'd get private keys and things like that. So, it's a major, horrible vulnerability. Right? Was really trivial to find that vulnerability in the code. So, everyone running nginx basically had this vulnerability. It was a real problem. But people were waiting for their package managers to release updates that would have that fix in it. And meanwhile, we're kind of standing there looking at the one line of code that you need to change to fix the vulnerability.
It was really clear. And if you looked at the patch for it, it's a really benign patch. Right? You can see how it's not going to blow things up. I get how if you're a small startup, it maybe makes sense. You're not a huge target. It maybe makes sense to wait for the official patch for things. But if you're a relatively large company or if a lot of people are depending on your application or if security's a big deal for what you're doing, it always made sense to me that you should be able to, if push came to shove, apply a patch to nginx and rebuild it yourself. It always seemed to me you should be able to make an inventory of the major dependencies that you have in your environment. This is not as doable in Node as it is outside of Node.
But for things you run on servers for databases, maybe even if you're in Jengo or Ruby where dependencies are a little bit more overt, it should be possible to kind of make an inventory of the major dependencies you have and then vendor them and be prepared to apply patches to them. Also forces you to kind of update those patches deliberately. Nobody does this. Right? I think it's a realistic thing to do, but I think people flinch at the idea that they would ever themselves build a version of a package they run. It's just a difference between the way software was built when I was starting in the industry and the way it is now, where it's all just kind of shrink wrapped and pin compatible, plug and play, take the gem file or the MPM package and you're off to the races.
So, I think if you're serious about dealing with these problems, if you're serious about addressing those things head on, those are some of the sacrifices that you make is you'll spend engineering time to vendor out them major dependencies you have so if you had to patch them, you wouldn't have to wait for three days for a package update. The flip side of all this though is for most companies, it doesn't make a whole lot of sense. For most small companies. It doesn't make a whole lot of sense to put a huge investment in security, just from a business perspective. Security is unlikely to be the thing that blows up your business. Product market fit is the thing that's going to blow up your business. Not building something that people want or want to use. Right? That's the real problem. And so there's a fuzzy space between when you start the company with two engineers and the business person or whatever that founding team looks like, and when you get to around like 50 employees.
That's the place where most people start thinking about adding full-time security. Where you're consciously making trade offs. You're trying to get to market and find product market fit. And that's the thing that's most important for your business. I'm not here to tell companies that they shouldn't do that. I think that makes a lot of sense. I think there are exceptions, like if you're building things where people's health or safety are on the line, obviously that's out the window. But if you're building a cat sharing app or something like that, it's just as likely that the way that security will burn you is that you'll pay too much attention to it. You'll do a lot of Kabuki theater trying to do kind of the cargo cult thing that you think you're supposed to be doing, because this is what big enterprises are doing, keep things secure.
And the reality is that when you grow from five people to 50 people and you bring in a security team full-time or whatever, they're just going to rip all that up anyways. It's going to be the wrong stuff that you're doing to begin with. I think there are some basic things that people should get right to begin with that are worth investing in, but that a lot of it is not worth spending a lot of effort on. I don't know if that sounded incoherent or not, but that's a shotgun blast of my answer there.
David: That all makes sense. So, what are the things that those companies in between one or two people and 50, what should they focus on first?
Thomas: As people are probably aware, there is a huge industry of products that are intended to increase security for networks or for software. There are things that kind of will promise to watch your application to detect attacks or block attacks or scan your code for vulnerabilities and things like that. I'm not a big believer in any of those things. I think, in general, your first reaction to any security product should be it's not for us. It's not going to work here. It's intended for big companies with big IT staffs that might get some 5% value from deploying it, but that's good enough for them. And it's not going to work for you at all. Security products in general, I'm very down on. With, I think, two exceptions. Right? One thing I think people could get right from the beginning that's worth investing in and it's kind of a minimal investment, I would, as soon as I possibly could, if I was starting a new tech business or new project or whatever, I would get that project onto an FSO IDP like Okta or Google Cloud auth, something like that, right away.
We spent some time interviewing chief security officers for large tech companies that had been small tech companies recently. And every single one of them said one of their first five things that they would do if they were taking over as the chief security officer for a new big tech startup or a new little tech startup is get Okta deployed. All of the applications that you're accessing, all the software as a service things and all your internal tools as well, you're logging in through Okta. You have a single central place where you're managing who's allowed to access things and you can enforce two factor auth and things like that.
So, a security product I would invest in, something like Okta or Google Cloud auth. The easier thing to do there is probably just Google cloud auth. Right? Once you do that, everyone logs in with their Google account and you can enforce when they do log in, they have to use a two factor auth key, a Fido key, something like that. Right? So, that works great. You should definitely do that. Another thing I really like is Tailscale. I'm generally a big fan of a VPN protocol called WireGuard. WireGuard is kind of a modern replacement for VPN protocols like IPSec and OpenVPN, which are both terrible. Just piles of software garbage. And WireGuard was kind of consciously designed to avoid all the security missteps that previous VPN protocols had made.
WireGuard is part of Linux. It's upstreamed into the Linux kernel. If you install a new distribution of Linux, you get WireGuard for free. Tailscale is a company that has kind of taken WireGuard and made it kind of work magically well on every mainstream operating system. On your phone, on your MacOS computer. Works great on Linux. I love Tailscale. I can't say enough good things about it. So, getting all the things that you're running internally off the public web and onto a Tailscale network also makes a lot of sense to me, especially because most people inevitably end up setting up some kind of access VPN at some point. And it's always a terrible project that no one wants to work on. And the VPN has always set up pretty badly. Tailscale just does all that work for you. It integrates with your existing SSO IDP. It's fantastic. I love it. I am not paid by Tailscale to say this.
So, two things I really believe in are centralizing access control to the extent that you can also centralizing how you're managing your network. Probably a lot of people that hear this have already heard of Tailscale, And all I'm really saying is yeah, Tailscale is really great. You should look at Tailscale. But things like security scanners, security scanners generate a lot of noise. And if you're taking them seriously, if you're doing anything with them, they're generating a lot of work. And it's hard to say that that work is really like they're not finding necessarily the most important vulnerabilities. They're just finding the low hanging fruit in your code that may or may not even be exploitable. The market for those products, rewards companies and scanner products, that flag more things.
And most of the things they flag are not the things that you should be paying attention to. I've never seen a security scanner that did a good job with anything. Detection I think is out the window. There's a weird thing with security attack detection products where the trick here is that if you start watching for attacks on your application or on your network, what you quickly learn is that you're constantly under attack. Everything on the internet is being attacked 100% of the time. So, we run kind of an application hosting company here at Fly and so a thing that people are doing is they're booting up applications and they're coming up on our Anycast network and you can watch the internet start to talk to these applications. And the first thing that happens with any application that we boot up, even if I just boot up a little test application that no one's heard of before, the first things I see in my log are just random internet scanners hitting them and looking for dumb PHP vulnerabilities from 2005 or whatever. Right?
The thing about that is it's easy for detection companies to demonstrate value because if you set something up there to listen for attacks, it's going to find attacks. You're constantly being attacked. But almost all of those attacks fail. The real figure of merit is is it detecting the attacks that are likely to succeed and there's nothing good that does a good job with that. I'd probably think more about before I spent any money on any product that was going to watch for attacks or things like that, or scam my code,. I think about setting aside money to have somebody come in and do an assessment of my code and find vulnerabilities. I'm not a security consultant anymore. So, I'm not talking my book when I say that. That's, I think, a more valuable thing to think about doing than deploying products. I'm really jaded about products, but I do really think that people tend to put off getting single sign on set up and the earlier you do that, the easier it is to scale it across your company. You really want that set up early. So, I'm a big fan of doing things like that.
David: Flipping it over to developers working on open source, then on their own projects, what can they do to improve the security of those projects for their users?
Thomas: Well, I don't want to be too patronizing, but the first thing I'd say is work in a memory safe language. So, I think there are big classes of vulnerabilities that you can get around just by working in modern languages that don't have these problems. In my head right now, I'm trying to think of what the unique risks of being on an open source project are. And I think they're kind of obvious. Right? I think that people that are running major open source projects are already doing PRs. They're already watching what code they're emerging in. It's not like it's really super easy to trick people into adding malicious code to your own project or things like that. Right? And for the most part, when people are building open source projects, they're not so much running their own infrastructure, which is a nice thing about being an open source project, is that you're mostly thinking about the code and not thinking about the cloud software as a service provider that runs that code.
So, the problems of being the person running that code are trickier than the problems of being in the open source project itself. But most of my thoughts here for software security are about vulnerabilities. Are about paying attention to the right vulnerabilities. About foreclosing the possibilities of other vulnerabilities. So, for example, I think that some of the most important vulnerabilities are underappreciated outside of security. I think that people don't pay enough attention to, for instance, what's called SRF or a server side request forgery, which is just the idea that if you're building a product, if you're building a piece of software, that makes its own HDP requests on behalf of users, if your program can be turned into a proxy, that's usually a game over vulnerability. Surprising, right? It's not intuitive that just being able to proxy HDP request would be that big of a deal.
But when software is deployed for real in production environments, that ability to make an arbitrary HDP request is usually the end of the story. Right? From there you can bounce to some internal tool that people didn't expect you to be able to talk to and then take over machines and things like that. I think it's worth trying to get a beat on what the most important vulnerary abilities are for the platform that you're building on. It's good to think about there are things that you can build that are riskier and there are things that you can build that are less risky. I think it's probably a little bit underappreciated how risky it is to build with cryptography. I would always recommend that people avoid building new cryptography and instead use things like libsodium that have already gotten these things right. I think those things are surprisingly hard to get right. It's a good question. I feel like my answer here isn't that great, but that's the first thing off the top of my head is just thinking more in terms of vulnerabilities and how you're getting your head around with the vulnerabilities are that you're trying to mitigate.
David: How do you think that's changed over the time you've been in the industry? Or how do you think the vulnerabilities or the security issues developers have to face and mitigate have changed? Or is everyone just making the same mistakes?
Thomas: We're making new mistakes. When I started the default language that almost any open source project would be built in was C and thankfully C is now a rarity. Good bit of advice for people thinking about adding dependencies to their stack is to avoid the dependencies that are written in C. We're on our way towards eliminating memory corruption vulnerabilities. I think that's increasingly a problem for Google and Apple, where they have tens of millions of lines invested in memory unsafe code, and they can't get rid of them. And they just have to keep playing cat and mouse with people who are finding ways to weaponize memory corruption vulnerabilities. And if you're a smaller company or an open source project, you don't have that problem. You can just use Go or use Python or use Rust. And those are all great decisions. And you should do that.
I've been doing this since the '90s and when I started the vulnerabilities that I'd mostly care about would be buffer overflows. Would be things I could write shell code for and take systems over. And then in the early aughts things shifted towards web vulnerabilities. So, SQL injection was the big one for a long time. And I think people still aren't really great at thinking through web vulnerabilities. And that's the place where I'd probably pay the most attention to, especially because almost anything you build now is going to somehow be connected to HDP. So, I think a lot more in terms of web vulnerabilities than I would classic buffer overflows and things like that. You can pay attention to things like PortSwigger is a really good example of this. PortSwigger is a website and a company and I'm a little bit hazy on how all this works, but they make a product called Burp Suite, and Burp Suite is the industry standard web application security testing tool.
Literally every single consultant that does any kind of security assessment work has a license for Burp. Burp is pretty good as a tool. There's a free version of it. You can download and use to test things. Minimum proxy or MITM proxy, man in the middle proxy, is kind of an open source equivalent to Burp. I kind of recommend that people take a look at Burp itself. So, another random little bit of advice is if you're at the point where you can actually invest something into doing security for your startup or for your open source project, I'd look at getting a license for Burp if web vulnerabilities are relevant to what you're doing. Burp is pretty great tool just in general for testing applications, not just security vulnerabilities, but PortSwigger itself, the company that does Burp, has a really good news site for new vulnerabilities.
If you're serious about keeping up with this stuff, I'd probably recommend paying attention to they highlight new vulnerabilities. They're kind of connoisseurs of new web vulnerabilities, which I am too. So, maybe I'm biased. I just like it more than ordinary people would. But pay attention to when people announce. In particularly when places like PortSwigger kind of highlight new vulnerabilities that they think are really clever and just think about how those things might apply to what you're building as well. They're usually pretty ahead of what's going on. Every year there's a conference called Blackhat, which is kind of the major industry software security vulnerability research conference. And every single year, there's a talk from the PortSwigger people on some new vulnerability they've discovered. Sounds kind of silly when I say it, but pay attention to PortSwigger. A bit of advice I have there.
David: Would you say that a lot of the attack surface has now essentially been outsourced to the cloud providers, given that we're running, at least a lot of new projects are deployed, on the cloud? Is that how developers should think about that?
Thomas: I'm a hundred percent biased here. Right? Fly.io is a cloud hosting provider. Right? We're doing essentially the same things that AWS does. We run some of AWS's code in that we run Firecracker for virtual machines. Speaking against my own interests. Right? There's definitely a big sense in which when you deploy things on cloud providers, you are handing over a lot of security decisions to those providers. And your default assumptions about what those providers might or might not be doing for security are... I trust Amazon a lot. I trust AWS a lot for security stuff. They have just an army of very smart people just working on security problems there. When you're working with smaller companies, and we're a smaller company, you should think about how they're keeping up with that same work with far fewer people doing security work there.
For instance, there's lots of places that will run a Docker container for you or that'll check code out of GitHub for you, build it and run it on their cloud environment. I think it's worth asking yourself how that stuff is being actually run. What's the multitenant environment for it? Are they just running Docker containers? How have they locked down those Docker containers? Things like that. It's worth asking that question and getting a sense of it. There is something to the idea of you're trusting the engineering team for that hosting provider to provide a big part of your application security.
David: Before we wrap up that then, I have two quick fire questions for you. So, first is what is the current favorite dev tool that you're using at the moment?
Thomas: I will tell you, but it's a terrible answer. Right? So the favorite dev tool that I've kind of adopted over the last year is Emacs Tramp. Emacs' the editor everyone knows. Emacs' the list operating system that is also an editor. Has a feature called Tramp, which has been around forever but I've only been using it for the past year. I'm in a place now where my daily driver is a MacBook, but a lot of the software that I build now is very Linux specific. So, for instance, kernel BPF code to do network management and stuff like that. That code is very difficult to test even in a virtual machine on my MacBook. So, I have a dev machine, which is like an Intel Nuc that sits there running precisely the version of Linux that we run in production. And I build all my software there.
And Tramp makes it completely transparent to edit code on that Nuc as if it was on my local daily driver dev machine. Everyone that's ever used Emacs apparently, besides me, knew this forever. I always got the idea that you could use Tramp to edit the file on a remote system. Sure, it'll copy the file over and you can edit it and it'll copy it back. But no. Tramp does way more than that. Anything you do in Emacs to syntax highlight your code or check for problems in your Go code, everyone that uses Emacs also uses a tool called Magit, which is the Emacs' get interface. And Magit alone is a reason to use Emacs. Magit works over Tramp. I sort of know intellectually how that works, but it still blows my mind that that works. So, my recommendation is switch to Emacs because it's the best possible editor and then use Tramp because Tramp is amazing.
David: You partially answered my second question, which is what is your current tech setup? Hardware and software that you're using on your daily driver?
Thomas: I have a MacBook. It runs Emacs and it runs terminal Windows. It could be any MacBook. It would be just fine for me. So, I'm typing right now on a 2018 Intel, 15 inch MacBook. And my keyboard reads Q, W, 6, B, T, Y. I've gradually harvested keys from other computers that I have in my office as the keys on this one fail and just looking at the new keyboard I'm super psyched. So, I maxed out a new... It's pretty ridiculous. I have no use for almost anything in the new computer, but I've got a maxed out 16 inch MacBook coming to me next Tuesday. I'm very excited to have a keyboard where the keys will actually be the keys that I'm typing on. But really it's almost just a dumb terminal for me. Right? I just need Emacs to work. I need Emacs and Twitter to work and then I'm kind of off to the races. So, I'm pretty easy.
David: Excellent. Where can people find you online?
Thomas: I'm noisy on Twitter. My Twitter name is TQBF, the quick brown fox on Twitter. I'm pretty easy to find there. There's a site many people are familiar with called Hacker News. I think if you took the three highest karma, kind of biggest losers, on Hacker News and combined them, they still would not be as big of a Hacker News loser as I am. By a factor of three I'm the highest karma person on Hacker News. Which is where I waste a lot of time. So I'm tptacek on Hacker News if you want to see me ranting about random legal things or the tech industry there as well. So, Hacker News and Twitter are where to find me, unfortunately.
David: Excellent. Well, that's all we have time for today. Thanks very much for joining the Console Podcast.
Thomas: Thank you very much for having me. I appreciate it.
David: Thanks for listening to the Console Dev Tools Podcast. Please let us know what you think on Twitter. I'm @DavidMytton and you can follow @consoledotdev. Don't forget to subscribe and rate us in your podcast player. And if you are playing around with, or building, any interesting dev tools, please get in touch. Our email's in the show notes. See you next time.
Console is the place developers go to find the best tools. Each week, our weekly newsletter picks out the most interesting tools and new releases. We keep track of everything - dev tools, devops, cloud, and APIs - so you don't have to.
Subscribe to the weekly Console newsletter
An email digest of the best tools and beta releases for developers. Every Thursday. See the latest email