text stringlengths 0 2.35k |
|---|
**Johnny Boursiquot:** Yeah, it makes sense. |
**Baruch Sadogursky:** Right. So this is kind of our vision, and obviously, it's just a patchwork of the problems that arose in the last years, and the solutions that we saw working or not working. I can tell you an example... So Maven Central, for example, has been struggling with the problem of this authenticity fore... |
So basically, if I send an email to you, and that's our first correspondence, you don't know who am I, but Natalie knows who am I. She can propagate this knowledge to you. Now, it works in email, but for distributing binaries throughout the world, you have no idea who am I. |
\[25:42\] So you downloaded my PGP key. It says "Baruch Sadogursky", or it might say "Brad Fitzpatrick." How do you know that I'm really whoever I claim I am? I can generate a pair of keys for any identity that I want. And if the only requirement of the central repository is to provide them with this key, I can be whoe... |
On the other side, you don't really care. If I created a module, you don't care what my name is. And even if I'll tell you my real name, it probably won't help, because you don't know me. What you need is someone who you trust to vouch for this module. And this is where when we have those golden registries, or golden C... |
**Johnny Boursiquot:** Okay. So in this new world, if I'm building software, that first decision to use a dependency - I'm still making that choice, right? So if I go on GitHub, Bitbucket, what have you, and I'm looking for a library that does X, then that job of identifying a library and deciding to trust it - maybe I... |
**Baruch Sadogursky:** And here we are back to the problems that cannot be solved by computer science. And there are a bunch of them. We're just getting started. We work on that set of problems which is solvable. Those are the low-hanging fruits. How about the problem that the project is hijacked? You'll remember proba... |
So whoever tells you they're going to solve the problem of supply chain security, spit them in the face. They are lying. |
**Break:** \[29:57\] |
**Natalie Pistunovich:** So a question from the crowd, from the GoTime Slack channel. Somebody's asking whether binary compatibility decisions are different among organizations. And Louis St. Martin, who's writing this, is saying that at work they have this situation that they know what is a good code, and reference th... |
**Baruch Sadogursky:** I'm not sure I understood, frankly... |
**Natalie Pistunovich:** If there is something like good usage patterns, and good style, and recommended whatever RFC equivalent would be for things like binary compatibility decisions specifically. |
**Baruch Sadogursky:** Yeah, so I think -- again, if we look for standards, something codified, I would say that semantic versioning is actually a very, very good system to guarantee backwards compatibility. But again, with a twist, obviously, and the twist is humans. |
The idea is that "Hey, we can rely on whatever version of patching, patch version to be binary compatible, and minor version be public API compatible, and then minor not compatible to anything, and that actually should work for everything, until we hit the problem that "Hey, someone did incompatible things and called i... |
How you can prevent humans from breaking the guarantees that we build our machine systems upon? That episode - that's a really sad episode. Sorry about that. |
**Johnny Boursiquot:** \[laughs\] |
**Natalie Pistunovich:** Maybe it will inspire somebody for something. |
**Johnny Boursiquot:** It's like everything -- it sucks. \[laughs\] |
**Natalie Pistunovich:** Yeah. Social engineering, but for packages. |
**Baruch Sadogursky:** I mean, social engineering is malicious, but even honest mistakes. Let's not go and assume the worst... But even if we assume the best in people, people make mistakes, and there is no easy way to catch them always, and this is why we have bugs; and that's fine. But we have a system that envisions... |
**Natalie Pistunovich:** \[34:14\] So another point that came up on Slack by Henry Snowpack is in the context of npm registry compromise. So Henry is saying that vendoring could solve that, to a point... |
**Baruch Sadogursky:** Yup. Yup. |
**Natalie Pistunovich:** ...if you require review checks before the dependency is being updated. Or I guess it's a question... |
**Baruch Sadogursky:** So two things here. First of all, let's talk about vendoring. Vendoring is the worst type of working. You take someone else's code, you detach it from their version control, you dump it into your version control, and you're all for the races. You diverge from the original development; bringing th... |
The benefit of vendoring - and this is why Go relied on vendoring for a decade - is because it kind of ignores that supply chain problem. As a matter of fact, you're saying "I don't use any third-party dependencies at all. All the code is mine. I take full responsibility of whatever I'm using, whether I brought it thro... |
So yes, it solves this particular problem, but the solution is usually more problems than actually benefits. And the fact that Go switched to Go Modules from vendoring is, I think, kind of -- it attests to the fact that vendoring is not a scalable solution. |
Vendoring is terrible in some ways, probably not the right solution for the supply chain problem, and we need to look for other solutions. |
**Johnny Boursiquot:** I would say my counterpoint to that would be if what I care about -- if I have sort of a set of things that are important to me, for my builds, reproducibility of my builds, making sure I have the right version, that I've vetted, that I know work with my stuff, I'm not worried about a patch being... |
**Baruch Sadogursky:** The good news are you definitely don't need vendoring for that. Like, at all. Any modern package management today will give you a full reproducible build by using dependency managers, because you have multiple layers of caching that you can lock what dependencies you use for this particular build... |
\[37:53\] So you definitely can get repeatability, you definitely can reproduce the builds, you definitely can lock your versions that you are interested in after your vetting. Using modules or any dependency manager doesn't mean getting out there and grabbing the latest, the freshest dependency and dumping it in your ... |
If you just have a script that does something for testing Go, or whatever, you can say "You know what - I don't care. I can download the latest version and just try to run with it." Worst-case, it just fails. |
If you have a sensitive build - security, financials, you name it - then obviously it's a different game. You have a closed system that has to produce a reproducible build every time. The upgrades will be tested in multiple scenarios on different levels of testing, because before they will be allowed to be a part of yo... |
**Natalie Pistunovich:** Henry said "Thank you for your response, Baruch", and I am proposing we will jump to the fun part of the show, of an unpopular opinion. |
**Jingle:** \[39:59\] to \[40:18\] |
**Natalie Pistunovich:** So I wonder if most of the thoughts you were sharing, Baruch, will be ending up popular and unpopular, but I also wonder if you have an unpopular opinion for us, that is on any topic you want. It could be dependency managers, it could be anything else. |
**Baruch Sadogursky:** Yeah, we spoke for the last 45 minutes about one huge unpopular opinion... I will just summarize it to put it up to the vote. Dependency management was sent to us from hell to make our lives miserable. |
**Johnny Boursiquot:** \[laughs\] |
**Natalie Pistunovich:** Alright, alright, so any unpopular opinion goes through the vote on the Twitters. So we will soon find out what does the crowd think. |
**Baruch Sadogursky:** Okay, I'm going to turn on my Twitter bot farm right now. Because of my Twitter bot farm, that's why Elon Musk didn't buy Twitter. \[laughter\] Because they did manage to find it, and they know there are a lot of bots them, they know who owns them, and that's me. |
**Natalie Pistunovich:** And you operate them all with DevOps. |
**Baruch Sadogursky:** Of course, of course. |
**Johnny Boursiquot:** CI/CD pipelining for your bots. |
**Baruch Sadogursky:** Yeah, yeah, built with another package manager that was sent to us from hell to make us all miserable. Yup. |
**Natalie Pistunovich:** Alright. |
**Johnny Boursiquot:** Cool. |
**Natalie Pistunovich:** I will say thank you very much to everybody who participated on Slack. It's fun when the crowd participates. And I will say, Baruch, a very big thank you for joining and sharing your thoughts. That is definitely interesting to hear. Johnny, thank you for asking practically all the questions in ... |
**Johnny Boursiquot:** \[laughs\] Yeah, I was definitely interested in some fresh thinking, and definitely called out a few of the pain points that I've definitely felt all across my career... And it's interesting to know that there's some new thinking being applied to these problems. It doesn't solve all of our proble... |
**Baruch Sadogursky:** I really hope so. We are learning and getting better step by step, but yeah, what we need to keep in mind - we cannot solve everything with computer science, because... Humans. |
**Johnny Boursiquot:** Right. |
**Natalie Pistunovich:** \[laughs\] Thank you everyone again for joining. |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.