text stringlengths 0 2.35k |
|---|
**Baruch Sadogursky:** But also, it's the name of our conference, the conference that JFrog organizes; a community DevOps conference in Israel. |
**Natalie Pistunovich:** So Yalla is the name of the conference? |
**Baruch Sadogursky:** Yeah, Yalla DevOps, actually. |
**Natalie Pistunovich:** \[04:12\] Yalla DevOps. |
**Baruch Sadogursky:** Mm-hm. |
**Johnny Boursiquot:** Cool. |
**Baruch Sadogursky:** Which is a very cool name. |
**Natalie Pistunovich:** Now we are here to talk about dependency managers. |
**Baruch Sadogursky:** Yes...! |
**Natalie Pistunovich:** And before we jump into that topic, Baruch, would you like to introduce yourself to the crowd? |
**Baruch Sadogursky:** Yes! So I hate dependency managers with passion, for the last (I would say) at least 15 years. That's kind of the relevant context. I've been a developer advocate with JFrog for (I think) 11 years by now, before that working as a Java developer, senior developer, or whatever it was... Architect..... |
My dream is to eventually sit down and write a talk about how terrible every dependency manager on the surface of Earth is. |
**Johnny Boursiquot:** Okay... \[laughs\] |
**Natalie Pistunovich:** I'm looking forward to that talk. Please tell me what conference, I will join. |
**Baruch Sadogursky:** Yup, yup. That'll be everywhere. Once I write it, I will go and preach about how terrible they are to everybody. The most annoying thing is there's no solution, really... And we can talk about that, why it is the issue, we can talk about later. And by the way, Johnny, thank you for wearing this T... |
**Johnny Boursiquot:** Yeah, that's actually a JFrog -- |
**Baruch Sadogursky:** Yeah, it's the \[unintelligible 00:05:59.14\] |
**Johnny Boursiquot:** It's completely coincidental. That was not planned at all. \[laughs\] |
**Baruch Sadogursky:** Um, okay... |
**Johnny Boursiquot:** It really wasn't, yeah. |
**Natalie Pistunovich:** It's just because the T-shirts are indeed cool, every year. |
**Baruch Sadogursky:** It's true, yeah. |
**Johnny Boursiquot:** I am interested in -- obviously, this is a space you're very familiar with... I come from basically a long history of using various dependency managers for various languages... And yeah, none of them are really perfect. Nor do I ever expect them to be, but each one has different pain points, each... |
**Baruch Sadogursky:** Right. I will quote Sam Boyer a lot. You obviously know Sam; I hope you hosted him on your podcast, and if you didn't, you should. |
**Johnny Boursiquot:** Oh, we have. |
**Baruch Sadogursky:** And if you did, you should do it again. |
**Johnny Boursiquot:** \[laughs\] |
**Baruch Sadogursky:** Sam is amazing... And basically, he summarized all the problems with dependency management perfectly when he said that dependency management looks like a computing problem, and we try to solve it as a computing problem, but it's actually a people problem. The biggest problem with dependency manag... |
\[08:43\] So you say "Well, I can always rely on the fact that a patch-level upgrade will work great because of backwards-compatibility", but it's not, because someone made a mistake. There is no way to promise binary compatibility without relying on people's opinions on whether their software is compatible or not. Doe... |
**Johnny Boursiquot:** Yeah, that totally makes sense. I can certainly see, for those who are not super-familiar with semantic versioning and how versioning is in theory supposed to happen - basically, you're not supposed to be changing existing APIs... Imagine a particular method that does something - you're not suppo... |
So we're relying on individuals sort of doing the right thing, but obviously, for popular projects, even though it happens, these tend to be sort of mistakes... So really what we're saying is human mistakes. This is sort of a grey area where algorithms can't really help you there, because it's a human making a decision... |
My question to that is "Is it really sort of a completely grey area that cannot be solved using an algorithm or using computers to solve the problem?" Do we not have tooling for that, or maybe we just haven't put in the time yet into solving to prevent these kinds of problems. It sounds like we could create some toolin... |
**Baruch Sadogursky:** Right. But that's the easy challenge, right? If you say like "Just check that the signature is the same and the parameters didn't change" and whatnot, that's obviously doable. But how about the behavior? So at the end of the day, you made the patch, and you made it for a reason, because you want ... |
In terms of the binary compatibility per se, your build won't break, because the parameters are the same, and everything -- and unless you test for this behavior, you won't even know. And then you release to production, and then you discover it, and now you need to roll back. Now, roll back - it's another hell of its o... |
So it's really a problem that has no solution, because it actually depends on people always doing the right thing, which is not what's happening. |
But this is only one side of the problem. I can tell you another one, which is also very human-driven... And that's the external dependencies. How do you get them where they are? How do you trust those who produce those dependencies etc? |
\[12:24\] Another quote -- and I don't remember whom... I think it was Brad Fitzpatrick that said that when you use external dependencies, it's like you just take someone else's random code from the internet and just dump it into your production systems. I think it was Brad. And he's on point on that. His solution was ... |
So he pinpointed the problem correctly. It is a terrible idea to just grab other people's dependencies and throw them into your production system... But what is the answer? And we as an industry try to come up with the solution for many years now, and Go Modules, the central repository, the signature server of the cent... |
This is another huge challenge, and especially today, when we hear all those stories about how the supply chains of those dependencies get hacked. SolarWinds for one example, and multiple others. That's obviously the problem which surfaced lately, but been with us for as long as we tried and used dependency managers. |
**Break:** \[14:20\] |
**Johnny Boursiquot:** So what would you say has been -- like, I have some idea of how you'd go about trying to not solve, but really mitigate... But I wanna hear what you've seen in your experience of how people sort of mitigate that second class of problems. |
**Baruch Sadogursky:** There were and still are a lot of efforts in trying to do that... And it looks like - this is from our experience as well; we tried to come up with -- again, back a couple of years ago when Go Modules only appeared, we heard this vision of "Hey, we will have some kind of a network of central modu... |
But then it was a very useful experience for us to really try and evolve this idea of negotiating what's \[unintelligible 00:17:34.25\] And I would say there are now two most important aspects of this system. The first is how do we know which module is the right module, in terms of "Hey, I have two of them, they're nam... |
\[18:30\] That brought us to envisioning a system that everybody will be able to use in order to guarantee those two, and it's called Pyrsia. Actually, Pyrsia - that's an interesting story of the name... It's actually a system that ancient Greeks used to convey messages with flames between each other. So this is why yo... |
The idea is a decentralized P2P network that will provide a consensus of what modules are authentic, and then make sure that they cannot be tampered with, with some kind of blockchain-backed ledger. I know that we lost two thirds of the audience right now, after I mentioned blockchain, but whoever stayed - you did the ... |
So basically, what we say is "Okay, everybody now can build modules from source, and publish them wherever they like." The system will communicate and decide if that's the golden build by comparing it to the build that the original producer of this module creates. Let's say I have a library now that I wanna publish, an... |
Now, we record it, so no one can hack into that and pretend that another result is the right one, and then we distribute it through a P2P network, because now we don't care where it comes from, as long as we can guarantee that it's the right results. So now Natalie builds it on her machine, and she's behind a firewall ... |
So this looks like, from our perspective, the solution to some of the supply chain issues, because it guarantees the authenticity, and it protects the supply chain itself, because we verify on your machine that the package you have is exactly the package you need to have, regardless of where it actually came from. And ... |
**Johnny Boursiquot:** \[21:59\] Does this approach require those who create those libraries or dependencies, that they basically follow a similar set of principles? I'm thinking of the recently made public Software Bill of Materials. Stuff that came out within the last year, for example... Basically, that requires a s... |
**Baruch Sadogursky:** Yeah, so this is the question of how we can verify that these golden module is really golden, is really good. And this can be achieved in multiple ways. First of all, we can say hey, let's say the Go community trusts the Google registry to a reasonable extent. And that means that if I produce a m... |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.