text stringlengths 0 2.46k |
|---|
**Jerod Santo:** \[laughs\] What is the box? |
**Leslie Lamport:** But here, I can tell you exactly what the box is. The box is thinking programmatically, as expressed in what I'm now calling coding languages, instead of thinking outside of that box. And the way to think outside of that box is to think mathematically. So a TLA+ spec is a formula, a real mathematica... |
**Jerod Santo:** Some of the challenge in this -- so I'm with you, I'm following you, and I agree with what you're saying. Some of the challenge is as programmers who are trained up here in the United States as I was, and go into the industry and begin our work, math is part of it, but it's not like a main part of it. ... |
**Leslie Lamport:** Yeah. And also, the math that you learned hasn't been taught very well. |
**Jerod Santo:** Well, a lot of the stuff I learned wasn't taught very well... \[laughs\] But that depends on your school, I guess. |
**Leslie Lamport:** No, simply because you've learned about sets and functions, right? Everything I write in the spec, these sets and functions, plus arithmetic - I presume you do pretty well with arithmetic. |
**Jerod Santo:** I've got that part down, yeah. |
**Leslie Lamport:** You've learned what some functions were, but you haven't learned how you would use them to describe what your program is supposed to do. And that's the part that you have to learn. That's the part that you're forced to learn when you start writing in TLA+. Now, I do what I can to educate, but as I'v... |
**Jerod Santo:** \[32:23\] Is there open source TLA+ specs out there that people can read? |
**Leslie Lamport:** Oh, yeah, go to the TLA+ website. There's a GitHub repository with a few dozen specs there. One of the things I hope to do before I retire is, after my current project, which is writing a book - but after I finish that, I hope to start organizing a set of examples that will help people learn. Whethe... |
**Jerod Santo:** \[laughs\] Well, if I might ask you to predict the future a little bit... I just noticed, I mean, you've been in industry, you've been at Microsoft Research, of course... So in the academic circles, in the industry circles for many, many years. We've had recent shows, we spoke with Kelsey Hightower. He... |
**Leslie Lamport:** No, it's not going to be too long. Not many years. |
**Jerod Santo:** Okay. |
**Leslie Lamport:** I'm not ready to announce a retirement date, but it's not going to be too far off. |
**Break**: \[34:06\] |
**Jerod Santo:** What's changed? When you first got into the game, what's changed and what's stayed the same? You've been in the industry a long time... Certain things change; as the song goes, the more they change, the more they stay the same. What's different today, 2023, versus when you were first in computer scienc... |
**Leslie Lamport:** Well, the big change since I started - and I started programming in 1957 - is that now nobody can understand programs. |
**Jerod Santo:** \[laughs\] |
**Leslie Lamport:** I mean, you just have to build things on top of stuff that you don't understand. Things have just gotten so complicated, because - as I like to say, things have advanced by evolution, and not by intelligent design. |
**Jerod Santo:** Ah. \[laughs\] |
**Leslie Lamport:** And that works great if you've got a million years to do your designing. But when you're doing it over a matter of decades, it doesn't work too well. |
**Jerod Santo:** It's certainly been compressed, I agree with that. |
**Leslie Lamport:** I'm sort of amazed that there hasn't been -- |
**Jerod Santo:** An extinction-level event? |
**Leslie Lamport:** Yeah, some meltdown. I mean, a really serious one. I suppose there -- well, I guess you sort of see some of them... |
You read about the state of California, it was -- I forget what it was, whether it was for their motor vehicles, armored motor vehicles, or something... A 100 million dollar project that just failed, that was just abandoned after a few years. So what happens on that scale? |
**Jerod Santo:** On the other side of the coin, we have young people today who can pick up a computer and a programming language like JavaScript, or like Swift if you're talking on iOS, and in a matter of days or weeks they can go from never having written any software, to deploying a game, or an idea of theirs into th... |
**Leslie Lamport:** Yeah. Well, I mean, coding languages have improved a lot since I've started - not difficult - in machine language... |
**Jerod Santo:** \[laughs\] |
**Leslie Lamport:** But yeah, there's a lot of good things in programming languages. |
**Jerod Santo:** But TLA+, that's the main thing that you're here to talk about, the main thing you want to pass on to our listeners... |
**Leslie Lamport:** Well, I would say not TLA+, I would say thinking outside the box of code. TLA+ is one way of doing it. It has its domain of applicability. The place where it shines is when you're faced with problems of concurrency. But that's not the only way of thinking above the level of code. |
**Jerod Santo:** Well, at the top I mentioned this Turing Award which you won in 2013 for your work with distributed systems... I want to ask you about the bakery algorithm; that seems to be both your favorite, and probably the one that you can explain to our listeners without multiple hours... I'm not going to ask you... |
**Leslie Lamport:** I think I would probably say that distribution is the mind of the beholder. |
**Jerod Santo:** \[laughs\] Okay... How so? |
**Leslie Lamport:** Well, this laptop that I have in front of me - I view it as just a single thing, system. |
**Jerod Santo:** Yeah, it's all right there. |
**Leslie Lamport:** \[41:55\] But to the hardware designer, this is a distributed system. You have all these chips talking to one another, and even inside the processor chip -- I mean, that processor chip might have been designed with somebody who was using TLA+. People at Intel used to use TLA+ in designing their chip... |
The distinction I make is between the traditional programs and concurrent ones. Traditional programs, you have one process specifically that's just doing the computing of the answer and stops. And in a concurrent system, you've got multiple independent processors acting, and they're not supposed to stop, they're just s... |
**Jerod Santo:** Yeah. An apt metaphor for concurrency might be the idea of movies. So you have a movie that has a series of events; it starts, things happen, 24 frames per second, however many seconds, and then it ends. And that's usually pretty normal, unless of course you're Quentin Tarantino, and now we're gonna go... |
**Leslie Lamport:** Actually, that's beautiful. That's an example I love, because that's the way to describe systems. Basically, think of them as a movie. And what distinguishes a digital system from an automobile is that you can't think of an automobile as operating in distinct steps; it's going continuously from one ... |
**Jerod Santo:** Right. |
**Leslie Lamport:** But you can think of a program or a distributed system, or the internet, as proceeding in discrete steps. And the way physics, science describes cars, or the universe, is in terms of its state. And that's the same way I describe a distributed system, in terms of its state. You can think of the state... |
But the other important thing, and the thing that a lot of people didn't use to do; I don't know what they're doing now, I've stopped paying attention to what theoreticians, or people who are designing specification languages are doing... But they would think of the movie as a movie of this particular device. And then ... |
\[45:59\] And there's one thing about - this a technical point, but it's the way that I think differently from most people... It's that if you take a movie and you add, you repeat frames, since -- you know, the thing about moving the frames is they don't represent any particular timescale... Because you might be able t... |
But what I say is that the frames don't meet any particular time scale, so that a frame, if a system, if you describe this frame, this system with a certain movie, then if you add frames that simply repeat what the previous frames have done, you haven't changed the description. Your system should still satisfy that. An... |
I realized they had that insight in 1983. And to my knowledge, TLA+ is still the only - well, the TLA that TLA+ is based on - formalism that I know of, that people use, that has that property, that if you write a specification of an hour and minute hand, an hour/minute/second clock satisfies it. |
**Jerod Santo:** And what's the name of that property? What would you call that property? |
**Leslie Lamport:** Well, I call it stuttering insensitivity. I think of these as stuttering steps... It's what I call them. Stuttering insensitivity. |
**Jerod Santo:** It's insensitive to those stuttering steps. And the reason is because - I'm asking you, I'm telling you... The reason why it's insensitive to the stuttering is because of the way that you think about it is about the order of events, not as much about the actual distinct timestamps. Is that right? If yo... |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.