Mike and Erik return to the CAP Theorem to finish the discussion started last week. Their goal is to try to find answers to this question: why do software engineers love to talk about the CAP Theorem so much? This episode covers the 2002 Gilbert and Lynch proof of CAP, as well as more recent critiques of the CAP Theorem, mostly based on Martin Kleppeman's article “Please Stop Calling Databases CP OR AP”.

Links

• Brewer’s “Towards Robust Distributed Systems” (slideshow of the talk!)
• FLP Paper: Impossibility of Distributed Consensus with One Faulty Process (1985)
• Lynch: “A Hundred Impossibility Proofs for Distributed Computing” (1989)
• Lynch and Gilbert prove CAP Conjecture: “Brewer's conjecture and the feasibility of consistent, available, partition-tolerant web services” (2002)
• Martin Kleppeman “Please Stop Calling Databases CP OR AP”
• “Highly Available Transactions: Virtues and Limitations”

Send us a text

Your podcast hosts have a suspicion about the CAP Theorem: if you're a working programmer and you've heard of any results in the field of distributed systems, we think you'll have heard of the CAP Theorem. But did you ever wonder where it comes from? In this episode, we try to tell the story of where the CAP Theorem comes from. THERE WILL BE BLUEBERRIES!!!

Reach us by email: podcast@picturemecoding.com
Sponsor us on Patreon: https://patreon.com/PictureMeCoding

Links

• 2000 PODC website
• Brewer’s “Towards Robust Distributed Systems” (slideshow of the talk!)
• FLP Paper: Impossibility of Distributed Consensus with One Faulty Process (1985)
• Lynch: “A Hundred Impossibility Proofs for Distributed Computing” (1989)
• Brewer Interview in 2015: https://medium.com/s-c-a-l-e/google-systems-guru-explains-why-containers-are-the-future-of-computing-87922af2cf95
• Brewer interview with Software Engineering Daily 2023: https://softwareengineeringdaily.com/2023/05/12/cap-theorem-23-years-later/
• A Theoretical View of Distributed Systems: Nancy Lynch (2021 Talk)

Send us a text

This week Nathan Mull, a type theorist and CS Professor at Boston University, came on the show to help Mike and Erik understand what the phrase "Propositions as Types" is all about. This is an idea about how programs are connected to logic and mathematical proofs, whether we want them to be or not! You know that program that orders pizza from Dominos?! Yes, even that program is a proof of something. Find out what it proves on this episode of Picture Me Coding!

Links

• Nathan Mull's personal site
• 2014 Philip Wadler Paper: Propositions as Types
• 2016 Strangeloop Conference recording (Youtube): "Propositions as Types" by Philip Wadler

Send us a text

This week Mike and Erik discussed the work of Frances Allen, who worked for IBM for 45 years starting in 1957. The first female Turing Award winner, Allen authored a number of papers on compiler optimizations that describe techniques that are still in use!

Send us a text

In this episode we looked at the tragic story of tech company CEOs doing drugs and the various knock-on effects.

We also played a lightning round of "Aspiring to be a nation state or on drugs?" with various headlines.

This is the final episode of Picture Me Coding for 2024 and we’ll be taking a break for a few weeks after this. But we’ll be back next year!

Mike made a playlist of all the music we talked about this year too! Find it here: https://open.spotify.com/playlist/6jbM32yzvi8MK3zv2cXPSr?si=f40f8df000404a11

REFERENCES ROUNDUP

• NYT: “The C.E.O.s Are Tripping. Can Psychedelics Help the C-Suite?”
• The Verge: “Google reveals quantum computing chip with ‘breakthrough’ achievements”
• Google blog announcement of quantum computing breakthrough
• Scott Aaronson: Shtetl-Optimized » Blog Archive » The Google Willow thing
• TechCrunch: “Court orders Automattic to restore WP Engine’s access to WordPress.org”
• Wordpress’s Mullenweg interview with The Verge.
• Goodbye, unreliable weather forecasts? Google DeepMind's AI model sets new benchmark for 15-day predictions | TechRadar
• Google Blog: Introducing Gemini 2.0: our new AI model for the agentic era
• the Spectator “Are you ready for agentic AI?”
• Google Genai: new Python repo with “Part” data type
• Azure’s AI Agent Service announcement
• Toms Hardware: “AI GPU Clusters with one million GPUs are planned for 2027”
• Big Tech Embraces Nuclear Power to Fuel AI and Data Centers - IEEE Spectrum
• https://subseacables.blogspot.com/2024/10/breaking-story-facebook-building-subsea.html

Send us a text

We carried over from last week and kept doing algos! This week we talked about:
- Diffie Hellman
- Radix Trees: https://dl.acm.org/doi/pdf/10.1145/321479.321481
- Kadane's Algorithm
- Raft's Leader Election: https://raft.github.io/raft.pdf

Send us a text

Mike wanted to talk about our favorite algorithms this week, so we covered a few:
- Fast Fourier Transforms
- Hyperloglog

Some references are below:

- Article in Communications of the ACM: What Is an Algorithm?
- Al-Khwarizmi (mathematician, wikipedia page)
- HyperLogLog Paper: https://algo.inria.fr/flajolet/Publications/FlFuGaMe07.pdf

Send us a text

In this episode, Mike and Erik go back to the topic of time in order to explore what is specifically frustrating to programmers about dealing with time. Why's it so hard and annoying? And why do people get so irritated when we ask them a bunch of questions about what they really want us to build with respect to time?

We also mentioned on this and our last episode that there are some new ways to reach out to us (in addition to our email address), but they are:

- Bluesky account: https://bsky.app/profile/picturemecoding.bsky.social
- Threadless store (in case you want a Picture Me Coding coffee mug or stickers!): https://picturemecoding.threadless.com/
- Existing email address: podcast@picturemecoding.com

Send us a text