Monday, April 22, 2019

Joe Armstrong, RIP

Francesco Cesarini (Hacker News):

It is with great sadness that I share news of Joe Armstrong’s passing away earlier today. Whilst he may no longer be with us, his work has laid the foundation which will be used by generations to come. RIP @joeerl, thank you for inspiring us all.

Sugandha Lahoti:

Dr. Armstrong was best known for helping lay foundations in the ’70s and ’80s to the most widely spread concurrency models as we know them today. In concurrent programming, multiple events, code snippets or programs are perceived to be executing at the same time. Unlike imperative languages, which uses routines or object-oriented languages, which use objects. Concurrency oriented languages use processes, actors, and agents as the main building blocks. Dr. Armstrong helped propel concurrency programming at a time when there was no IoT, web, massive multi-user online games, video streaming, and automated trading or online transactions.


Erlang was originally built for use only at Ericsson, as a proprietary language, to improve telephony applications. It was designed to be a fault-tolerant, distributed, real-time system that offered pattern matching and functional programming in one handy package. It was then open-sourced to the public in 1998. Since then, it has been responsible for business, big and small, to create reliable systems.

Greg Hurrell:

His last tweet:

Joe Armstrong:

Once upon a time my boss asked me to study if we should use C++ or Erlang for a specialist XML parser to be used in a product (for reasons of speed not energy).

My recommendations was an FPGA

We built an FPGA.

Relative speed of C++/Erlang was irrelevant compared to FPGA.

Michael Gasch:

Apparently, the creators of the language found a way to trick the mighty failure gods. What have they done differently? Well, amongst many carefully evaluated design decisions that went into the language, they came up with the philosophy of “Let it crash”. I.e., instead of writing a lot of defensive code to handle every possible corner case, you accept that there will be failures, where you separate the concerns and take corrective actions.

The separation of concerns in this case is based on the concept of a supervisor and supervision trees, where the supervisor deals with failure (crash) handling when faults in the business logic (worker processes in the tree) occur. Another advantage is that these processes don’t have to run on the same machine to benefit from the supervision concept. Furthermore, in Erlang, processes are isolated from each other and don’t share state. So you can further reduce the blast radius of failures on a process-level.

See also: Wikipedia, Let’s Talk Concurrency (via Hacker News), Why OO Sucks (via Hacker News, note).

Previously: A Taste of Erlang.

Update (2019-05-10): See also: uses this (Hacker News) and The Guardian (Hacker News).

Comments RSS · Twitter

Leave a Comment