37638894breeze

The semantically significant bit is "technology" and that is what is tripping up the "mah local model" group.

Technology develops as the result of focused application of capital and shaping the evolution of that technology to serve the interest of that capital. The driving force behind AI -- when ranked according to impact and significance -- is being funded by surveillance capitalists. In that sense one can say "AI is fundamentally a surveillance technology".

AI and Capitalism are a toxic mix, in my opinion.

I like the "X, [the platform] formerly known as X" its evolution. Latest in the wild sighting was "X, formerly Twitter".

What I like about it is that it lifts the skirt on the superficial divides. Good soldiers of the information regime all toe the line here and dutifully help Elon with his illadvised brand name destruction. I mean, I go to geopolitical websites run by people who like Putin, for example, and even there, the resistance blogger/propagandists dutifully remind us the "X, formerly Twitter" just in case we forget or haven't heard the news. We know things are bad if even North Korean output is using the "formerly Twitter" insert. Anyone checked?

You could go the route of having a zoo of many small libraries and just version the library, so mod-x becomes lib_x. That, dependency management, is not a convincing argument for something 'new' called "module". You can do it with libraries as well.

The question (then) remains: are modules really just libraries? Was it always just about coexistence of related functions?

I have to listen to the Joe's talk on this but from OP this is more accurately "Why do we need Erlang modules at all?" or generously "Modules in FP languages". A key motivating pattern (fib/3) is a pattern in FP.

More directly in terms of Joe's brainstorming:

- I don't see how the versioning matter is simplified by a flat space of functions. Before you had Nm modules to track and now you have Nf functions to track, with Nf >> Nm. Aggregating functions in library/modules to version is actually helping with versioning effort, not hindering it. More generally, the versioning of multi-component systems are complex affairs that can only be addressed by constraints - general engineering systems have standards + catalogs as the means of addressing this general engineering issue.

- Broadly I disagree with conflating modules and libraries. They are distinct conceptually. Modules could have state (and meta-data state), conceptually. Modules potentially could also have active elements internally. Modules can have life-cycles. To sum: modules conceputally are not just collections of (related) functions.

So the general question is 'can we live with just libraries of functions?'

I think PLT excitement here is not 'a k/v bag of richly annotated functions' -- the guaranteed end result of that approach is n variants of elaborate 'structure' encoded into the metadata Joe is talking about -- but rather pushing modules to extreme to make the distinction from libraries crystal clear.