I think this project is a great idea but I think it's overly ambitious to be reasonably feasible in a near-term timeframe as it's basically talking about re-implementing 20+ years and thousands of man-hours put into BEAM dev into the browser.
It's also competing time-wise with simply "compiling Erlang itself to emscripten/WASM", since computing capability will still scale over time in speed and bandwidth and memory capacity to the point that simply "being able to get ALL of Elixir in the browser" will outvalue "being able to run some subset of Elixir in a single-threaded context only in the browser" (assuming performance is not an order of magnitude different).
I was just reading another HN article discussing the fact that OLAP is going away simply because computers have gotten enough memory and databases have gotten fast enough at querying columnar data stores that it's no longer necessary to add the "optimization complexity" of OLAP to get reasonable data analysis performance.
AMD has the Ryzen Threadripper now, 32 cores with 64 threads... If CPU development continues in that direction, this will be a BIG win for Elixir but only if it can take advantage of more than one thread.
That all said, I'm sure a naïve recompilation of Erlang BEAM to emscripten/WASM might not automatically take advantage of those threads properly out-of-the-box due to the abstraction layer, so there'd also have to be some work there to massage the scheduler to use "real" threads "on the metal".
It's also competing time-wise with simply "compiling Erlang itself to emscripten/WASM", since computing capability will still scale over time in speed and bandwidth and memory capacity to the point that simply "being able to get ALL of Elixir in the browser" will outvalue "being able to run some subset of Elixir in a single-threaded context only in the browser" (assuming performance is not an order of magnitude different).
I was just reading another HN article discussing the fact that OLAP is going away simply because computers have gotten enough memory and databases have gotten fast enough at querying columnar data stores that it's no longer necessary to add the "optimization complexity" of OLAP to get reasonable data analysis performance.
AMD has the Ryzen Threadripper now, 32 cores with 64 threads... If CPU development continues in that direction, this will be a BIG win for Elixir but only if it can take advantage of more than one thread.
That all said, I'm sure a naïve recompilation of Erlang BEAM to emscripten/WASM might not automatically take advantage of those threads properly out-of-the-box due to the abstraction layer, so there'd also have to be some work there to massage the scheduler to use "real" threads "on the metal".