This is our AT Protocol journey here so far:
Starting with a toolkit to evaluate Blueskyʼs open source dataplane, we picked Elixir from all the languages in our toolbelt to build a proof of concept to serve timelines.
We still owe you the code.
Today, we are open sourcing our proof of concept of a dataplane.
The Closed Core of Bluesky
We talked about this before:
Itʼs great that Bluesky is running mostly on open source components.
However, they draw a line at the dataplane that sits at the core of the system.
Although Bluesky provides an open source implementation of the dataplane, this is not what they run in production.
They use a second, closed source variant based on ScyllaDB because only this implementation is ready for the production scale of Bluesky.
Itʼs a pity because we already see interoperable alternatives to Bluesky emerging, delivering on the promise of hard decentralization. Members of the Eurosky team have even been invited to the European Parliament.
Running alternative networks should be as easy as possible but we see each of these projects reimplementing their dataplane. There is a need for an open source dataplane implementation that will sustain the load when the network grows.
We built our proof of concept in Elixir, for many reasons. Most notably, itʼs just an excellent runtime to build such systems, while we can always mitigate its weakness in raw compute performance by bringing in native code with a NIF.
You can find the repository here: https://github.com/bitcrowd/dataplane_elixir/
We invite you to evaluate its performance with our simulator project.
Remember, it is a proof of concept, it is not a production ready implementation.
Give us a shout if you want to help us take it further, faster.