This is inspired by DuckDB in which rows, not columns, are the primary thing operated over.
By allowing to specify what the table interaction model is, we would also get lightning speed for cross-row aggregation and other such benefits.
Because the entire project is aimed at fast response time and allowing for massive amounts of requests per instance,
this would enable new usecases/improve those massively.
It is basically summed up as: bring the benefits of SpacetimeDB over SQLite to DuckDB, but inside SpacetimeDB.
And while at it, allow tables of both formats.