Good question! I've spent some time looking at Calcite and similar systems.
My hope is that it isn't an "either-or" but an "and". I also believe that Calcite is primarily database-oriented, as are most similar systems. Trustfall is designed to be able to delegate queries (or portions thereof) to underlying systems: for example, you could take a portion of a query and compile it directly to SQL instead of executing it via an interpreter -- no need to worry about query planning over a SQL database when the db itself can almost certainly do that better. In principle, Calcite should be possible to be plugged in like that as well.
Same goes for other systems that do federation over data sources, whatever they may be. Whenever there's an awesome database or federation solution out there that covers some data sets, plug it right in! No need to reinvent the wheel -- Trustfall development can then focus elsewhere, and everyone wins. The goal here is to have a universal data system that works great for any data source in any programming language, not a set of small-time turf wars over whether SQL or GraphQL or $NEW_TECH is better.
My hope is that it isn't an "either-or" but an "and". I also believe that Calcite is primarily database-oriented, as are most similar systems. Trustfall is designed to be able to delegate queries (or portions thereof) to underlying systems: for example, you could take a portion of a query and compile it directly to SQL instead of executing it via an interpreter -- no need to worry about query planning over a SQL database when the db itself can almost certainly do that better. In principle, Calcite should be possible to be plugged in like that as well.
Same goes for other systems that do federation over data sources, whatever they may be. Whenever there's an awesome database or federation solution out there that covers some data sets, plug it right in! No need to reinvent the wheel -- Trustfall development can then focus elsewhere, and everyone wins. The goal here is to have a universal data system that works great for any data source in any programming language, not a set of small-time turf wars over whether SQL or GraphQL or $NEW_TECH is better.