BTW, John, thanks for the pointer to CONNECT, it's saved away in my notes in https://github.com/Schemepersist/sdbi-design

From: John Cowan <xxxxxx@ccil.org>
Date: Friday, September 27, 2019 1:45 PM

On Fri, Sep 27, 2019 at 12:00 PM Lassi Kortela <xxxxxx@lassi.io> wrote:

So this covers some of the same ground as JDBC, but also some extra

Since JDBC is one of the types, it covers all of it.

Unless that CONNECT engine is embedded in your Scheme, quite a wild path, Scheme <-> CONNECT <-> JDBC <-> a database like Cassandra, 4 processes minimum, unless Cassandra can run on JDBC's JVM.  TBD, working on Neo4j today.

Does this mean the queries to MariaDB using CONNECT are identical to
ordinary native SQL queries?

Yes, exactly.  So we could have two approaches to Cassandra:  native Cassandra QL, which has one set of strengths, or SQL via Maria, which has another set.

Eh heh heh, the only reason I'm dealing Cassandra into the mix, at least this early, is precisely because it has its own query language (and it's popular, appears to be sane, etc. etc.).  The idea is that we're less likely to bake in SQL specific restrictions if we out of the box support for real 3 different query languages, and for SQL, at least 2 varieties (PostgreSQL and SQLite3, both of which are now at the "Hello, world" stage).

- Harold