> Since MariaDB's over the wire protocol is backward compatible with MySQL,
> it is part of our first-class support set.

Yes

> And Maria (but not My) has a
> hidden treasure we can exploit: the CONNECT storage engine.  This is
> basically an engine for talking to sources of data other than classical
> base tables, such as are supported by MyISAM, Aria, InnoDB, etc.  Here are
> the sources it currently supports and rectangularizes:
>
> Fixed-length binary records
> CSV and friends
> dBase III/IV files (FoxBase etc.)
> Flat text files
> Filesystem directories (either hierarchical or non-hierarchical view)
> INI files
> JDBC
> JSON files (if the data is reasonably rectangular)
> MongoDB
> ODBC (including unixODBC)
> Column-oriented binary with one file per column (alpha)
> XML (if the data is reasonably rectangular)
> HTML tables
> Zipfile directories

Awesome stuff!

> In some cases you have to use a CREATE TABLE to specify the
> rectangularization, in other cases not.  Some are updatable directly;
> others are updatable but destroy the original (updating XML gives something
> matching the rectangular view, not the original more complex view), yet
> others are read-only.  See <
> https://mariadb.com/kb/en/library/connect-table-types-overview/> for links
> to more detailed docs.

So this covers some of the same ground as JDBC, but also some extra
(NoSQL databases and eccentric conversions of everyday file formats).

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

> There are also engines outside CONNECT that provide similar abilities:
> Cassandra, MyRocks, Object Query Graph.  See <
> https://mariadb.com/kb/en/library/storage-engines/> for links.

A lot.

I guess going the route with the fewest adapters is generally best,
unless the database at the other end is hugely complex and an adapter
simplifies the interface a lot.