Row-row based logging in a blackhole replication filter setup?

Question

I am using two MySQL server instances on the same server to filter replication to a third external server. My filter slave is using the blackhole engine as described here: https://dev.mysql.com/doc/refman/5.7/en/blackhole-storage-engine.html

Both master and slave use statement based replication. The documentation says:

Inserts into a BLACKHOLE table do not store any data, but if statement based binary logging is enabled, the SQL statements are logged and replicated to slave servers. This can be useful as a repeater or filter mechanism.

The above statement makes me assume that if I had both of my MySQL instances set to row based replication, nothing would make it to the third, external database. Which kind of makes sense since there are no actual rows in the filtering blackhole database.

However, I have been thinking... Would it not be possible for the filtering middle instance to simply pass on any row based instructions it receives to its own binlog, i.e. would a row-row filtering blackhole setup not work?

Answer 1

https://dev.mysql.com/doc/refman/8.0/en/replication-features-blackhole.html -->

The BLACKHOLE storage engine accepts data but discards it and does not store it. When performing binary logging, all inserts to such tables are always logged, regardless of the logging format in use. Updates and deletes are handled differently depending on whether statement based or row based logging is in use. With the statement based logging format, all statements affecting BLACKHOLE tables are logged, but their effects ignored. When using row-based logging, updates and deletes to such tables are simply skipped—they are not written to the binary log. A warning is logged whenever this occurs.

For this reason we recommend when you replicate to tables using the BLACKHOLE storage engine that you have the binlog_format server variable set to STATEMENT, and not to either ROW or MIXED.