I have seen a few of these go into production. What behavior is desired?
1) user gets table name when table changes
2) user gets table name and PK for row when that row changes
3) user gets table name and all columns when row changes
Also, how reliable must the notification be? If the notification subscriber is disconnected when the change happens will they miss a notification? Below I use "persistent subscription" to mean that change notification is not lost when a subscriber is disconnected.
The systems I used usually did #2 (user gets table name and PK) and then the user can query the table to get the other columns. This can be inefficient because of the extra queries but it was simple. These systems also made notification subscription persistent. One example is:
https://www.facebook.com/notes/facebook-engineering/wormhole-pubsub-system-moving-data-through-space-and-time/10151504075843920
I think this type of system can make it a lot easier to integrate MySQL/MariaDB with other systems. Row-based replication makes it much easier.