Peter wrote about this recently, but I don’t know if it was really clear what was going on.
Point One: Sphinx can be contacted by the MySQL protocol. Not “as a MySQL storage engine.” Not “from MySQL.” It understands the MySQL protocol itself. So from the protocol point of view, the Sphinx search daemon can look just like a MySQL server.
Point Two: Sphinx understands a SQL-like query language. Don’t be fooled. You’re not writing SQL. It just looks like you are.
Point Three: Because of point One and point Two, you can use the mysql command-line client program to talk directly to Sphinx, with absolutely no MySQL server anywhere in sight. This also means you can connect to Sphinx from your application and query it, exactly like connecting to a MySQL server and querying it.
Go take a look at Peter’s blog post. He’s not writing MySQL queries. He’s writing queries to Sphinx.
Now think about how cool this is—how easy this is to integrate with your code that already communicates with MySQL. Is there any other external full-text search system that masquerades as a MySQL server? I don’t know of one.