Ask HN: Why do relational databases index instead of a separate module/process?

Sphinx [1] provides SphinxQL which is a subset of SQL. Assuming you follow Sphinx's table formatting rules and SphinxQL can handle your SELECT query then your SELECT query can be much faster than querying the database directly [2].

So my question is why not create something like Sphinx, but have it focused on indexing relational databases for the purposes of generic SELECT statements? Is this a good idea? Would this be faster? Why or why not?

[1] http://sphinxsearch.com/

[2] http://www.mysqlperformanceblog.com/2009/04/19/talking-mysql-to-sphinx/