The context in which I mentioned was that RDBMS companies have been refining join/sort algorithms for years and be more useful to just use it, rather than refine/redefine again.