By traditional approach we means all engines that use a unique score to rank documents (like all engine based on Lucene).

SQL queries are not relevant for text query, for example you have no notion of tokenization or proximity between words.

We plan to write a blog post on limit of SQL based search.