Fascinating. I haven't finished reading the entire presentation yet, but I already stumbled on something.<p>The author lists a drawback of the first antipattern ("metadata tribbles"): a table split into multiples to keep its size down requires tricky querying, specifically, it requires a union across all splits (slide 10). It also requires extra effort to keep table structures synchronized (slide 11). Then, in a solution to his second antipattern ("entity-attribute-value"), he suggests doing something quite similar to his first antipattern: creating multiple tables with nearly identical columns and using a union to query across them (slides 27 and 28).<p>Hello? So to resolve antipattern 2 you just apply antipattern 1? This is why precisely why relational databases for most problems look like square pegs being hammered into round holes.<p>[EDIT] Not sure if it's worth finishing reading this presentation. Slide 56, a "solution" to the problem of storing a hierarchy in a relational database. Breadcrumbs. Not bad, except that the query now requires a "like" clause. Whoops, can't use an index on that column anymore! (Feel free to correct me if there exists a RDBMS which can use an index on a "like.") Hope the author doesn't mind a nice table scan on that query. He doesn't even mention this problem.