before: 'Hmm, I need to do the thing with the data'; hack together SQL in Squirrel until it gives the correct answer; copy SQL into program; think 'it worked in dev and test' when it turns out to be horribly slow in production; sprinkle indexes; see moderate improvements; leave alone.
now: 'Hmm, I need to do the thing with the data'; write SQL to get data, bearing in mind table indexes, join criteria, and the ease with which the database optimiser will be able to narrow down to the target data; run explain plans and check costs; remember that I need real data volumes for the explain plans to be representative of production, and repeat; copy SQL into program; never hear about it again.
And all in less than two hundred pages which--while they could be quickly summarised as 'use indexes effectively'--don't waste any time with history lessons, edge-cases, or other forms of padding. Throughout this, it still manages to highlight the differences between Oracle, SQL Server, MySQL and PostgreSQL.
I can't really offer any criticism here, hence the five stars, but if there is ever going to be a follow up then I'd like to see some coverage of DB2.