It's not easy doing XML transformations in XQuery and it's not easy doing XML queries in XSLT. Meaning that they are solutions geared for different problem spaces. Not that I particularly like the complexities of either but they both have their reason for existence, even if they happen to overlap.

Similary, the consummate Turing Machine is not particularly easy to do day to day computing chores, but it's reason for it's existance is mostly theoretical and mathematical - to establish the outer limits for computability. In terms of expressing the concepts of computability, the turing machine is a very good (and easy to use language).

The emphasis in theoretic computability these days lies more with Church and the Lambda Calculus. Some languages have arisen from this effort - ML, Haskell, Scheme. But the primary emphasis in Lambda Calculus (and associated Combinatorial Logic) is providing a mathematical model of computability - in that frame of reference, the Lambda Calculus is much easier to build a foundation upon. Though there are other efforts such as the Relational Calculus (RDBMS), the Predicate Calculus (Logic & Declarative), and the Pi Calculus (distributed processing) - and that's not even getting into the usefulness of Category Theory.

Anyhow, the point (if there was a point) is that merely claiming turing complete is not much of a feat. It doesn't tell us much about how well a language can tackle a given problem given resource constraints. So, yes, Turing Tarpit languages like BrainF*ck are Turing Complete, but that doesn't mean they are useful beyond the offhand parlour tricks.