I was intrigued by this post because you seem to prefer the use of the BerkeleyDB and are only considering the MySQL method if you are given compelling enough reason, or at least that is what I inferred.

We however are not given enough information in my opinion to give you a compelling reason in either direction. You express concerns about running the MySQL daemon, but you really don't get into the nature of the data you are working with.
The lifespan, stability, and interaction over the lifespan would effect what is the better way to store the information.

That said, I would recommend the MySQL storage method because I feel/think it offers less/easier modifications during the data lifespan for most record oriented material.