Some reference info on the topics discussed here and in the RML thread that might be useful: 2001 USENIX paper: PSTL - The Persistent Standard Template Library for C++ http://www.infosys.tuwien.ac.at/Staff/tom/Publications/Gschwin d2001-pstl.pdf http://www.infosys.tuwien.ac.at/NewsCache/pstl.html This papers covers the issues and has simple app benchmarks that compare to the widely used Berkeley DB (alse > note Berkeley DB provides the underlying b-tree for MySQL) Reflection support by means of template > metaprogramming http://www.di.unipi.it/~attardi/Paper/GCSE01.pdf Which has a good discussion on interfacing to a relational table using
meta programming Those are very helpful links. While benchmarks obviously only approximate real-world performance, I would be very interested in comparing the various approaches using a standard task. By providing a solution to a small "real world" problem, it enables you (the Boost developers) to compare both the performance and how a library looks like to use in practice, and is much more preferable to hypothesising. And we can also compare to the "STL" solution to see whether relational templates/queries really do make an application easier to develop. Regards, Calum