While the discussion continues regarding policy decisions in the serialization library, I think I might contribute the following observation, having only recently become a user of this library: The library has the feel of a library for building serialisation, and serialisation-based libraries, rather than a fully-featured library suitable for a programmer to pick up and start throwing objects at. In this regard, the documentation might make explicit the choices and policy decisions that users have to make, and the limitations that obtain (for example, you can't just pick up a text oarchive, and hook it to a text iarchive with connected streams have things Just Work(TM), which is quite reasonable, but not totally obvious) because of the way it has (largely necessarily) been implemented. Please understand that I am not saying that the library cannot simply be used for simple use-cases - quite the reverse, when you only want to do simple things, then the library does just work out of the box. However, things are rarely that simple for very long. It may help to think of the library this way.