What do I need to known on front-end(e.g. JavaScript) about remote(back-end) classes? about cereal

[AzothAmmo]

When cereal saves something like an std::shared_ptr, it ensures that it only actually serializes the data once. Subsequent saves of data at the same address simply store an id (some number), which refers to the data already saved.

When cereal does a load of an std::shared_ptr, it will properly load the data on the first instance and after that look up the id instead of doing a full load.

If you are trying to parse the output from cereal by yourself, you will probably need to take a look at some of our logic for how we use these tags. We also use similar metadata for storing information for polymorphic classes, which operates in the same fashion but does some more complicated things upon loading.

In general you should be able to look at the serialization functions we provide for whatever type you care about (e.g. looking in memory.hpp) to see how metadata is generated and used.

I would strongly recommend that you do not hand generate pointer based serialization as this involves hash values generated using built in C++ features. If you do need to create things by hand and then give them to cereal, try it without using a smart pointer.