I understand your use case but flatbuffers are currently read-only once they're built. If you want to understand the gritty details read
http://google.github.io/flatbuffers/md__internals.html .
For example, flatbuffers tables with unspecified values for members or members who have the same value as the default value are *not* stored in the flatbuffer.
In other words, if a monster's HP is set to 10 and the default value is 10 no storage is allocated for the HP value in the monster's flatbuffer which is nice since we end up with a compact data structure but the downside is that you can't assign a new value to the data. This is primarily why we're not generating code to set values in flatbuffers.
If you want to use flatbuffers as a wire format, it's possible in a couple of ways:
1. Build the object you want to send to the remote (if it's not a struct).
2. Just send a struct to the remote.
To be honest, if you're careful with allocation during the flatbuffer build process, the overhead of sending a flatbuffer via a typical network stack is going to be *way* more expensive than the process of building the buffer.
Are you working on a game or application that you can share the details of?
Cheers,
Stewart