Schema evolution?

[jfallows]

Given that the flyweights are used to communicate via shared memory, which is concurrently visible to multiple processes, it is likely that on occasion different versions of a flyweight schema will be in use.

We either need to detect and fail such scenarios, or allow older flyweights to continue to work if possible.

For example, if a new required field is added in v3 of a schema, then a v2 schema flyweight can read the data, but if a v2 schema flyweight is used to write some data, a v3 schema flyweight would fail validation.

We also need to take inheritance into consideration, such that the base type might add a new field, either fixed size or dynamic size.