A serialization/deserialization mechanism that do not require a static definition of schema

[zhangwei217245]

A serialization/deserialization mechanism that do not require a static definition of schema

This mechanism should allow schema-free organization of data, and being able to provide serialization and deserialization to the dynamically formed structure.

This is going to be very useful for the dynamic aggregation/organization of heterogenous data, especially when there is a need to aggregate different type of querying result, to organize compound result of a request in a way that is not predefined.

Describe the solution you'd like

Logically, it is a big key-value format, and can be self-embedded, and inside, it has to be self-describing, meaning each element in it should contain the type/size information so that we know exactly how to parse the bytes.

Describe alternatives you've considered

Solutions that requires predefined schemas:

1. the serde framework in mercury
2. protobuf
3. avro
4. thrift serde

Solutions that does not require predefined schemas

1. JSON (string, not compact)
2. BSON (binary, compact, introduces external dependency)
3. MessagePack (binary, compact, introduces external dependency)

Additional Information

N/A