From the mailing list discussion, we need to define the structure of data headers to describe a "page" of Arrow data. In Apache Parquet these are referred to as the data page headers and are separate from the schema metadata.
As part of this JIRA:
- Choose a technology (it need not be the only one, but it would be good to have a reference tool) for representing the data page header in "serialized" form (as can be stored in shared memory and interpreted by any system with access to that memory). Google's Flatbuffers libraries has been discussed for this use case.
- Draft IDL specification for the data headers