Hyrax JSON

Overview

With funding from the Australian Bureau of Meteorology we have developed prototype JSON data and metadata responses for Hyrax. After reviewing some the existing JSON encodings for DAP content we chose to implement two prototype encodings.

The first, and most likely the most useful, is based on the w10n specification as realized by the good folks at JPL. This encoding utilizes an abstract model to capture the structure of the dataset and it's metadata. In this model the properties of the JSON object are made of a controlled vocabulary. This means that clients utilizing these responses can always "know" what to look for in each returned object. No matter what dataset is being accessed the client has a consistent mechanism for extracting variable names and values.

The second encoding utilizes an "instance" representation model wherein the datasets variable names are used to create the properties of the returned object. This means that each dataset potentially has a different set of properties and that client software must be written to navigate each dataset. For data providers with large sets of homogeneous holdings this representation allows the quick development of targeted clients that can work with these data. However since the variable names form the dataset become JSON properties there is no promise that the JSON objects will actually be valid as variable names in DAP datasets have few content restriction and the JSON property names must be valid Javascript variable names. Because of this this second representation probably doesn't have the required flexibility to become an official JSON representation for the DAP.

The intention is to develop this work (in particular the w10n representation) into a DAP4 extension that defines the JSON representation for the DAP4 data and metadata responses.