The main statistical standards for data and metadata exchange are XML-based (DDI, SDMX, or its semantic siblings: SCoVo/SDMX-RDF...): they are usually complicated and verbose: probably not best suited as an exchange format for apps.
Could stats benefit from JSON as an exchange format? How should data and metadata be expressed?
The ultimate goal of json-stat.org is to define a JSON schema for statistical dissemination or at least some guidelines and good practices when dealing with stats in JSON.
This question is relevant because today none of the existing statistical standards supports the JSON format. JSON is a lightweight data-interchange format best suited for (web/mobile) apps, a space where the presence and use of statistical information will probably grow in the future. But the benefits of JSON will likely not be limited to that web/mobile apps’ space.
This question does not assume that a new standard should be created from scratch, ex novo. Existing standards (like SDMX-ML) could be used and automatically converted from XML to JSON.
Translating an exiting standard, like SDMX-ML, into JSON has to face the following issues, though:
So even if you think the right path is to translate SDMX-ML (or another existing XML-based statistical standard) into JSON, we’d better agree how to do it.
All standards are defined with some use cases in mind. Take SDMX. SDMX is not different. SDMX was about heavy statistical organizations exchanging information regularly with heavy statistical organizations. It is not a coincidence it didn’t support a REST interface since very recently or that it does not support the JSON format yet. It is not a coincidence that it is verbose and complicated but flexible and rich.
SDMX was clearly not about statistical producers exchanging information with many different sporadic third-party hackers developing web and mobile apps for a broader audience. This and other use cases were not taken into account when SDMX was defined.
That’s why deriving JSON from SDMX-ML might lead us to inefficient JSON. Again, that does not mean that a new JSON standard should be defined from scratch: it can benefit, of course, from the information model of existing standards (like SDMX-IM).