I’m more biased towards ‘transformers’ like the the Cambria project than common schemas because:
- it seems simpler to assume multiple ‘formats’ and small translator functions to move between them. At least in JSON this should be straightforward and it’s a setup more organized for responding to change.
- it requires no coordination between people.
- similar apps may use similar data in differing resolutions or complexity.
- I would say standards move slower by definition, which I tend to find limiting.
The other day I was imagining a JavaScript library that could provide a common API to remoteStorage, Fission, and perhaps SOLID as well (if it’s possible to idiomatically represent JSON objects there). Different from the issue you’re bringing up but maybe part of a solution. I think that things like schema translation and versioning should be ‘built into the library’ developers use so that it happens in a consistent way.