Require more stringent backward compatibility for resource.type

[peterdesmet]

What Data Package version are you using?

v2

Describe the Issue

Data Package v2 has the following backwards compatibility note for resource type:

If a resource has profile property that equals to tabular-data-resource or https://specs.frictionlessdata.io/schemas/tabular-data-resource.json, an implementation MUST treat it as type property were set to table

In my opinion that is not sufficient. "profile": "tabular-data-resource" implies a v1 Tabular Data Resource, which has more requirements:

property	requirement	implications on note
profile	required	Implied
profile	must be tabular-data-resource	I think we should drop https://specs.frictionlessdata.io/schemas/tabular-data-resource.json as that was never a spec-defined approach
schema	required	Mention this
schema	must follow Table Schema	Implied when referring to schema
data	must be JSON tabular data	The rules slightly different than what is currently defined for inline data, but not sure we should mention that
path	"Be a CSV file"	Mention this (e.g. as mediatype). Note however that "must be a CSV file" contradicts the beginning of the spec, which says "specialized for describing tabular data like CSV files or spreadsheets."

Update note

I think the backwards compatibility note could be rephrased to:

If a resource has a profile property that equals tabular-data-resource, an implementation MUST treat it as a resource where type is set to table, schema is defined, and mediatype is set to text/csv for files provided in path.

Good old Tabular Data Resource

I like the stringent approach of Tabular Data Resource: it is easier to implement for. Data Package v2 offers more flexibility in "type": "table": it can describe any tabular format and schema is not required. Would it help to bring back a more stringent implementation as a new value for type (in addition to "table")? For example "type": "csv"` explicitly expects:

• schema MUST be present
• path MUST be present
• mediatype MUST be "text/csv"

Ping @roll @pierrecamilleri

Participation

• I am willing to submit a pull request for this issue.