Why dataflow only?

[sosna]

The SDMX-CSV spec states: "For the first column, the dataflow column, always is the term DATAFLOW".

Typically, SDMX data messages require a reference to structural metadata but this reference can be a data structure, a dataflow or a provision agreement. Considering this, is there a reason why SDMX-CSV accepts one of these 3 types only (i.e. dataflow)?

Also, the minimum required to parse an SDMX data message is a data structure. So forcing a dataflow might seem to push the bar a bit too high, especially in cases where dataflows are not available.

Last, CSV messages have no header and therefore no way to pass the provider ID (unless I’m mistaken). At least, if you could reference a provision agreement instead of a dataflow, there would be a work-around.

A solution would be to:

• Rename the 1st column Structure
• Specify the artefact type (e.g. Dataflow=BIS:CBS(1.0) instead of BIS:CBS(1.0))

What do you think?

[dosse]

Good idea. Will discuss in the TF3.

[dosse]

Released