From the beginning the Mental Model has been "At a given time, poll one data point from the Internet and store the value".
We see more and more deviations from this over time, and I think it is time to challenge the validity of this approach.
HTTP documents (json or xml) may contain many data points of interest.
The timestamp of the datapoint can already be extracted from the document.
But some documents are summaries or "offline" collections, with many timestamps and associated data for each.
When connecting to many APIs, we get the same thing of "one request, loads of data"
I think we should break the connection between datapoint and polldeclaration. It is currently a one-to-one relationship. Polldeclaration is the "request to external party", but then something else to be triggered/linked to an incoming "document" or "dataset" to sort out what is what and send that off to Timeseries.
From the beginning the Mental Model has been "At a given time, poll one data point from the Internet and store the value".
We see more and more deviations from this over time, and I think it is time to challenge the validity of this approach.
I think we should break the connection between datapoint and polldeclaration. It is currently a one-to-one relationship. Polldeclaration is the "request to external party", but then something else to be triggered/linked to an incoming "document" or "dataset" to sort out what is what and send that off to Timeseries.