Revise recommendations around non-standard data models

[ldodds]

Section 4.4.3 and 4.5 include recommendations to use a standard data model and, where a feed isn't doing that, some recommendations about how to structure data to provide consistency, around e.g. data and URLs.

I propose that we revise the specification to do the following:

• Make the recommendation to use standard model more prominent (i.e. the default recommendation)
• Gather the other recommendations, including the discussion in 4.5, under an additional heading "Using non-standard models" to clarify that they apply to that scenario. People using the standard model can then ignore this.
• Make the kind property optional, as I think it's only there to help clients understand what might be in a feed item when there is a non-standard model. The property is under-specified as we don't indicate what values it might contain.

In addition, we can review the recommendations around dates, urls, etc and ensure these are addressed in the Modelling Opportunity Data specification.

[nickevansuk]

See https://github.com/openactive/realtime-paged-data-exchange/issues/86 for comments about kind.

My key concern would be to keep a good separation between RPDE and the modelling spec, as they are designed to work together, but we probably want to avoid them coalescing over time.

[ldodds]

I can see logic in trying to keep RPDE reasonably generic, but don't think we need to strive too much to create a totally generic harvesting protocol: we don't have any strong requirements around need to harvest other types of thing, in arbitrary models at the moment.

I would expect that as new requirements come up, we'd encourage people to adapt/extend the existing standard model.