Add some helper text to user guide re what constitutes a rendition

[bquinn]

should go here: https://www.iptc.org/std/ninjs/userguide/#_renditions

we have some text in the G2 docs: https://www.iptc.org/std/NewsML-G2/2.34/specification/NewsML-G2-2.34-specification.html#newsitem-content

Alternative renditions of the news content, i.e. different technical representation of the same logical content, are wrapped by a contentSet wrapper element. Their order of appearance in contentSet is of no relevance. Their presence is optional: this allows for a lightweight and extensible representation of information.

Each rendition SHOULD by defined by a rendition attribute.

All alternative renditions SHOULD be derived from an original rendition by a software processor. For example: images in different resolutions, vector graphics and alternative bitmap images, text in different formats (e.g. NITF and PDF). The rendition from which all other renditions originate is indicated by the original attribute of contentSet; this attribute takes as a value the local identifier (id) of the original content component included in the contentSet.

The example we were discussing is an AI-generated summary video of a larger video - this should not be a rendition, it should be an association (with its own renditions if required).

Philippe's rule of thumb is that the metadata for all renditions should be the same. If the metadata doesn't match (eg the description is different or the people referenced are different), then it shouldn't be a rendition, it should be an association.