I think Alison Taylor's video is excellent and that it has a great deal of relevance to RM and to RM users.
Among many other thoughts I had while watching the video, I was reminded of the change that happened when RM added the ability to tag media to other objects such as people, places, citations, etc.. Previous to the change you started with the object and linked to the media. After the change, you could still start with the object and link to the media and you also could also go the other direction - starting with the media and tagging to the object.
Previous to the change in RM, each object's link to a particular media file could have its own caption, and two different objects that linked to the same media file could have different captions for the same media file. For example, for a photo of two brothers, John Doe's caption could say "John Doe and his younger brother William" and William Doe's caption could say "William Doe and his older brother John". After the change, the RM caption essentially became associated only with the media file and therefore each media file could have only one caption in RM. I was pretty bummed out by the change because I had spent hundreds of hours captioning media files with multiple captions. I lost all my captioning work with no hope of getting it back.
When I expressed my concerns on these forums, I received basically no support. I'm going from memory here, but other forum members expressed the view that a caption for a photograph really should be associated only with the photograph itself and not with the context in which the photograph appears. And if the caption for the photograph really should be associated only with the photograph itself without regard to context, then the photograph really should just have one caption.
That brings us back to Alison Taylor's video. She talked about a number of ways to "caption" a photo, including some very low tech ways. For example, you can hand write your captioning information on a piece of paper and scan in your piece of paper along with the photo. But she also talked about using the IPTC data that's actually a part of the media file. In looking at IPTC data, it seems pretty apparent to me that it follows the "one caption per photo" model. And yet it still seems to me that different captions in different contexts make sense. Newspapers frequently print a file photograph of a newsworthy person that they have printed before, but for a new story the file photograph may be printed with a different caption than it was the previous time.
The alternative to having captioning information to be stored in the media file itself is to have a database system where the captions are stored in the database. In this environment, it would certainly be possible to have multiple captions. Whether real world software that manages photos has multi-captioning ability I don't know. But I do think that some of them have the ability to include pixel coordinates that for example would isolate one face out of a group photograph without having to make separate copies of the photograph for each face that is thus isolated. Surely under these circumstances, there would have to be a "caption" of some sort associated with each set of pixel coordinates for each face? I don't really know how these systems work.
That brings us back to RM. The RM database seems to have unused data fields for such things as pixel coordinates. Perhaps this was planning for the future? I don't know, and such a feature may or may not ever appear in RM. But if ever it did, then you would think that each "sub-image" created electronically with pixel coordinates would need to have its own sub-caption. If so, then that would restore to RM the ability to caption by context, not just caption on the basis of one caption per photo.