I had a stray thought about fragments in JSON documents, after a colleague asked about if we had considered using GraphQL to allow allow querying subsections of a larger document. It seemed like you could reasonably index into a JSON document and reference a subsection of it in a simpler fashion. I went looking for information about this and found RFC 6901 on JSON pointers, describing exactly what I was theorizing. However, it doesn’t appear to have been adopted yet. The gist is that you can index into a document with #/<key>. Little mystery solved, but while researching this I stumbled into something else more interesting.
I found this about best practices in designing fragment IDs when creating a media type. But it discusses these best practices in relation to several schemas that were out in the wild. This had the JSON pointer example, it also had XML and PDF examples, which seemed reasonable. But it also had plain text and CSV, which was interesting as the format i’s a line number and a character/field which works but I had never really considered attempting. The more interesting one was the SVG, since it lets you index into an image. You can index into a named section or any arbitrary view box.
This was an interesting diversion that really reminded me about the amazing flexibility and foresight of some of the technologies underlying the web.