2022-06-28-.ipynb

thanks for the message @agoose77 . i like where you are going with these thoughts. stoked to see folks pushing this topic.

i definitely agree that the notebook is a rich structured document, but i feel that minimizes the polyglot nature. my interpretation is notebooks are "hypermedia collage".

what is funny is we probably agree, but I actually am of the mind that the notebook format is too permissive. my unique experience with pidgy has demonstrated when you elevated markdown to the primary language you naturally lose the ternary of raw, code, markdown cell conventions. instead cells become a binary of ON and OFF cells.

i could see something added to the schema that makes things statement like:

{
  ...
  "if": {
    "cell_type": {"const": "code"}
  },
  "then": {  
     "properties": {"outputs": {"type": "array"}}
  },
  "else": {
     "properties": {"outputs": False}
}
  ...
}

i think y'all might have the best case to make if you can keep the vocabulary near to the existing vocab with minimal changes. the more than can be reused the better. the new json schema approaches are more composable and could allow us to maintain backwards compatibility while evolving the existing schema.

schema vs context

its going to be hard to add a lot of contextual information to a schema. that is not really the goal. schema impose structural consistency, not contextual information. for that tools like json-ld may be a preferred standard for described the context of the markdown in a given document.

further i would suggest looking at the ietf rfcs as there are some recommendations in dealing with markdown variants https://www.rfc-editor.org/rfc/rfc7764.html