When you write a Micropub implementation the primary spec involves an unholy JSON format. About 80% of the code I wrote was just dedicated to finding out wheth

When you write a Micropub implementation the primary spec involves an unholy JSON format. About 80% of the code I wrote was just dedicated to finding out whether or not I'm updating an article, creating a new one, deleting the article entirely and which attributes need updating. The JSON format has an immense amount of issues relating to irregularity, some fields have html and text sub-attributes but this doesn't seem documented properly anywhere, not even the MF2 page, everything is an array even where it makes 0 sense to do so and as an implementer I'm somehow expected to then process and store this mess as a coherent article. Part of the spec validation involves storing a publication which contains almost no text and some user defined elements for some reason. And I'm expected to understand and properly update these custom elements if requested.