Notes on Assertions (for adding to the doc)

[metasj]

Meta: post a section at the bottom of the doc that's conversational. "so you want to..." and discussion

Idea: Some things to add to the doc

• fuzziness and disambiguation
• having a dial to turn from very precise to vague

Idea: we may want restricted fields. UUIDs: tricky if individuals self-report; they may not be UU. Perhaps a "UUID4" field should only be created by some supernodes.

Q: how to combine lots of assertions about one idea into a node?
TR: Say I post X=1000 w/ no prov, you post X=1400 w/prov, two others post 2500 and 2800 w/ different provs. How do you find /see all of those assertions? Say you query "things w/ type country and name Austria". You'd want all claims of that sort, perhaps to a few depth-levels.
--> nodes aren't real, they are just clusters of properties. (see also: shapefiles)

TR: Short term, what if we build a quick reduce function that looks at all assetions ever made about a topic. You have a db-layer of assertion-structure that determines what turns up.

Idea: Be aware of what unique identifiers are, and what properties are idents. It's incumbent on a [server/query-engine] to track this, to know when two assertions refer to the same thing and can be merged.

Aside: J-LD has language as first-class context [so all strings know their lang]

[joeltg]

Couple things that need to be fixed (noting mostly for myself):

• [x] Assertions need to be objects at the top-level, not arrays, so that they can be signed (the signature needs to be expressed in properties on a top-level object). This means the array that was previously described needs to live in the @graph: [...] property of this (potentially signed) top-level object.
• [x] prov:Entity objects should have a prov:value that references a named graph instead of an asserted object directly. This gets messy quickly (lots of nested @graphs) but it's important since prov:Entity should reference the assertion-as-a-digital-object, which must have an id distinct from the objects within the assertion.
• [ ] @ids in JSON-LD can't be @index maps, so we need some other way to "append" properties to existing nodes. A weak version of this is just using http://schema.org/sameAs or equivalent, but would love a more formal scheme.