A datum is a role played by a proposition, as John Sowa said. A single
value is not a datum unless it represents a proposition.

Sure, but doesn't context determine role in this situation? Of course, 
said determination might be implicit rather than explicit which is 
ultimately inevitable. Thus, when _values_ are in the role of 
representing the description of an observation subject, they denotation 
"Data" applies. Likewise, if the aforementioned is achieved via a single 
value, then the denotation "Datum" would apply.

I think this confuses John's assertion that a "datum" is a "role" of a proposition. He takes the view that the proposition plays a role in the assertion that it is true, and that role is called "axiom" or "fact". I would just have said that a datum is a proposition that is taken to be, or asserted to be, true. The context for that role is any context in which the proposition is taken to be true.

If the context of appearance of the datum is in a report, and you trust the author of the report, then the datum/proposition is taken to be true. That context has nothing to do with the subject of the report, or the "subject" of the datum, or the the spreadsheet it appears on, or anything the like. The context is only the acceptance that the proposition is true.

The context of representation of a datum is "quite another thing entirely". I agree that a "value" may be considered to represent a datum, when the context of its appearance determines both the relation to which it is (in most cases) an argument, and the "subject", which is an/the other argument. So, in the context of my driver's license, the "value" T labeled date-of-birth is interpreted as the proposition (person.has.date-of-birth EdBarkmeyer T), and that proposition is taken to be true. (Some "values", such as "true" or "false" and "yes" or "no", are representations of assertions involving unary predicates. For example, the value "yes" in the blank following "US Citizen?" is taken as a representation of the assertion (person.isUScitizen EdBarkmeyer).)

So, in such a representation context, each value or tuple of values is in the role of representing one proposition about the subject. Assuming one respects the observer and takes these propositions to be true, each proposition represents an observation, and "data" is simply a set of observations. (I think Kingsley's term "description of" means "set of observations about".) That, unsurprisingly, pretty much coincides with the use of the term "data" in science and statistics.

Propositions (or data objects) are not restricted to triples (subject /
predicate / object, subject / verb / object, entity / attribute / value,
entity / key / value, ...). Some propositions naturally require higher
arity, e.g., X is 3 meters from Y.

Sure, but triples are an effective base upon which higher arity can be 
built, right?