"Non-ancillary" is convoluted and imprecise

[fantasai]

Coming from the FO debate over the word "ancillary"...

Negating a relatively obscure word is a pretty awkward way of identifying the concept of "primary" or "essential". :)

Also, if the definition of ancillary we're using is "necessary support for the main work", then non-ancillary could mean either "the main thing" or "completely unrelated/unnecessary thing"...

---

Separately but relatedly, it's also awkward to use "ancillary" to categorize "uses" in two buckets (non-ancillary vs ancillary) and also for "APIs" in three buckets (non-ancillary, ancillary computed information, ancillary new information) where the use of "ancillary" in the second sense doesn't quite map to the first sense. (The "computed information APIs" are often used for non-ancillary uses.)

---

Suggestion:

• Keep ancillary uses, contrast it with primary rather than non-ancillary as needed.
• Change non-ancillary APIs to essential APIs as proposed in Rename "ancillary" to "supporting" and "non-ancillary" to "essential." #438.
• Change ancillary APIs computed from existing information to secondary or derivative APIs, since these are used for both primary and ancillary uses.
• Might also consider renaming ancillary APIs that provide new information to use a distinct word like auxiliary (so that the vocabulary is disjoint between use categories and API categories). Or maybe it's fine since this category really does map to ancillary uses.