Shared Normalization¶

Different source families enter the repository in very different shapes, but they still need a shared normalization discipline if they are going to appear together without collapsing into one shapeless export.

Normalization is the repository's way of translating unlike inputs into repository-owned evidence artifacts that can be compared, reviewed, and published responsibly. It is not a cosmetic formatting pass. It is the point where raw source material becomes something the repository can stand behind.

What Shared Normalization Means¶

Shared normalization does not force every source to have identical fields. It means the repository gives each family a predictable stage where the material is converted into repository-owned evidence artifacts with known roles.

That discipline is what makes it possible to compare:

source capture versus public publication
one domain against another
visible points against the narrower evidence files that support them

What Normalization Must Not Do¶

Normalization should make unlike families legible together, but it should not pretend that unlike evidence becomes identical evidence.

It must not:

flatten away source-specific limits
invent certainty that was not present upstream
hide the difference between context layers and direct sample-backed evidence

If it did those things, the repository would look cleaner while becoming less honest.

Why Readers Benefit¶

Readers usually notice normalization only when it is missing. Without it, a mixed repository turns into a pile of raw tables, ad hoc exports, and public pages that cannot be compared honestly.

With it, readers can move across source families and still understand what is direct evidence, what is context, what is sample-backed, and what is a broader framing layer.

Where The Repository-Owned Records Live¶

Normalized family outputs live in paths such as:

data/landclim/normalized/
data/neotoma/normalized/
data/sead/normalized/

The public publication side then builds from those repository-owned surfaces into atlas, regional, and country outputs.