A fact lives in exactly one document, and every other document cites that one by id. A library kept this way can be corrected in one place, and a reader can always follow a claim to where it is held.
Four steps make the case: what a fact's home is, what goes wrong when a fact is copied instead, how much of the rule a machine can check, and how documents about the same thing share it without sharing sentences.
Step one
The rule
A fact is a definition, a measured number or a decision. Its home is the document whose genre holds that kind of fact: a concept for a definition, a result for a number, a journal entry for a decision. Every other page links the home and says what the fact means for its own reader, in its own words.
A definition
lives in a concept
Its defn holds the one sentence.
Others link it with class="defn-link", and hovering shows the definition.
A measured number
lives in a result
Its number field holds the value, with its evidence and the command that re-derives it.
Others cite the result by id in the sentence that uses the number.
A decision
lives in a journal entry
One dated entry of kind decision, permanent once committed.
Others link it, and the entry shows on every document it names in about.
A guide chapter needs the definition of a term that a concept already holds. What does the chapter do?
class="defn-link", so hovering shows the definition. It does not write the definition again.Step two
Why one home
Two copies of a fact agree on the day they are written. Then one is revised and the other is not, and the library holds two answers with nothing to say which is right. With one home, a correction lands once and reaches every page that cites it.
Below is one real decision in this library's history, and three pages that mention it. The pages are made up for the example. Pick a habit and see what the pages say after the decision.
The home: a journal entry
The journal became one short Markdown file per entry, permanent once committed.
Recorded as a decision: One journal entry per file.
Restated in three pages
- a guide chapter
Add your note to the journal file.
stale - an entry
Each journal entry is its own file.
revised - a map row
The log, one running document.
stale
Only the page someone remembered was revised. A reader now meets two answers and has no way to tell which is current.
Cited from one home
- a guide chapter
Record what you decided in the journal.
current - an entry
The journal keeps the dated record of decisions.
current - a map row
The journal, and why it is laid out as it is.
current
No page says how the journal is stored. Each links the journal concept, whose definition says it once, so the change happened in one place and none of the three was touched.
The rule also makes a library checkable. A number that cites its result can be followed in one click to its evidence and the command that re-derives it, as R-1 shows for the time this library takes to check.
Step three
What is checked and what is judged
Part of the rule is mechanical, and the gate checks it. The rest is judged by the agent, by reading.
The gate checks
- No two concepts define the same term.
- With the lab pack on, a number shaped like a measurement cites a result in the same sentence.
- Every citation resolves to the document it names.
The agent judges
- A paraphrase of a definition, written as if it were new.
- A decision restated as if this page made it.
- A number from another document, given as this page's own.
Why does a concept's definition cite nothing?
Step four
Siblings share ids, never sentences
A journal entry, a report and a paper about the same result each cite the result and reuse its figure. None is generated from another, because each has a different reader. What they share is the id, so they cannot disagree about the number.
How corrections work follows what happens when that number turns out wrong.