Rubric 1.0.0 · column 12 of 13
Installed fonts
How many of 78 probed font families this browser can render, measured by text metrics rather than by any enumeration API.
Under rubric 1.0.0 this column is worth between 1.74 and 5.64 bits, and the likeliest outcome the table knows — 28 of 78 — costs 1.74. Those figures are assumptions printed before your run. The counted box beside them is not: it is how many of the devices these sites have read share your value.
How a browser produces a font count
There is no API that lists installed fonts to a page. The count is produced by measurement: a string is rendered in a candidate family with a known fallback, and the width and height of the resulting text box are compared against the fallback's. A difference means the system resolved the family; an identical box means it fell back. Seventy-eight families are probed this way, so the value is a count of successful measurements, and every one of them is a real typesetting decision made by the operating system.
How rubric 1.0.0 prices a font count
Five buckets over the count, and they are not monotonic. 21 to 35 detected is assumed at 0.30 and costs 1.74 bits — the quietest bucket, and the one a stock install tends to land in. 1 to 20 is 0.20 and 2.32 bits, 36 to 50 is 0.12 and 3.06 bits, and 51 or more is 0.03 and 5.06 bits. Detecting none at all is 0.02 and 5.64 bits, which is more expensive than detecting most of them.
The spread between the cheapest and the dearest outcome is 3.91 bits, and that spread is the whole of what this column can ever do to a total. Where the four band edges sit →
Every outcome this rule can produce
This is the printed envelope, written out. The pen cannot land outside it, because these are
all the answers rule fonts has.
| Outcome | Assumed p | −log₂p |
|---|---|---|
| 28 of 78 21–35 detected modal outcome — the printed line | 0.300 | 1.74 |
| 14 of 78 1–20 detected | 0.200 | 2.32 |
| 44 of 78 36–50 detected | 0.120 | 3.06 |
| 62 of 78 51–78 detected | 0.030 | 5.06 |
| none detected | 0.020 | 5.64 |
What changing a font count costs
This is the hardest column on the form to move deliberately, because there is no list to rewrite — every entry is a rendering result. Blocking font measurement drops you to none detected, which costs 5.64 bits, more than the 51-plus bucket you were probably trying to leave. Uninstalling fonts is a real change to the machine with real consequences for everything else it does, which is why fonts is the sole member of its group and always counts in full.
The rubric's own reasoning, verbatim
Detected count from a fixed 78-name probe list. A stock OS lands in a narrow band; a design or development machine sits far outside it.
What moves it
Installing or removing fonts. Design and development machines carry far more than a stock install, which is why this signal separates professionals from everyone else so cleanly.
Where this prior is weakest
A count throws away which fonts. Two machines with 44 detected fonts can share almost none of them, and the rubric prices them identically. The set would be far more identifying than the count — and far more invasive to publish.
Correlation
The only member of the "typography" group, so it always takes full weight. That is worth knowing when you compare it against a grouped signal: no discount is protecting the total from it.
How the correlation discount works →
Challenge it
If you have a source that contradicts this assumption, the honest thing is to say so in
public against this specific prior. Open a challenge
against fonts →
What people ask about a font count
Answered from rubric 1.0.0 itself. Where an answer quotes a probability, that probability is a constant in the shipped scoring code and is printed in the table above.
- Why is detecting no fonts worse than detecting fifty?
- Because the rule assumes zero detections at 0.02 and 51-plus at 0.03, so none detected costs 5.64 bits against 5.06. A browser that renders nothing measurable is treated as the odder machine, and blocking the probe lands you there.
- How are installed fonts detected without an API?
- By measuring text. A string is typeset in each candidate family against a known fallback and the resulting box dimensions are compared. A different box means the family resolved. There is no enumeration call to intercept.
- Why does a design machine score higher here?
- Because it resolves more of the 78 probed families, which puts it in the 51-plus bucket at 0.03 and 5.06 bits instead of the stock bucket at 0.30 and 1.74 bits. The rule prices the count, never which families were found.
Where to read next
- The complete prior table for all 13 columns — every rule, every outcome, and the rationale the rubric ships with each one.
- Assumed and counted: what a prior is, what the counted column beside it measures, and why the two are never added.
- Compare two passes for the delta on this column — change one setting and read what actually moved, instead of guessing from a total.
Where the pens move
Knowing which prior is weak is only useful if something can actually move the signal.
Engine level, not JS level. Donut Browser spoofs through CDP injection in its Wayfern Chromium fork, so the value a page reads is the value the engine reports — there is no patched accessor to catch.
It refuses to contradict itself. A profile whose proxy exit geography, time zone or language disagrees with its fingerprint does not launch. That is a pre-launch gate in the app, not advice in a manual.
Open, and quiet. AGPL-3.0, desktop app and sync server both in the repository, and no analytics SDK anywhere in the source. macOS, Windows, Linux.