Method and limits
How Atlas chooses what it shows.
Atlas is a transparent presentation system for published planetary data, not an independent scientific authority. Definitions come before values, and uncertainty, disagreement, provenance, and absence stay visible.
Public claim
No hidden authority.
No planetary scientist reviews this atlas. Values are selected from published sources by a documented, automated rule. The rule, its inputs, and every change to a published value are public. Corrections are welcome and are logged.
What the rule cannot detect
All available sources are wrong
The rule cannot detect whether every available source is wrong.
The definition misses your intent
The rule cannot detect whether a selected definition is misleading for a reader’s intent.
A newer source is missing
The rule cannot detect whether a newly published source has not yet entered the registry.
These are limits of the selection rule. Separately, read what Atlas’s checks cannot prove.
Definitions
Define the fact before choosing a value.
Candidates compete only when both their quantity and definition match. Values under different definitions are different facts, not conflicting measurements of one fact. Context values may explain an unknown, but they cannot silently fill it.
Atlas’s bounded human authorship is to:
- define what each page question means;
- select the comparable field or explicit groups;
- author both a confident answer and a disagreement answer; and
- declare executable assertions that must keep the answer true.
The automated resolver performs none of those editorial tasks.
Source policy
Authority depends on the quantity.
Atlas assigns a tier to each source for each quantity. A source can be authoritative for one kind of value without becoming authoritative for everything it publishes.
Tier A
Reference standards and constants
Physical constants, reference frames, rotational parameters, and reference parameters.
Typical sources: JPL Solar System Dynamics and IAU/WGCCRE reference reports.
Tier B
Mission and agency values
Mission-derived and agency-published environmental values.
Typical sources: NASA Planetary Data System archives, mission publications, and agency fact sheets.
Tier C
Peer-reviewed research
Quantities without an appropriate reference-standard or mission value.
Typical sources: Peer-reviewed literature.
Tier D
Identity and discovery
Identity, stable identifiers, and citation discovery only.
Typical sources: Wikipedia and Wikidata.
Tier D is never an admissible preferred numeric source. When it is the only candidate, Atlas publishes the value as not established.
Resolver order
How one published value is chosen.
The resolver is a pure build-time rule. Identical inputs produce byte-identical output in this order.
Step 1
Match the fact
Keep only candidates whose quantity and definition exactly match the metric.
Step 2
Apply source priority
Keep the highest admissible source tier present for that quantity.
Step 3
Break top-tier ties
Prefer the latest applicable epoch, then whether an uncertainty is stated, then the lowest uncertainty, then the stable source identifier.
Step 4
Compare the candidates
Compare every admissible candidate using its stored uncertainty or the quantity-specific agreement tolerance.
Step 5
Keep disagreement visible
When candidates disagree beyond tolerance, mark the value contested; retain and name the competing value.
Step 6
Protect the answer
When disagreement can change an answer, rank, or material comparison, mark it answer-critical and use the authored disagreement answer.
Resolution
Resolution states include honest absence.
- Settled
- Admissible values agree.
- Derived
- Calculated from published inputs.
- Contested
- Admissible values disagree.
- Not established
- No admissible value fits the definition.
Presentation policy
When Atlas refuses one clean answer.
Atlas refuses a single confident presentation in exactly these situations. These rules preserve disagreement, absence, and physical differences instead of smoothing them away.
Answer-critical disagreement
Atlas uses the authored disagreement answer and names the published values and sources that differ.
Definitions are not comparable
Atlas splits the comparison into labeled groups or excludes non-comparable rows. Surface and one-bar values share a rank only when the answer and every one-bar row name that reference level.
A ranked value is not established
Atlas ranks the known subset and visibly identifies every excluded unknown value.
A count changes with time
Atlas puts the epoch beside the value and qualifies any superlative with it.
There is no solid surface
Atlas forbids standing-on-the-surface language and names the pressure or reference level in the answer and row.
Corrections and dataset releases
Challenge the atlas in public.
A change to a published resolved value is a dataset release, not an invisible edit. Every value passport pre-fills its stable value ID, current value and source, page URL, and dataset version.
Public and permanent
Value corrections and general feedback become public, permanent GitHub issues.
A free account is required
Submitting either report requires a free GitHub account.
No response-time promise
Atlas makes no response-time promise.
A report may close
A report may close without action.
Corrections need evidence
Only value corrections ask the reporter to name a published source; general feedback does not.
Reports are welcome; no response-time promise is made.
Privacy
What Atlas does not collect.
The static site has no Atlas account, runtime server, or database. Reading a page does not send Atlas a usage event.
No analytics
Atlas performs no usage analytics.
No background third-party requests
Atlas makes no background cross-origin requests: no counter, beacon, error-reporting service, third-party font, or remotely loaded runtime asset. Outbound links navigate only after you activate them.
Local presentation only
The only browser storage Atlas permits is namespaced local presentation preferences; they are never measurement.
The web host is separate
Hosting providers may retain security logs that Atlas neither accesses nor uses.
Licensing and reuse
Authority is not permission.
A scientifically authoritative source may still restrict reuse, so Atlas records source priority and licensing separately.
Application code is licensed under the MIT License. Original Atlas prose and Atlas’s original selection, definitions, arrangement, and compiled dataset are licensed under CC BY 4.0 to the extent Atlas can license them. Public-domain facts and upstream material retain their own status or terms.
Unknown licensing blocks publication. Generated composites also exclude ShareAlike and NoDerivatives material unless a specific compatibility review records otherwise.
Moon data and sky geometry
A catalog entry is not a sky model.
The satellite catalog records 465 confirmed moons of eight planets and five recognized dwarf planets in its2026-09-09 snapshot. This catalog is independent of the twelve comparison worlds. Missing measurements stay unknown; moons of asteroids and other small bodies are outside this edition.
Sky viewpoints use validated geometric positions and rotation for the Moon, Europa and Charon during 2026 UTC. The same dated positions drive their local Orrery markers. Titan's Orrery path remains illustrative, so it has no surface viewpoint. These views do not model terrain, atmosphere, stars or eclipses.
Sky keeps a fixed observing site while time changes. Outside the covered year, Sky moves a requested date to the boundary with an explanation; Orrery keeps its date and hides the dated moons. Agreement with the pinned source models is a numerical check, not a claim of observational accuracy.
What the quality gate cannot prove
Passing checks is not the same as being right.
Passing the quality gate proves that Atlas behaves as specified. Passing it does not imply scientific peer review.
Comprehension
Automated checks cannot prove that an explanation gives every reader the intended understanding.
Physical fidelity
The two-body model can be implemented exactly and still omit atmosphere, rotation, terrain, oblateness, and third bodies that matter to real launches.
Interior knowledge
No gate can turn model-dependent or missing interior boundaries into observed facts.
Assistive technology
Automated accessibility rules cover only part of the browser and assistive-technology experience.
Source correctness
Spot checks can reveal pipeline errors, but they cannot certify every published number or its upstream source.
Asset interpretation
Consistent sphere portraits may imply contemporaneous appearance or natural color unless processing and epoch limitations remain inspectable.
Discoverability
The person building Atlas already knows where its controls are; automated checks cannot reproduce a first encounter.
Question quality
A consistently applied definition can still be a poor or misleading way to answer a reader’s question.