There is no stable answer because “format” can name several layers at once. A container describes how parts are bundled. A codec describes how audio or image data is encoded. A profile narrows a standard for one setting. A version changes the rules over time. A preservation registry can list those separately, while a software company may call them one product family.
What can a registry tell us?
PRONOM, maintained by the UK National Archives, records documented file formats and the signatures and relationships that help identify them. It is valuable precisely because it makes its collection visible. It is not a census of every private application file, abandoned experiment, or undocumented database on a hard drive. A number without a boundary would create false precision.
Why count at all?
An inventory helps an archive decide which readers, converters, documentation, and emulation environments it needs. It can reveal that a collection depends on one obscure tool or that a supposedly common format has many incompatible profiles. Counting is useful when it leads to a preservation decision; it is trivia when it pretends that two names alone describe the risk.
What makes a count reproducible?
An archive should say what it included, when it counted, and whether variants were grouped. It should record identifiers and signatures so another team can check the boundary later. That discipline turns a number into a maintenance tool. A changing registry is not a contradiction; it is a record of new documentation and newly understood dependencies.
The practical inventory usually has several columns: format name, version, file signature, software dependency, documentation, and the actions a future reader must support. That is more useful than a grand total because it tells an archive where a failure would happen. A collection can contain a familiar extension and still be hard to preserve if its profile or codec is undocumented.
That inventory is also a map of institutional risk. A format used by one specialist application may deserve more attention than a common image format if the collection cannot be exported without losing meaning.
The long tail therefore cannot be reduced to a leaderboard of extensions. It is a map of dependencies: who can read a file, what information it carries, and which future actions would fail if its interpreter disappeared.