A file format is a convention for arranging information so software can interpret it. It may specify order, labels, numbers, compression, metadata, and relationships between parts. An extension such as `.pdf` is a name on the envelope: it can be missing, wrong, or shared by more than one variant.
What does a format preserve?
That depends on the object and the work that follows. A raster image preserves colour samples at positions in a grid. SVG preserves geometry, so a circle can be redrawn at a new size. MIDI preserves performance instructions rather than the sound of one recording. DICOM carries image data alongside information about orientation and physical spacing. A format is a decision about what should remain explicit.
Why is the extension not enough?
The bytes have to be read according to rules. A file called `report.pdf` can be renamed `report.mp3` without changing its contents; the new label does not teach an audio player how to interpret a document. Some containers also hold several kinds of data. A DOCX file is a package of related XML parts and media, not one uninterrupted stream of prose. The outside name is useful only because software has learned the convention behind it.
What is the reader agreeing to?
A reader is agreeing to interpret fields in a particular order and with particular meanings. If one system stores a duration in seconds and another assumes musical beats, both can parse the same number and still disagree. Format documentation therefore describes semantics as well as syntax. The hard part is not only opening the file; it is knowing what the opened values mean.
Where does the agreement live?
Sometimes the specification is public and stable; sometimes it is inferred from a working application. A parser may need a published schema, a byte signature, and rules for optional fields. Archives care about all three because a future reader must identify the file before it can interpret it. “It opens on my laptop” is evidence of one tool’s support, not proof that the format is understandable in general.
Formats are small languages. Their long tail exists because the world contains more kinds of work than a short list of extensions can describe. The practical question is always what a later reader needs to change, measure, search, or reproduce.