Esri gave the shapefile spec away and it became GIS's universal exchange format

In the 1990s every serious GIS vendor shipped its own proprietary file format, so moving a map layer between packages meant re-digitizing it or writing one-off converters. Esri's own ARC/INFO coverage format was powerful but closed. When Esri introduced the shapefile with ArcView 2 in the early 1990s, it made an unusual choice for a market leader: instead of guarding the format, it documented it completely.

A shapefile is not one file but a trio: .shp holds the geometry, .shx is an index of record offsets, and .dbf — a 1980s dBASE IV table — holds the attributes. The design was modest: no topology, a 2 GB ceiling, null values read as zeros. By rights it should have been an also-ran.

The July 1998 white paper 'ESRI Shapefile Technical Description' (J-7855) put the entire byte-level layout in the open, free to read and free to implement. Any programmer could now write a converter in a week. Open-source libraries adopted it, then every rival GIS, then every government data portal that needed to be readable by everyone.

Each new reader deepened the network effect. Interchange, which only has to be common rather than good, converged on the one format anyone could legally and cheaply support — Esri's. The technically superior alternatives stayed niche because supporting them was work with no payoff.

Publish the bytes, not the product: openness beats elegance

Implementation was free and unencumbered, so rivals and public agencies had no reason to invent their own interchange format.

Every reader and writer added network effects; once data portals assumed shapefiles, escaping them meant paying a tax nobody wanted to pay.

Interchange tolerates ugliness. The format's defects — dBASE IV attributes, 2 GB caps, no topology — mattered less than universality.

Esri won strategically: its native format became the pipe through which even competitors' users moved data, pulling them toward Esri tools.

A format wins on adoption cost, not merit. The alternatives were cleaner but undocumented or licensed, so nobody adopted them. Make yours free to implement and accept flaws as the price of ubiquity.

Nearly three decades on, the shapefile is still the most common way vector GIS data moves between organizations. The Open Geospatial Consortium standardized GeoPackage, a SQLite-based successor, in 2014 precisely to replace it, yet adoption remains partial — data portals, censuses and open-source tools still default to the file everyone can read. GDAL, the open-source data library most GIS runs on, maintains shapefile support as a core obligation. The format everyone loves to hate outlived the formats everyone was supposed to love.