Summary

  • W3C's first public WOFF 1.0 Working Draft appeared on 27 July 2010. It packaged existing table-based fonts for use with CSS @font-face, optionally compressed their data and carried metadata. The WOFF 1.0 Recommendation followed on 13 December 2012.
  • WOFF2's first public draft appeared on 8 May 2014 and its Recommendation on 1 March 2018. Content-aware preprocessing and Brotli coding improved the packing of font data. The browser still unpacked a font for the web document that linked it.

The masthead looked exactly right on the designer's machine. On another computer it borrowed a substitute typeface and suddenly looked as if it had put on somebody else's spectacles. CSS @font-face offered a way to ask the browser to download a particular font. The request was only half a delivery plan. The font had to travel in a package that everyone at the handoff could recognise: its maker, the website and the browser. Even a beautiful letter needs a reliable address label.

The W3C's contemporary WOFF FAQ described the gap plainly. The CSS mechanism for linking a font had existed for years, while an interoperable font format remained unsettled. Browser implementers, type designers, foundries and web designers needed a common way to send the asset. This was not a fight over whether an a should have a curl at the end. It was a transport and packaging decision for the existing letters.

The first public WOFF 1.0 Working Draft, dated 27 July 2010, chose a practical suitcase. It wrapped the table-based structure used by TrueType, OpenType and Open Font Format files. Font tables could be compressed, and metadata or private-use data could travel alongside them in separate areas. The browser could decode the package and use the restored font with the webpage. The specification drew a useful boundary: selecting glyphs, redesigning them, or otherwise editing the source font was a different job from packing it.

That boundary mattered to the masthead. A type designer could supply a properly licensed font, a website could package it, and a browser could render it as intended. The suitcase carried the typeface; it did not sketch a new one during the journey. Optional metadata could tell a recipient about the font, including licensing information, without changing how the glyphs appeared. The package was a way of getting an existing design across the network with less friction.

The WOFF 1.0 Recommendation of 13 December 2012 made the format a finished W3C reference. It kept the link between the font and the document that requested it. A WOFF font delivered through @font-face was for that webpage's rendering, not a global desktop installation made available to every application. The browser had a parcel for a named recipient. It was not handing out copies of the typeface to every window in the building.

The first suitcase worked, but large type families could still make a journey expensive. The first public WOFF 2.0 Working Draft of 8 May 2014 sought denser packing. Instead of only applying the older compression approach to separate font tables, the new format used content-aware preprocessing and improved entropy coding. Its compressed font data used Brotli. The point was not to make a font look thinner; it was to reduce redundancy in the data that carries the same shapes.

The WOFF 2.0 Recommendation of 1 March 2018 set that approach in a final standard. Its text compares the method with WOFF 1.0's Flate compression and explains why preprocessing the font structure helps. The browser can reconstruct a usable font from the package, while details of table order and transforms may make the resulting bytes differ from the original file. A reader judging the masthead sees the typeface doing its job; a person comparing binary files must ask a narrower question about exact byte identity.

WOFF2 also kept the web-document boundary. A linked font serves the page that requested it. Browser rendering and system-wide font installation are different acts, with different reach. That scope gives publishers a way to set a publication's voice without silently turning a web download into a device-level font for every other app. It also puts the choice of typeface with the designer and the handling of the package with the browser.

There is a pleasingly unglamorous lesson in the dates. The 2010 draft and 2012 Recommendation settled the first common suitcase. The 2014 draft and 2018 Recommendation refined how tightly to pack it. The format did not ask the network to invent typography. It let a font that already existed travel to the page where its letters belonged. The masthead reached another computer looking like itself; the suitcase had done its quiet work and stayed out of the photograph.