A bit of an aside post. I’ve been playing with extracting images from WARC files so I can try to play with some computer vision techniques on them. One part of extracting images was that I needed to know which file extensions to look for. While I began with the usual suspects (initially JP*Gs), after some crowdsourcing and being pointed towards Andy Jackson’s “Formats Over Time: Exploring UK Web History” I came up with a good list of extensions to extract.

These are from the 2011 Wide Web Scrape web archive.

The following regular expression, '.+\.TLD/.+(\.gif|\.jpg|\.tif|\.jpeg|\.tiff|\.png|\.jp2|\.j2k|\.bmp|\.pict|\.wmf|\.emf|\.ico|\.xbm)' where TLD represents the top-level domain I am searching for (i.e. CA or COM) helped find the extensions that I am looking for. A variation of this script extracted images and numbered them, preserving duplicates and identically-named files.

The findings were interesting in terms of what top-level domain contained what. I’ll paste my findings below in case they’re interesting. It can help you know what to look for, and gives a sense of the web c. 2011. The next step is to analyze these images, of course..

Table One: Absolute Appearances of File Formats Across Samples of Various Top-Level Domains, Wide Web Scrape.

Table Two: Relative Frequency of File Formats Across Samples of Various Top-Level Domains, Wide Web Scrape