favicon content-type sniffing

Looking at a hundred random icons, extracted from <link rel="icon">s in 
pages from dmoz.org, I see quite a lot of legitimate icons served with 
incorrect Content-Types:

text/html
  http://www.sandpphotos.co.uk/favicon.ico

text/plain
  http://www.ircle.com/favicon.ico
  http://www.jobe-industries.com/favicon.ico
  http://www.iss-software.com/images/favicon.ico
  http://www.estudi16.com/favicon.ico
  http://www.barcelonavacation.com/favicon.ico
  http://www.gargamel.jp/images/gargaweb.ico
  http://www.lotuscarclub.org/favicon.ico
  http://www.acdcrew.nl/favicon.ico
  http://www.newadvent.org/images/icon1.ico
  http://www.vespa-servizio.com/favicon.ico

text/plain; charset=ISO-8859-1
  http://www.dr-prager.info/favicon.ico
  http://merlin.vcpl.lib.fl.us/favicon.ico
  http://www.cc-paysfouesnantais.fr/design/standard/images/favicon.ico

text/plain; charset=ISO-8859-2
  http://www.mariaradio.ro/favicon.ico

text/plain; charset=UTF-8
  http://static.suite101.com/favicon.ico

There's also 53 image/x-icon, 4 image/vnd.microsoft.icon, 4 image/gif, 3 
image/png, and some redirects and 404s and server errors.

At least some of the image/x-icons are not Windows ICO files, e.g. 
http://www.frontweb.com/images/favicon.ico and 
http://www.vcpa.com.au/images/favicon.ico are JPEG.

Apparently HTML5 doesn't allow sniffing when these images are used in 
<link> (whereas it does for <img>). For compatibility with the web, that 
should probably be changed.

-- 
Philip Taylor
pjt47@cam.ac.uk

Received on Tuesday, 17 June 2008 22:52:15 UTC