This site will look much better in a browser that supports web standards, but it is accessible to any browser or Internet device.
The flaw in sudo is that a failure in step #2 was not treated as a fatal error.
The only installations that are affected are those that have configured sudo to link directly with the Kerberos 5 libraries and where Kerberos has beeen configured on the machine. Only hosts that are operating as a Kerberos 5 client only (and not a server or slave) are affected. Kerberos clients do not have a Kerberos 5 keytab file present on the local machine.
Exploiting the flaw requires that a malicious user be able generate a bogus KDC response, thus tricking sudo into believing that authentication was successful. Since the user almost certainly had to authenticate in order to login in the first place, there is little impact unless sudo has been configured to use a password other than the user's for authentication. Even so, sudo will still only allow commands to be run that the user is authorized for.
Much has been made on Bugtraq about MIT Kerberos 5's use of the KRB5_KTNAME environment variable. This has no impact on sudo since the authentication routines (including Kerberos 5) are run with a zeroed out copy of the environment.