A number of bugs, some minor (invalid HTML in reports)
and some major (broken web service) would have been
detected by a test suite for check.cgi, running over
HTTP. The tests could be run against a local
installation, to verify changes, or against a remote
deployment to ensure consistency.
Tests should reflect the relevant standards, but also
cover potential bad requests from real browsers.
At least, no requests should cause internal server
errors. Additionally, all content returned should be
valid HTML. As a further improvement, the validation
results should be checked against those generated locally.