A recent mail from Bjoern to our QA Interest Group mailing list reminded me the importance of the work on normative references in specifications; while I hope the Working Group will be able to address this topic in more details, especially given how recent history has shown how important this question is, I’ve decided first to work on a pragmatic approach, that had been on my todo list for quite some time (and others’ too, apparently): a tool to detect outdated references to our W3C Technical Reports; see for instance its results on the list of references of DOM Level 2.
As other tools in this category (e.g. the bibliography extractor), this tool relies on the data available through the TR Automation project, an XSLT style sheet that reads these data and compares them to the links in the given document, and the XSLT servlet to make an on-line tool out of it.
Now, what would be really cool (yet another project in my so called ‘someday pile’) would be to have all the tools available to check the quality of a document output in a common format (EARL being the most likely good candidate) so that you could create a meta-service that would allow you to pick and choose which checks you want to apply to a given document. The pubrules checker does already a bit of that to check the HTML and CSS Validity of the submitted document; the difficulty in generalizing this approach being more to make a proper Web interface (which might be hard due to fairly big number of reports you may end up getting if you add broken links reports and error spelling, for instance), and also to define a confidence level that the error needs fixing.