Questions to address:

There are a lot of interesting little questions such as how Wget should deal with trying to parse IRIs from local files, whose character encoding may not be firmly established (the proposed "meta-database"/session info db would help toward this end, but would be outside the scope you're trying to deal with). And, should Wget deal with IRIs specified on the commandline, perhaps taking locale environment settings into account to interpret them? What about dealing with older URIs that contain non-ASCII byte values, but expect them to be interpreted directly (say, an ISO Latin-1 encoded URI): should Wget fall back to the literal byte values if the UTF-8 transcoded version results in a 404? (Please explicitly mention in your details section which of these questions you wish to implement solutions for, and which you don't.)