[Openid-specs-ab] User input normalization and parsing

[Justin Richer]

I've been working on the webfinger handler for our client software, and 
I am having a terrible time trying to get the actual parsing rules 
straight and provide consistent output with the example input values. 
Specifically, when using the Regex in Appendix B of RFC3986 (and several 
derivatives such as that used by Spring's UriComponentsBuilder), the 
"path" component seems to eat things that it shouldn't. For instance, 
with the input "joe at example.com", I get the entire string put into the 
"path" component and everything else null. Similar thing happens with 
the "acct:juliet%40capulet.example at shoppingsite.example.com" example, 
with the "acct" being parsed as the scheme and everything else getting 
dumped into the path.

How is everyone else parsing user input? Are you able to follow all of 
the input parsing and normalization rules described in the discovery 
document? And if you're working in Java, can you point me at your code 
or the library that you're using to do it? (Note that Java's built-in 
URI parser falls over for other reasons.)

  -- Justin