What I wish, is that despite the well-meaning intentions of folks like Sanjiva and Clemens (and even Marc Hadley, from a couple of months ago), that they’d step back and look at what they’re proposing from a developer POV.
Developers want a library that assumes the same things they assume. Forcing them to be explicit about those assumptions is the quickest way to turn them off, IME.
Consider: java.net’s support for HTTP and URIs, for all its problems, is still able to turn a URI into data – the most magical of Webish incantations – in about three lines of code. If your solution can’t match (or beat!) that, I’d be heading back to the drawing board.
And +1 to what Stefan said.