Hi Steve. I agree that we could do even more to help new users. But as some silly philosopher once said, "perfect is the enemy of the good". Another philosopher said something like "patches welcome" Mind if I commit as is and we can improve it further with another JIRA if you want?
In response to your particular points:
Now, if you are really devious and the Server instance was told the port the HTTP server for that service was up on
This is problematic since some servers have lots of different HTTP endpoints, and also because the IPC portion of Hadoop is nicely divorced from the HTTP portion of Hadoop right now. Adding a dependence between them seems messy.
Incidentally, what if the client is an HTTP1.0 browser? Shouldn't the HTTP version be grabbed from the incoming request? Or is that getting too close to building a mini-web-server?
I tested the patch with curl and google chrome. Both seemed to work fine. Given that HTTP 1.1 was introduced in 1999, I don't know if any HTTP 1.0 browsers are even out there anymore. Again, if someone has an incompatible browser, they're still no worse off with the patch than they were before.