Re: new interfaces io_close, io_reidentify, file_record_lock

Hrm, I just had a flash about the pid analog to the auth server.

Why not *make* this auth ids?  Here's how it would work:

First, we specify a range of auth ids which indicate "possession of
task N".  (Just flip one of the high bits.)  There is no problem here,
because nobody is using giant ids.

The proc server can give you the necessary auth handle anytime you
like.  We would add a relevant call.  (It's root, so it can create it
at will, and it knows the true pid, so it can restrict access
rightly.)  You should be able to get the "pid" auth id given the task
port or the process_t control port for any task.

To demonstrate you are pid foo, you just provide your auth handle in a
normal auth transaction, and the server looks for that pid.

Thomas