Hello,

An Elisp program communicating with a subprocess is typically written
as:

  (let* ((process-connection-type nil)
         (process (start-process "cat" nil "cat" "/etc/motd")))
    (set-process-filter process #'my-filter)
    (set-process-sentinel process #'my-sentinel)
    ...)

It is a bit tricky and error prone.  On the other hand, Python and GLib
provide a simple way to spawn and communicate with subprocess:

http://legacy.python.org/dev/peps/pep-0324/
https://developer.gnome.org/gio/unstable/GSubprocess.html

So I would like to propose something similar:

* Provide a function 'make-subprocess' similar to 'make-network-process'
  or 'make-serial-process', which can be used as:

  (make-subprocess :name "cat" :program "cat" :args '("/etc/motd")
                   :filter #'my-filter :sentinel #'my-sentinel
                   :coding '(utf-8 . utf-8))

* Rewrite 'start-process' as an Elisp wrapper around 'make-subprocess'.

This also has the following benefits:

* We could collect stderr output naturally.  'make-subprocess' could
  have a keyword, say :error, to prepare a pipe for stderr when spawning
  a process.

* Maybe it could share the same code to handle the standard keywords
  (:coding, :filter, :sentinel, etc.) with 'make-network-process' and
  'make-serial-process'.

What do people think?  I'm attaching a patch as POC.