Shouldn't `mml-parse' use `save-excursion'?

[Lloyd Zusman <ljz@asfast.com>]

In the lastest cvs version of gnus, I'm looking through the code
within `mml.el'.  One thing I noticed is that the `mml-parse' function
changes the buffer pointer without doing a `save-excursion'.  Here's
how `mml-parse' looks now:

  (defun mml-parse ()
    "Parse the current buffer as an MML document."
    (goto-char (point-min))
    (let ((table (syntax-table)))
      (unwind-protect
	  (progn
	    (set-syntax-table mml-syntax-table)
	    (mml-parse-1))
        (set-syntax-table table))))

However, I think it should actually look something like this:

  (defun mml-parse ()
    "Parse the current buffer as an MML document."
    (save-excursion
     (goto-char (point-min))
     (let ((table (syntax-table)))
       (unwind-protect
	   (progn
	     (set-syntax-table mml-syntax-table)
	     (mml-parse-1))
         (set-syntax-table table)))))

I'm thinking that `mml-parse' should probably used in more places
within gnus (I'll explain in a moment), and the `save-excursion' will
make it easier to use it without unwanted side effects.

The reason I would like to use `mml-parse', is because I'd like to fix
the `mml-secure-part' function in `mml-sec.el' so that it doesn't
insert duplicate tags, and so its functioning isn't dependent on the
current point's position within the message buffer when it gets
invoked (it is dependent now).  For this reason, I think it would be a
good idea to call `mml-parse' early within `mml-secure-part' so that
the results of `mml-parse' can be used to determine whether or not the
part that's being requested already exists in the message buffer
... if so, then `mml-secure-part' could be a no-op.

Anyway, whatever turns out to be good to do within `mml-secure-part',
it still seems to me that wrapping the logic of `mml-parse' within a
`save-excursion' will allow this very helpful function to be used in
more places within gnus.

Thoughts?

-- 
 Lloyd Zusman
 ljz@asfast.com