Hi Mosè 
   > Hi Uwe,
   > 2014-11-25 18:48 GMT+01:00 Uwe Brauer <oub@mat.ucm.es>:

   > I agree backward compatibility should be preserved as long as possible
   > (but not at any cost), but about what?  Most users only customize
   > variables, don't fiddle with functions, if they write some elisp we
   > hope they're also able to read the doc string of a function and see
   > which are its arguments, if they've been changed.  But please consider
   > the first version of a program (nor the second, the third, and so on)
   > is not perfect, when you develop it you arrive at a point in which you
   > must choose between keeping it bugged/broken, and fix it and break
   > compatibility (or fork it).

My point is the following: if you improve a function or variable by
adding more options it should be done, in my opinion, in way the user
has not to change his old settings.

   > Regarding the change to `LaTeX-label', the whole point of it was to
   > let users choose to which environments label should be inserted.  The
   > addition of the second argument was needed to discriminate between
   > environments and sections as suggested by Vladimir. 

I don't want to start this discussion again, since  I also use reftex,
my labels look typically 

  \label{rem:fixpoint-scheme:2}

Meaning that this  is the second remark in the file called fixpoint-scheme. For
me this is enough I wouldn't need to add more information like the one
concerning the section, but I understand there are users with other needs.


   > Defaulting `prefix' to an empty string when no type is provided (in
   > order to make this argument optional) would defeat the whole
   > purpose of the change. Only defaulting `prefix' to nil wouldn't
   > break old codes using `LateX-label' function, but keeping the
   > second argument mandatory helps users be aware of the change of the
   > syntax of `LaTeX-label'.

I had a look at the code and it is really a complete rewrite. From my
philosophical  point of view, the "appropriate" approach would have been
to leave the second argument optionally not mandatory, and a user
interested in this enhancement could consult the documentation and not
the other way around: that the long-term-user gets an error and presumes
a bug.

Something like this. 
(defun LaTeX-label (name &optional type)


   > Moral: I'm not going to change `LaTeX-label'.

Would you accept an (ugly) patch? (Also the changes that I do this any
time soon are  unlikely due to my workload and other priorities, such as
the xemacs pkg sync.)


regards

Uwe