On Tue, 2005-12-06 at 18:00 +0900, Stephen J. Turnbull wrote: Others on the python-mode(a)python.org list also help, as time allows. It's been a /long/ time since these issues were hashed out, but I'll try to give some history as best I can recall. There was a lot of debate about this many years ago, and Guido had very strong opinions about how he wanted it to work. I think that some of that was based on the way python-mode historically worked back when it was originally written, in, oh 1992 or thereabouts. If you look back in the cvs log, you'll see rev 2.32 1995/10/18 where I changed underscore to symbol class, and then 2.49 1996/07/03 where I reverted it back to word class "even though I don't agree with it". ISTR getting beat up pretty badly about it by users until it was changed back. We probably have had no complaints about the current behavior since then (until now, of course :), so I submit that python-mode users are pretty happy generally about the current behavior. IIRC there was some debate about the treatment of underscore w.r.t. user-facing commands vs. internal functions. It was brought up that internal parsing functions might want to treat underscore one way, while users might want to treat it a different way. AFAIK, there is still no 'good' way of handling this. I have a phantom memory that changing the class of underscore is going to break some parsing or font-locking. IMO, you can't change the current behavior without doing severe damage to the muscle memory of existing python-mode users. If you can provide a variable that switches underscore's class, while maintaining backward compatibility with the way it currently works /and/ make sure that nothing breaks, then I think that might be acceptable. -Barry