When `unload-feature' is called, it unbinds all of the symbols provided by a package, and removes the feature symbol from the features list. When a `defvar' or `defcustom' variable symbol from a feature is `makunbound', does that remove values set by the user with `setq' or `customize'? If so, then this situation needs to be handled correctly by `unload-feature' somehow. I suppose that it also ought to handle unlinking DSO's as well. What about open processes or ports held by something in the feature? What other resources might be held that should be freed up? How can these things be accomplished? (YTMAWBK) This also implies that if a hook function is added to a hook as the function itself (using #') rather than as the symbol containing the function, it will remain in that hook list in it's original form. If the symbol whos function cell contains the function is put in the hook list instead, then it will refer to an autoload form. Therefore `add-hook' should always be called with a type `symbolp' for FUNCTION, rather than a type `functionp'. Is this correct? What about having an `unload-feature-hook'? When would it be run? I suppose that it obviously has to be run prior to unbinding the value and function cell of a symbol... Hmmm. Prior to unbinding any of them, not each time one is unbound. Has anyone ever tried to devise a package system for X?Emacs based on multiple obarrays? The idea has crossed my mind a number of times, but I have not thought it through very much. It would require changes to the evaluator, I think. What if an obarray had, like a keymap, a parent obarry? I suppose it would then need to have siblings as well... Hmmm. -- Those who do not study Lisp are doomed to reimplement it - Poorly. A few months in the laboratory often saves several hours at the library. mailto:karlheg＠debian.org (Karl M. Hegbloom)