Didier Verna writes:
A more important problem we have IMHO is the availability of
non-portable (system-specific) ways to customize the editor.
I don't really consider availability a problem, and on occasion I've
found it convenient to use X resources to prove to my own satisfaction
that XEmacs can do what Customize is steadfastly refusing to do for
However, certainly Lisp should be the first- and best-supported
channel for customization.
This also implies having a customization layer sufficiently
abstracted in lisp to satisfy the needs (hint: Xft, custom, face
inheritance etc... how well does that mix currently ?).
Well, that's not easy. Custom itself really only provides for a flat,
global namespace of variables. Groups are an afterthought, and
specifiers are not exploited by Custom at all.
Specifiers also have some deficiencies. Specifically, there should be
mode and extent locales. I believe that a mode locale would be more
efficient, robust, and intelligible than the current method of using
setter functions to "backpatch" all the relevant buffer locales for a
mode. And the extent locale would be a way to implement the "native"
mmm-mode you suggested earlier, I think.
The main problem with extent locales is deciding where they would fit
in the precedence hierarchy. (There isn't such a problem with mode
locale, unambiguously it would have a higher precedence than the
global locale, but lower than any buffer.)
XEmacs-Beta mailing list