"Stephen J. Turnbull" <turnbull(a)sk.tsukuba.ac.jp&gt; writes: Hrvoje> Speed is definitely an issue with large English man pages. Hrvoje> First, nroff takes a while to process them, and Emacs Hrvoje> takes time to clean them. This is less dire on newer Hrvoje> machines, but man pages really do get that big (try `man Hrvoje> zshall' on a slower computer.) OK, we don't want to do full-buffer regexp substitutions often, and probably man.el does a lot of them. Yoshiki> I'm reluctant to do the same thing for Chinese and Korean Yoshiki> characters as I don't know how they localize man pages. Hrvoje> Ideally, XEmacs should just do the right thing, either by Hrvoje> hardcoding the stuff or by guessing it. We should Hrvoje> generally be careful about introducing new user-options. We don't have the choice of "just working" here. We don't have Chinese or Koreans to ask at the moment, and we don't know whether there might be other languages that make the bogus width distinction that Japanese does. The defcustom (1) gives Japanese users who know that they'll be looking mostly at non-Japanese pages the chance to defeat what you say is an expensive operation, (2) allows non-Japanese whose situations we don't know yet to make the appropriate choice for themselves (including the fact that custom is self-documenting, and so they might actually find it while looking in custom for some option to fix bad behavior), and (3) documents what we did for the future day when the *roff folks remove the bogus width distinction. Unfortunately, I18N is going to mean dealing with bad choices like this one that have become standard practice in some locales, which we can't possibly know ahead of time precisely because they're so bogus. One possibility would be to make a "compatibility" group where we could put user options (like this one) that we would like to guess correctly, but can't be sure of doing so. -- University of Tsukuba Tennodai 1-1-1 Tsukuba 305-8573 JAPAN Institute of Policy and Planning Sciences Tel/fax: +81 (298) 53-5091 _________________ _________________ _________________ _________________ What are those straight lines for? "XEmacs rules."

Yoshiki> So you are proposing to change regexp to Yoshiki> \\(.\\)\\(\b\b\\1\\) and add an option to disable it? Yoshiki> I'd prefer to use \\(\\cj\\|\\ck\\|\\cc\\)\\(\b\b\\1\\) Yoshiki> if Chinese and Korean man pages are also broken. No. I'm proposing to use a defcustom, default to the one you're using (you've convinced me it is gauranteed not to introduce new brokenness), and document so that Chinese and Koreans have an easy interface. Something like (just guesses for the prefix and the name) (defcustom manual-bold-twocolumn-compatibility-regexp "\\cj\\(\b\b\\1\\)" "Matches bold two-column character. For compatibility with Japanese nroff, which puts one backspace in for each nominal column to create a bold character. Ie, a bold kanji looks like "K^H^Hk". Other multibyte languages can be added to the regexp if necessary by replacing the "\cj" with a group of alternative character classes (such as "\ck" for Korean and "\cc" for Chinese). This option may go away if it becomes clear how to do the right thing automatically without it." :group 'manual) Hrvoje is probably right, this should be a defvar or defconst instead of a defcustom. But putting in a declaration will help document it. -- University of Tsukuba Tennodai 1-1-1 Tsukuba 305-8573 JAPAN Institute of Policy and Planning Sciences Tel/fax: +81 (298) 53-5091 _________________ _________________ _________________ _________________ What are those straight lines for? "XEmacs rules."