These patches add a new option to ls(1), du(1) and df(1)
that enables sizes to be printed with thousands separators,
which is often much more human-readable than the so-called
"human-readable" output produced by the -h option.

Sample output from ls(1):

   -rw-r--r--  1 foo  bar    151,950,474 Oct  4 16:22 b1.img.gz
   -rwxr-xr-x  1 foo  bar          3,543 Oct  4 19:16 mirror.sh
   -rw-r--r--  1 foo  bar      1,249,240 Oct  5 13:22 t1.img.find
   -rw-r--r--  1 foo  bar  6,291,456,000 Oct  4 19:05 t1.img
   -rw-r--r--  1 foo  bar    149,764,301 Oct  4 19:23 t1.img.gz
   -rw-r--r--  1 foo  bar      2,083,980 Oct  5 13:22 t1.img.ls-alR

Sample output from du(1):

   11,828  /home/alfred
   470,344 /home/berta
   866,424 /home/chris
   9,808   /home/donald
   112     /home/emma
   1,358,516       total

Sample output from df(1):

   Filesystem    1K-blocks      Used       Avail Capacity  Mounted on
   /dev/ad0s1a   1,012,974    62,418     869,520     7%    /
   /dev/ad0s1d  10,154,158    31,424   9,310,402     0%    /var
   /dev/ad0s1e  10,154,158 1,544,566   7,797,260    17%    /usr
   /dev/ad0s1f 128,009,212 1,358,518 116,409,958     1%    /home
   /dev/md0        297,326        20     273,520     0%    /tmp

The thousands separators are enabled with a new option
"-," (a comma).  I'm aware that this is against style(9)
and it will be changed (see discussion below).  Using
the comma is just temporary.  The "-," and -h options
are mutually exclusive; they override each other so
shell aliasing works as expected.

It is expected that users create shell aliases or shell
functions for their convenience that contain the option,
for example:
   alias ll="ls -l,"
or:
   ll() { ls -l, "$@" | less -EM; }

The code uses the separation character and grouping
specification accoding to the current locale set by
the user (LC_NUMERIC, LC_ALL or LANG, respectively).
If the locale does not specify thousands separators,
but the user still requests them by specifying the
option, then a comma will be enforced to be used as
the separator, unless the locale already specifies
the comma as the decimal point, in this case a period
will be used.  If the locale doesn't specify grouping,
then groups of three digits will be separated.

These patches apply to RELENG_6 and RELENG_5 as of
2006-11-22, both of which have been tested.

The patches for ls(1) and du(1) apply cleanly to HEAD
(7-current), but the patch for df(1) requires minor
modifications because of the recent addition of the
"kflag" variable.  Other than that, all patches work
fine on HEAD.

The following files are patched:

df.patch (relative to src/bin/df):
     df.c
du.patch (relative to src/usr.bin/du):
     du.c
ls.patch (relative to src/bin/ls):
     ls.c
     ls.h
     print.c

TODO:

-1- Using "," as an option character is against style(9)
    (even though it is easy to remember and easy to type).
    Select a suitable option letter, preferably the same
    in all tools.  In ls(1) the following letters are
    currently unused:  -DEJKMNOQVXYejvyz
    All of them happen to be unused in du(1) and df(1),
    too, so using one of them would be OK.  Maybe -D
    ("Digit separators") would be a good choice.

-2- Add the new option to the Usage output of the tools.

-3- Add the new option to the manual pages.

-4- Consider adding the common code to libutil (that's
    also where humanize_number(3) already lives).

-5- Maybe it could be integrated somehow in the printf(3)
    code.  I have no idea how, though.

-6- Add the feature to more tools.  For example, it might
    be useful with "netstat -i", "ipfw show" and others.

--