Id: FEATURES,v 1.2 2004/01/23 18:56:42 vixie Exp

Features of ISC cron relative to BSD 4.[23] and SysV crons:

--      Environment variables can be set in each crontab.  SHELL, USER,
       LOGNAME, and HOME are set from the user's passwd entry; all except
       USER can be changed in the crontab.  PATH is especially useful to
       set there.  TZ can be set, but cron ignores it other than passing
       it on through to the commands it runs.  Format is

               variable=value

       Blanks surrounding the '=' will be eaten; other blanks in value are
       okay.  Leading or trailing blanks can be preserved by quoting, single
       or double quotes are okay, just so they match.

               PATH=.:/bin:/usr/bin
               SHELL=/bin/sh
               FOOBAR = this is a long blanky example

       Above, FOOBAR would get "this is a long blanky example" as its value.

       SHELL and HOME will be used when it's time to run a command; if
       you don't set them, HOME defaults to your /etc/passwd entry
       and SHELL defaults to /bin/sh.

       MAILTO, if set to the login name of a user on your system, will be the
       person that cron mails the output of commands in that crontab.  This is
       useful if you decide on BINMAIL when configuring cron.h, since binmail
       doesn't know anything about aliasing.

--      Weekdays can be specified by name.  Case is not significant, but only
       the first three letters should be specified.

--      Months can likewise be specified by name.  Three letters only.

--      Ranges and lists can be mixed.  Standard crons won't allow '1,3-5'.

--      Ranges can specify 'step' values.  '10-16/2' is like '10,12,14,16'.

--      Sunday is both day 0 and day 7 -- apparently BSD and ATT disagree
       about this.

--      Each user gets their own crontab file.  This is a win over BSD 4.2,
       where only root has one, and over BSD 4.3, where they made the crontab
       format incompatible and although the commands can be run by non-root
       uid's, root is still the only one who can edit the crontab file.  This
       feature mimics the SysV cron.

--      The 'crontab' command is loosely compatible with SysV, but has more
       options which just generally make more sense.  Running crontab with
       no arguments will print a cute little summary of the command syntax.

--      Comments and blank lines are allowed in the crontab file.  Comments
       must be on a line by themselves; leading whitespace is ignored, and
       a '#' introduces the comment.

--      (big win) If the `crontab' command changes anything in any crontab,
       the 'cron' daemon will reload all the tables before running the
       next iteration.  In some crons, you have to kill and restart the
       daemon whenever you change a crontab.  In other crons, the crontab
       file is reread and reparsed every minute even if it didn't change.

--      In order to support the automatic reload, the crontab files are not
       readable or writable except by 'crontab' or 'cron'.  This is not a
       problem, since 'crontab' will let you do pretty much whatever you
       want to your own crontab, or if you are root, to anybody's crontab.

--      If any output is generated by a command (on stdout OR stderr), it will
       be mailed to the owner of the crontab that contained the command (or
       MAILTO, see discussion of environment variables, above).  The headers
       of the mail message will include the command that was run, and a
       complete list of the environment that was passed to it, which will
       contain (at least) the USER (LOGNAME on SysV), HOME, and SHELL.

--      the dom/dow situation is odd.  '* * 1,15 * Sun' will run on the
       first and fifteenth AND every Sunday;  '* * * * Sun' will run *only*
       on Sundays;  '* * 1,15 * *' will run *only* the 1st and 15th.  this
       is why we keep 'e->dow_star' and 'e->dom_star'.  I didn't think up
       this behaviour; it's how cron has always worked but the documentation
       hasn't been very clear.  I have been told that some AT&T crons do not
       act this way and do the more reasonable thing, which is (IMHO) to "or"
       the various field-matches together.  In that sense this cron may not
       be completely similar to some AT&T crons.