(Note: this file has been re-arranged to be in reverse chronological
order, which is The Right Thing for ChangeLogs - DLC)

March 05, 2004 (fortune-mod-1.99.1)

	Most of the changes have occured at some point in time in the last
	5 years. 

	A high number of spelling, punctuation, formatting and grammar
	fixes.

	Internationalisation support.

	New -c option to see which file a fortune came from.	
	
August 29, 1997

  Incorporated a couple of minor changes made in the (old) Debian
  fortune-mod package, including the addition of an extra data file
  called 'cookie'.  Renamed some documentation files, and included some
  install information for non-Linux users.  Added a "-v" option to
  report the program version.

  I intend to submit this distribution to SunSITE RSN.

    -- Dennis L. Clark <dbugger@progsoc.uts.edu.au>

May, 1997

  This release fixes many of the portability problems with the
  fortune-mod program released by Amy Lewis in October, 1995.  The
  previous version had many Linux-isms in it, which left it unworkable
  on any other platform.  This version replaces most of these with more
  standard calls, making it more likely to work under other platforms.
  The Makefiles have been modified so that GNU's gcc and make are no
  longer required: any standard make and ANSI-compatible C compiler
  should work.  Sorry, pre-ANSI compilers are not supported (c'mon, this
  is the 90's, darn it!)

  This version has been tested to work on SunOS 4.1.x as well as Linux.
  All changes made to the sources were as platform-independant as
  possible.  Therefore, no "#ifdef LINUX" or "#ifdef SUNOS4" directives
  appear in the code.  An effect of this is that a number of 'implicit
  declaration' warnings are emitted by gcc under SunOS 4.1.x, but this a
  problem with SunOS's standard headers, not with the program or the
  compiler.  The benefit of this approach is that it eases the work of
  expanding the port to include other platforms. Bug reports and fixes
  for other platforms are most welcome!

  A few "standard" C function calls were replaced with more standard
  counterparts at various points.  Generally, when there was a choice
  between a BSD version of a function and a POSIX version, the POSIX
  version was favoured (even though fortune originated on BSD).  An
  exception to this was the regex functions: either POSIX or BSD
  versions can be used, with selection made via the top-level Makefile.

  Fortune and strfile also compiles on Solaris 2.5, but a discrepency
  between the declaration and implementation of 'struct dirent' on the
  test platform caused fortune to execute incorrectly there.  It is not
  certain whether this is a bug in the header file, the C library, or
  the test platform.

  While the making of this release was not meant to become a bug
  search-and-destroy mission, some bugs were inadvertantly discovered
  and fixed. including the known bug of using -a with a file that occurs
  in both the offensive and inoffensive directories.  Fixing this
  particular bug required a way to be able to seperately identify two
  fortune files with the same name, with one in the inoffensive
  directory, the other one in the offensive directory.  Now, such a name
  will be taken to be the inoffensive file by default.  However, you can
  now append '-o' to a fortune name, and the '-o' will be removed and
  the offensive directory will be searched.  Thus you can say (assuming
  you use the distributed datfiles):

        fortune 80% politics politics-o

  Which has an 80% change of giving an "inoffensive" political fortune,
  and a 20% change of giving an "offensive" political one.  Note that
  this makes fortune-mod backwards-compatible with BSD fortune, but only
  for users, not fortune database maintainers.

  Of course, this solution only passes the buck: with the above example
  again, if you have a 'politics-o' file in your inoffensive directory,
  you are back to square one.  OTOH, seeing that '-o' originally was
  meant for offensive fortunes, using it for inoffensive ones is simply
  asking for trouble.

  Bug fixed: Fortune's definition of a fortune length (for -s and -l)
  was inconsistant.  Unordered fortunes counted the 2 delimiting
  characters (as hinted in the man page), but sorted or randomized
  fortunes did not.  Now the delimiting characters are _NEVER_ counted,
  so you will always get the length limit you expect.

  Another bug fix: -l and -s can now work together with -m.  Previously
  -l and -s were ignored when -m was in effect.  The new behaviour helps
  me count how many long or short fortunes there are in a file.

  Ansify has been removed from the package, as well as some filter
  scripts from NetBSD that no longer appear useful.  Randstr has been
  kept, but has not been improved in any way.  It at least has a man
  page: maybe somebody will find it useful.

  Some of the documentation (including the man pages) has been improved
  and updated, and some files have been renamed so that the package
  looks less Linux-specific.

    -- Dennis L. Clark <dbugger@progsoc.uts.edu.au>

Late October 1995
  Ansify has been abandoned.  I'm going to distribute this working version
  of fortune, and then see how difficult it would be to add termcap/terminfo
  enhancements to fortune itself--I don't anticipate serious problems, but
  I'd rather go ahead and get this on the net.
  
  A last-minute change was made to the way that percentages are displayed
  with -f; it is now in the format nnn.nn%.  The reason for this is that
  with the multiplication of small files, fortune -af displayed a large
  number of "0%"s--no worse than the old version, but not helpful.

  The fortunes database was finally cleaned up, and this version is now
  being distributed (at least, I hope it is).  I don't consider the current
  division of fortunes among files absolutely canonical; some are certainly
  in the wrong places.  But things are *better*.

Amy A. Lewis	alewis@email.unc.edu

Mid-October '95
  Another utility, ansify, now compiles; it has not been tested at all, so
  it may not work even slightly.
  
  Ansify is a rather stupid program, all things considered, but the work on
  it does raise an interesting possibility for an enhanced fortune.  At
  present, the fortune databases contain x^Hy sequences for underline and
  special characters (and this can be extended to include bold, = x^Hx).
  Ansify is stupid because it doesn't use the proper tools, ie termcap or
  terminfo (hmmm ... since it works on files, that may not be so stupid);
  it appears that if that can be done, then a termcap/terminfo enhanced
  fortune could be produced, which would recognize the existence of ^H in
  a string and attempt to display using appropriate control sequences.
  This sort of modification would be of greater interest to casual users,
  I think, than even the bug fixes, and since it would not force changes
  in the storage of fortunes, it is eminently portable.  Consider this a
  'todo' announcement.

Added (early October 95):
  A new executable, rot, which is a rot13 filter (a caesar cipher). Most
  probably have caesar, but on the other hand, if you compile this mess
  as root, caesar probably isn't in the path.

  A new parameter to fortune, -n, which permits you to specify the length
  at which to break between long (-l) and short (-s) fortunes.
  
  fortune -f now shows probabilities.
  
  A bug: fortune -a nn% filename filename ... now fails without an error
  message, if the filename named following the percentage exists in both
  the inoffensive and the offensive directories (that is, if you have two
  files containing definitions, one called fortunes/definitions and one
  called fortunes/off/definitions, and call fortune as: fortune -a 10%
  definitions religion politics ..., then fortune simply fails).  This
  appears to be an artifact of the changes that were made in storage/
  naming of offensive fortunes.  It only happens with the combination of
  a percentage with -a and inoffensive/offensive files that share a name.
  Temporary workaround: rename one or the other of the files (*sigh*  I
  don't like that as a solution).
  
  The man pages have been updated.  The old man pages are also available,
  but are not installed unless you do it yourself (the new ones are).  The
  new man pages have the extension .man; the old ones have numeric 
  extensions.
  
  A place has been created for fortune files containing HTML tags (the
  reason I started playing with this mess was because I wanted to be
  able to format fortunes nicely for the web without having to run an
  enormously complex script to figure out from formatting how best to
  display things, a particular problem since the formatting isn't
  consistent).  Tagged fortunes don't exist yet, and I'm seriously
  considering creating a slightly different fortune binary that would
  output the necessary headers and trailers (reducing the CGI script
  to complete triviality) (-f isn't really needed for a webfortune).

Todo: I'm thinking of adding a -x to unstr, to rot13 the output.  This
would have the effect of putting all the necessary tools in one package.
It further breaks compatibility with BSD tools (which has *mostly* been
maintained, merely enhanced slightly, although the change in how
offensive files are distinguished from inoffensive might be regarded as
breaking compatibility) by adding yet another parameter to unstr, which
didn't have any, before.  So I haven't decided, yet.

Todo: KOI8 encoded fortunes?  They couldn't be rotated without a great
deal of trouble, of course.

More minor fixes:

  The way that fortune -m prints its output has been slightly changed.  It
  used to print the delimiter first, then, if this were the first fortune
  from a particular file, it printed the name of the file in parentheses.
  It now prints the first fortune without an initial delimiter; if the
  fortune is the first from a particular file, it then prints
  (filename), newline, delimiter, newline *to stderr*.  Redirect stderr to
  stdout to get something *similar to* (but not the same as) the old
  behavior.  The new behavior, if stderr is redirected to stdout, and
  both are then redirected to a file, produces fictitious entries, one
  per file in which a match was found.  However, whether stderr is
  redirected or not, the new format produces files that strfile can
  parse without choking (the old format, since it placed the filename
  on the same line as the delimiter character, effectively forced editing
  of the file in order to make it usable by strfile, unless the option of
  concatenating two fortunes with an ugly "% (filename)" line separating
  them was considered acceptable output).  Under the new display format,
  if stderr is redirected into the file, you end up with filenames marking
  the separation between files (as before), but they are now valid text
  strings (which should probably, therefore, be deleted).
  
  In other words, if you don't care what files the original text came from,
  and want a new file containing (let us say), quotes from Mark Twain,
  you might do:
  fortune -am '-- Mark Twain' >twain
  The files accessed would march down the screen; the fortunes would be
  stored in parsable format into the file twain.  If, however, you planned
  to edit (perhaps to remove the quotes from the original file, you might
  then wish to redirect stderr to stdout.  Using bash:
  fortune -am '-- Mark Twain' &>twain
    or
  fortune -am '-- Mark Twain' >twain 2>&1
  
  [The above is now in the man page, more or less]
  
Late September, 1995
  Too many changes to mention, really.  Look at the source code for
  comments on individual files.  LINUX.DIF has been removed.
  
  It is worth noting that strfile was completely broken as distributed,
  and fortune had code to make it report a different file list than the
  one it used to retrieve fortunes.  There's some rather strong language
  on the subject in strfile.c; if it offends you, tough.
  
  Bugs were fixed, and some enhancements were added.  Unstr, in
  particular, has had its command line considerably enhanced.  Strfile
  now *really does* sort, instead of merely setting the 'sorted' flag.
  Ditto for randomizing.
  
  Noteworthy: the way to distinguish between offensive and non-offensive
  files has changed for fortune.  A second directory (which may be a
  subdirectory of the main fortune directory; the program doesn't
  add files recursively down a directory tree) has been added to 
  pathnames.h.  Offensive files should be placed there.  There is no
  longer any need to add the -o suffix to file names, and the problems
  with finding files (especially offensive ones) seem to have
  disappeared in the process.
  
  Currently, I'm working on breaking the fortune files themselves into
  smaller, more manageable pieces, checking spelling, punctuation, and
  grammar, and trying to reduce redundancy.  The eventual goal, after 
  the files are cleaned up, is another set of files carrying HTML tags,
  which would then massively simplify a CGI script that calls fortune.

  New Makefiles.
  
  A 'randstr' (I want to call it 'lottery,' but I won't) utility, which
  amounts to a poor woman's stripped-down fortune, to illustrate some
  other possible uses of strfile-type random-access strings files.
  
  Amy A. Lewis   alewis@email.unc.edu