vms-empire news See vms-empire.spec for more recent news. 1.4: Thu Aug 1 14:53:41 EDT 2002 Packaging tweaks. RPM spec fixes. READ.ME broken up into READ.ME and NEWS. Obsolete bug notes removed. Option -f to set savefile added. I'd like to move the documentation to XML, but can't yet because db2man doesn't render tables. 1.3: Fri Apr 19 05:13:33 EDT 2002 Walt Stoneburner's cleanup patch addressing all compiler errors and warnings, both in debug and in production mode; works with GCC v3.0.3. 1.2: Fri Jul 28 01:10:00 EDT 2000 The victory-odds table in previous versions was seriously buggy. I folded in corrections from Michael Self. I also took changes from James T. Jordan , who wrote some speedups, added ANSI prototypes, and cleaned up code. 1.1: 1994/12/01 16:07:03 I colorized and speed-tuned this and added a save-interval option. The rest of this history is Chuck Simmons's original notes. Because there are so many other Empire games out there now (all descended from this one!), I've reverted to the VMS Empire name. -- Eric S. Raymond History Apparently, this game was originally written outside of Digital, probably at a university. The game was ported to DEC's VAX/VMS from the TOPS-10/20 FORTRAN sources available around fall 1979. Ed James got hold of the sources at Berkeley and converted portions of the code to C, mostly to use curses for the screen handling. He published his modified sources on the net in December 1986. Because this game ran on VMS machines for so long, a previous version is known as VMS Empire. In early 1987 I reverse engineered the program and wrote a version completely written in C. In doing this, I used lots of structures and defined constants, and I attempted to make the code flexible and easy to modify. The algorithms used in this C version are completely new, the names of the commands have been changed to be more mnemonic, and new commands have been implemented. Only the format of the display is the same. I suspect that many of my changes are slower and less intelligently implemented than the originals. Also, I have not implemented some of the original functionality. However, my hope is that the commented C sources I have written will prove far easier to modify and enhance than the original FORTRAN sources. If you make changes for the better, by all means send Ed James and I a copy. The basic game has been heavily modified. I've changed the types of objects built, modified the parameters on others, and added lots of new kinds of movement functions. Read the man page for a complete description. The file 'bugs' contains lots of ideas for enhancements, and describes the bugs I haven't been able to find. Final Notes Unfortunately, I have a rather powerful mainframe at my disposal which is somewhere between 10 and 40 times as fast as a 68020 based computer. This means I can afford to use extremely inefficient algorithms. I suspect that running this program on a smaller machine, such as a Sun workstation or Vax will not be overly rewarding. In particular, the computer will take a very long time to move its pieces, and it may not be desirable to save the game after every move. (You mean your system doesn't write out 1/2 megabyte files in a few milliseconds?) This second problem is easily fixed, but I don't yet have any good ideas for fixing the first problem. The size of a saved file can be easily tuned by reducing the LIST_SIZE constant in empire.h. The only current simple tweak for making the computer move faster is to reduce the size of a map. Chuck Simmons amdahl!chuck Ed James edjames@ic.berkeley.edu ucbvax!edjames The low-end PCs of 2002 are more powerful than Chuck's 1987 mainframe. Performance tuning isn't the issue it once was. :-) My changes enable color on machines with terminfo color support, for a dramatic improvement in appearance and readability of the display. Color support, if present, will be auto-detected at compilation time. They also implement and document a `save-interval' option, addressing one of the misfeatures noted in the bugs file. I've also tweaked the sources so they compile clean under GCC -- they assumed the older K&R model of forward reference, causing many warning references. Finally, I've sped up expand_perimeter by cutting down on the number of array references it has to compute. This eliminates several multiplies from the inner loop, and is a technique that could be applied much more widely in the code. If efficiency matters that much to you, maybe you need to get outside more. Eric S. Raymond esr@thyrsus.com (home page: http://www.catb.org/~esr/)