What to do with Xconq

[Stan Shebs <shebs at shebs dot cnchost dot com>]

In my first message, I made the case that Xconq is not yet successful.
This message goes into some specific mistakes and proposes some
corrections.

While reading about and reflecting on the various games in Xconq's
area, and playing them, I began see to see some strategic mistakes in
the current system.

* Confusion between game and game engine.  Xconq the program is really
an engine for strategy games.  Game modules are the games.  Valve
doesn't call their effort "the Half-Life module for Quake" - it's a
game in its own right.  By being forced into the framework of a single
generic program, the games lose some of their identity and interest,
while the C/tcl-level programming gets harder and harder, to the point
that interested programmers can't figure out how to work on it.

* Too many unfinished games.  Players and game designers never really
get to see what the system might be capable of.  (``Finishing'' a game
might mean anything from providing polished graphics to writing better
AI support.)

* Unfinished subsystems.  Agreements, standing orders, morale, the
list goes on.  They take up space, make the code more complicated,
but don't do anybody any good.

* Too far behind state of the art.  Gamers who know about Xconq admire
it as an open-source game that is comparable to early 90s games, but
then they go back to playing Starcraft.  Xconq doesn't have to match
the latest extravaganza, but it needs to be close behind and stay
there.  Note that that this doesn't necessarily mean graphics -
Nethack is still very popular without fancy graphics.  (It's also much
more complicated than Jay Fenlason's original 1984 version - gameplay
has evolved considerably.)

* Genre confusion.  These days, strategy games fall into several
subcategories - historical wargames (TOAW), complex simulations (Civ),
real-time (*arcraft), and maybe strategic RPG (HOMM 3).  Xconq doesn't
offer strong support for any of these genres, instead falling into a
mushy middle of simple-ish turn-based games.

* Uninteresting games.  As someone here commented a while back, too
many of the games in the library are experiments or demos.  They're
not very engaging - I myself go through the motions of testing them,
but I don't find myself wanting to play them all the way to the end.


One *non*-problem is the lack of publicity.  At one point I thought
that was the problem, and spent some time advertising Xconq more.  And
indeed, more people visited the website and downloaded the game.
However, these seem to have been mostly one-time visits - there's been
very little followup or feedback.  If you consider the mistakes just
listed, it's easy to see how the avid gamer would look around Xconq a
bit and then lose interest.  So given the existing problems, publicity
is not an issue, and indeed may be a bad idea if people form a poor
impression of Xconq.


How to fix


OK, so one solution is to give up and go do something else.  But so
far there's no other strategy game or game engine that I would rather
work on (FreeCiv is perhaps the closest).  So it seems best to take
the existing code base and work it over.  Here are some things we
should do:

* Define the game/engine separation better.  Building different
programs is probably going overboard, but for instance one could
imagine that the new game choice sets the entire look-and-feel of the
game, not just units and terrain.

* Use GDL more to enable and disable code modules that have been coded
for specific games or classes of games.  This is starting to happen a
little bit, because the idea of factoring everything in general ways
was already breaking down.

* Do more graphics.  The only games for which players don't care much
about graphics are the established old-time Unix games (Nethack) and
some very specialized historical wargames.  In both of these cases the
gameplay is very deep, with years of refinement having gone into the
rules (the opposite of the thrown-together-over-the-weekend Xconq
module!)  The current state of the art is 3D in various forms; that's
not necessary, but something on the order of CivCTP or RRT2 would be
good; isometric, 8- to 16-bit color, canned animations of rendered 3D
models.  This is closer than it seems; there is already a prototype
isometric display in Xconq for instance.

* Adopt more standard game graphics conventions.  While there are
arguments for using multiple OS-specific windows, it goes against both
principle and tradition.  The principle is that the suspension of
disbelief doesn't happen if cookie-cutter dialogs and window panes are
always popping up, and the existence of the tradition should be
obvious to anyone that has played computer games for a few years.  In
practice, this means that the tcl/tk interface hasn't been such a good
idea, although it's handy for game design.  My current front runner
idea is to use SDL to run the main window(s), and add some GDL
mechanism to specify the graphics sets to use.

* Merge interface code.  The multiple-interface design is good, but
the work of maintaining multiple GUIs is very time-consuming, and they
end up with different sets of features.  An SDL + tcl/tk combo would
work on all platforms of interest, and perform better as well.

* Add the capability to do real RTS games.  The machinery is 99%
there, would be easy to finish the job.

* Focus on a handful of game modules, and finish them.  One or two in
each Xconq-supported genre should be enough.  Do the graphics, make
the AI good, etc.

* Design a "featured game" that is unique to Xconq.  It should be
complex enough and deep enough to interest the jaded Starcraft or Civ
player, and stress the engine's abilities.  I was tinkering with a
"new standard" game last summer that features modern military
strategy, but that's still too timid.  I'm now thinking of a campaign
series, perhaps where you start out directing battles, and progress
through scenarios on successively larger scales, or perhaps an SF
game where you go from planet to planet.  What kind of a strategy
game would capture *your* imagination and keep you at the machine
all night?


Although this all sounds like a lot, the assumption here is that
all the other random stuff would go on the backburner and only be
revived if it's relevant to the new strategy.

What does everybody think?  Is this something you want to work on?
How about your friends?

Stan