[glob2-devel] A roadmap

[Andrew Sayers]

Stephane and I talked tonight on IRC about where Glob2 is heading in the
near future.  The conversation log is available at:
        
http://www.pileofstuff.org/globulation2/logs/globulation2/2005.11.02.html
Search on that page for "19:26:31" - the conversation started at that
time.

We agreed that I would write up what we decided, but the following goes
a little beyond that, trying to draw some threads together.  I would
appreciate everyone's comments about this.

I propose that we make the broad aim of the next release to be a
gameplay mechanics code clean-up.  This would involve:

* Documenting those areas of the code that relate directly to the rules
  that dictate gameplay.  This means:
  + Units
  + Buildings
  + Maps etc.
  + Everything directly relied on by the above
    (but not things that are relied on by things that rely on the above)
  + *not* AI, or other high-level things like that
* When documentation naturally leads to a better implementation, the
  better implementation *should* be written, unless doing so would hold
  everything else up too much
* Other than bug fixes, *no* changes to gameplay
* Completion of the map rewrite
* Complete integration of the new stream code
* Integration of the new debugging code into all new code, and into old
  code when it's convenient to do so

It's very important that we stick to a limited rewrite so that we
actually produce something, rather than getting trapped by never-ending
feature creep.  As such, the network rewrite and the move to Qt should
be left for another release.  I feel like a bit of a cheat for saying
that after our long and productive discussion last week, but it's the
only pragmatic decision, and the results of that thread will be useful
when we do get around to networking.

After the next release of Glob2, it should be possible for new people to
come to Glob2 and completely understand the mechanics of gameplay by
reading the documentation around the source code.  It might not be as
easy for them to start hacking at gameplay-related code.  Again, that's
for a future release.

Finally, one thing that we didn't talk about on IRC, but which occurs to
me would be a good idea.  How about we make it a goal that classes being
documented must be rewritten to have all variables private, and complete
public interfaces?  That could mean a major rewrite of the whole
codebase, but I think it would be worth it for overall simplification of
the code.

        - Andrew