Re: [glob2-devel] We must do a game, not an engine

[Cyrille Dunant]

On Thursday 26 April 2007 07.41:07 Kyle Lutze wrote:
> > On Lots of days many glob2 devs wrote:
> > words
>
> GAH! can't we all get along?!
>
> I didn't actually read the whole thread, skimmed mails here and
> there but it got long and I don't have enough time, but I thought I
> would throw in my .02. (note from after I finished writing this:
> this is long, but it is worth the read for the active developers)
>
> 1) bradley wants to do a core code rewrite, LET HIM! he's not being
> paid to help, he's volunteering his time and doing something that
> desperately needs to be done and not many are willing/able to do.
>  From what I gather he also doesn't want to work on the current code
> base because it's no longer fun and causes lots of headaches
> figuring things out and I don't want to see him leave over you guys
> trying to force him to continue working on that.

I sure can't prevent him. And as long as it works as well at the end, never 
mind.
But I am afraid that the "rewrites" are never quite complete (see nct and his 
skins)...

> 2) Cyrille: you seem to think rewrites are worthless and that time
> should be concentrated elsewhere, such as neat-o graphics. That's
> good, except that the guys doing the rewrites are coders, and that's
> their passion and they may not want/be able to do graphics that
> would do the game justice. On the other hand, what's stopping you?

I did the anims last week or so. Because I believe in nice graphics. 

> 3) I always like to point at the kernel development for this, if we
> had more developers it would be much more practicable too (at least
> from what I read this is how they do it)
>
> each major release, such as 2.4 to 2.6 is a complete rewrite. The
> rewrite is done when they find a block in the previous release that
> prevents them from moving forward, but they keep working on the old
> version to help it keep going in the meantime.

And we just got Yet Another Scheduler. But the usb layer still sucks, and some 
devices won't work. Of course everyone is free  to do what they want. But be 
honest, say it is for fun, don't rationalize it.

> that skinny version in context of glob2: as bradley said, the code
> started out decent, and steph and nuage did a great job, but then
> more futures were to be added that the current code couldn't support
> and two choices were presented: rewrite all the code to make room,
> or hack it in, many chose the ladder.

Actually, if you look at the codebase, you will realise that the vast majority 
of the code comes from nct and nuage. They did not, mainly, "hack" new 
features. However, certain parts of the code are written in a very convoluted 
way.

But again, you seem to assume things were hacked. No, things were don in ways 
that are more convoluted, sometime for optimality reasons. Some optimizations 
are probably nowadays not necessary anymore.

This does not mean they are unclean! Optimized code is painful to look at. 
Example: I needed sparse matrices. 

1) std::map<std::pair<size_t, size_t>, double> -- 0 lines of code, works.
2) operator overloading, valarrays and dedicated classes. speedup x 10 -- 50 
lines of code
3) clever aliasing tricks to get rid of temporaries. speedup x 3 -- 500 lines 
of complicated code

Someone seeing that and going into the internals would say huggh, this is 
horrible. I need to rewrite. And we would be back to step 1, which, let us 
admit it, is wayyyy, cleaner.

> This leads us at crossroads at this point as the code has been
> hacked so much that it went from the starting point steph and nuage
> had it to to something where it's been hacked by many developers
> over time (I've seen probably a dozen developers come and go while
> I've been apart of the glob2 community) to the point where we can no
> longer effectively work on the code to make the game better without
> bashing heads.

Uhhh, writing a campaign would make the game more valuable. With _no_ 
additional code.

> I have to be honest, I've gotten a bit addicted to a different game
> with my real life friends which is half the reason I haven't been
> around glob2 as much, the game is eve-online. They started out with
> a great game, but since then have been hacking mods in where the
> code wasn't in place for it, but since it's an MMORPG they don't
> have the time to do a complete rewrite, so they keep hacking the
> code to do patches and now any sort of battle in the game lags to
> all hell, node crashes, etc. which make the game not so playable
> making people quit, bad reviews, and players starting to disrespect
> the developers of the game for putting more "features" in before
> fixing the annoying little bugs that don't cause the game to cease
> functioning, but at the same time become very annoying to the point
> that players start questioning if the game is still worth it (like me).

You mean like rewriting the map format instead of fixing the desyncs? I mean, 
everyone can do what he well pleases, but let us not be deluded on what is 
actually useful _from_the_player's_point_of_view_.

Because we are forgetting the players with all this rewrite talk, are we not?

What are the issues they raise? what works not so well?

What are the game balance issues?

All of those points are important, and not one requires a rewrite, only 
tuning.

> Back to glob2. The code needs a rewrite so we can more easily
> implement new units, buildings, put an end to some of the pesky net
> game drop issues, etc. There is no way around this.

I an getting angry. 

THE CODE NEEDS NO REWRITE. IT NEEDS REFACTORING.

> On the other hand, I do believe it would be nice for the people that
> aren't interested on doing the rewrite to keep working on the 0.8.23
> version right now. images can still be made and used in the rewrite,
> maps that are made can still give an example and manually remade in
> the new glob2, and making campaign stories are hard, but recreating
> the maps, waypoints, etc. aren't nearly as hard as trying to come up
> with the whole thing from scratch.

most of the work in making a game is not code, it is graphics and writing and 
maps. Look at the number of each profession at Blizzard :)

> In a perfect world, how I think this stuff should be handled in the
> future:
> - we have a base (currently 0.8.23) that is worked on and updated,
> etc. until we hit enough issues that it becomes better to rewrite
> than continue hacking things in. At that point some of the people
> keep working on the old release, some people move to the next step

Once upon a time, a TODO was done which essentially said that the 1.0 is the 
current state, debugged, with gfx and campaign finished.

Are we working in that direction? NO.

> - we have a beta group working on a rewrite. This group should be
> able to take much of the extra libs and includes for a decent start,
> and if all code is documented in the base, and notes are included on
> where things were hacked in and why, it will speed up the
> development process of the beta to get to the point where the base
> is, and be able to add in extra features without the headaches that
> originally existed.
>
> ================
> I can give you guys more examples of other groups/companies that use
> this design but I think you can get the gist of it. From what I see,
> Bradley is working on the rewrite, k776 has been working on small
> updates, documentation, and other needed but somewhat boring work,
> steph has been helping here and there, but like me school is a big
> issue right now, I got addicted to a different game and feel behind
> on homework and helping glob2 for a while, so now I'm turning that
> around to get back on track, kai, leo, and and cyrille I have
> absolutely no idea, and nuage seems to have moved on.

I do gfx, sometimes. When I have the time and motivation. Yes, I am a 
bottleneck, sorry.

Actually, from Rama's drawing, I did ~90% of the graphics.

And you know what?

they need a redraw.

:)

CU

 -- CFD