frankc at maddog.com
Fri Jun 7 17:13:58 New Zealand Standard Time 2002
From: "Koster, Raph" <rkoster at soe.sony.com>
> There is one hugely critical difference, though. Game development
> tends to be much more iterative, because "fun" isn't really
> quantifiable. So in that post, I went through the initial stage of
> design. But after that comes much throwing away of documents,
> starting over, rework, etc, that isn't nearly as prevalent in
> other areas of software development. In most large-scale software
> development, the initial requirements set forth will, if done
> correctly, highly approximate the desired end result. The same is
> not true in game design.
It's the same in the "serious" side of software development.
Nothing ever gets done in a linear way. Even the mature industries
have their bag of tricks to make it appear that development is
Part of the reason why I am leaning more and more to the Adaptive
Model (one of the agile development models) is that it focuses on
deliverables. The Milestone Acceptance Test (Chris Hind in
GameDeveloper May 2002 talks about) also focuses on the
deliverables. Then you have someone like Brian Marick
(www.testing.com) who wants to completely throw out the V&V model
and replace it with a handoff model.
Brian actually outlines a more onion-style of testing and
development than what you see in the MAT. However, you can actually
see the time-box in the MAT -- something that is also characteristic
of XP and SCRUM -- two other agile development methods.
The handoff approach reminds me of the Japanese manufacturing method
that was part of JIT. In that case, the dependencies were fairly
simple. The model was a stream. On the downstream side someone
delivers to you so you are the customer. On the upstream you
deliver to someone so they are your customer. You don't deliver
incomplete things to a customer and you don't give the customer more
than they can handle.
I can already hear my own Silicon Valley techies say, "you can't
just depend on deliverable or handoff testing. You gotta do system
and acceptance testing." Sad fact is that system testing in most
companies is just days not weeks and accptance testing is the CEO
and someone from marketing doing some things until lunch time.
True story. I talked to a CEO of a tiny streaming media company who
wanted a big-time program manager to come in and do everything. So
at some point I asked about testing and he said something like
"that's not a problem. The whole company tests for a week." Sounds
like he has unlocked the secret. I am sure the whole company
develops for a week too.
MUD-Dev mailing list
MUD-Dev at kanga.nu
More information about the MUD-Dev