next »

[goldbuick]

I'm not sure this is appropriate for the feedback forum, but I'll give it a shot!

The Relay GDK project is about providing a low-cost open-source middleware solution for Massive Multiplayer games.

http://relay.ws/pdf/co-sim.pdf

It's a rough draft of a paper describing the algo behind the Relay GDK project. Right now I'm looking for people's thoughts/questions about the idea.

http://relay.ws/site/developers

[Guert]

It would be a good idea to post more details about the project for instance, what is it suppose to do? What's the Relay GDK project?

Later!

[Derek]

I moved it to Technical.  The Feedback forums are for games only, but you should be able to get feedback here for technical ideas.

Having read the abstract and introduction from your paper, and glanced through the rest, I'd suggest that you make it more clear what the problems are and what the goals are.  Theoretically, I should be able to understand your intentions from reading the abstract/intro.  As it is, it's fairly incomprehensible.

Some questions to ask:

What complex systems are we iterating through?

What is the current state of MMO games and what are we trying to fix?

How is this better than other approaches?

In a good paper, the basic concept can at least be understood by a layman, if not the details.  More specific ideas and concrete examples would go a long way in making your paper clearer.  A "distributed approach to delivering a multi-user experience" is pretty vague when you really sit back and think about it.

But I don't know, I was never much of an academic.  Just my opinion.

[george]

I think (with a small t) that he's talking about a MMO where, instead of the game hosted on one server (or multiple big servers) with all the players connecting to it with their clients, the game is 'hosted' on all the player's computers. In this case the game state needs to be synchronized across all the peers (i.e. players' computers) so that's what his math/physics/relaxation (?) is all about.

I guess this makes it easier and cheaper to run a big MMO with hundreds of thousands of players?

But at the same time, it seems like if you're going to have hundreds of thousands of players...you could spring for a bigger server?

Anyway MMO middleware is getting more popular and it'll be interesting to see what comes out of this.

[Zaphos]

Comments:

* Who is this paper written for?  I think a pdf whitepaper is an odd way to communicate to a general audience ...

* The mention of Jakobsen's paper is a bit odd since the link to it seems quite tenuous, and when you put it right up front like that you give the impression you're trying to implement a physics system for an MMO.  The (ab)use of the term Conservation of Mass adds to this initial confusion.

* The lack of references to anything from the networking community makes me wonder if any literature search was performed?

* The paper presents the way the system works but doesn't give an argument for why it's a good idea to use this instead of other ideas.  For example, in the synchronization system you could get a much faster convergence if instead of taking the average of your answer and their answer, you just took their answer.  Or if we had a quality function and took the answer of whichever peer had the higher data quality.  Presumably there's some advantage to taking the average, but I don't see this discussed in the paper.

* How do you interpolate between things that are binary states, like "dead" or "not dead"?  If you say "dead" is redefined as "respawned somewhere else" does that mean that the person who was unaware of the death will cause the respawned enemy to appear halfway between the place where he died and the place where he respawned, due to the averaging function?

* Finally I don't see anything on the topic of robustness to hackers, which is always a concern in online games I think.