[MUD-Dev2] [BIZ] Project Darkstar - my take on it

[Jeffrey Kesselman]

On 8/31/07, Bristle <bristle2008 at yahoo.com> wrote:
.eally look at it.
>
> Turns out it is the license.  The best I have been able to find is
> 180-day non-commercial license. That must be a learning edition because
> who is going to put in 6 months into a project that has to end.

The full GPL licensed open source was released yesterday.

Go get it

www.projectdarkstar.com
>
> I also found a reference to a developer's edition that is a single
> computer 50 people limit.  I don't know the details of that and maybe
> that is what I have on my computer but I didn't see a copy of the
> license in my distribution so I wont touch it. I can understand the 50
> user limit, but don't understand the single-machine limitation and you
> can't copy the software to run more than one machine.

That purely a guesstimate of the user capacity your  likely to get
from a single decent box in your office running as a development
system.  Dont be confused, that has nothing to do with licensing.

>
> ok someone is going to ask - why do I want to run a game platform on
> more than one machine when today's cpu is fast enough to handle more
> than 100 zone servers?  Major reason is fault tolerance.

Again you are quite confused.

That is a *single* machine.  And you can run as many as you want, but
the single node stack is a *development* stack.

We are finishing the multi-node stack now.  It allows you to run one
world, one "shard"across an entire back end machine room.  You don't
have "zone servers" merely zone objects (if you want zones) .  They
are dynamicly shuffled between systems as necessitated by run-time
user access patterns.


> You can't have
> 5 9s reliability with one computer. blow one power supply and see how
> quickly the reliability number drops. With multiple computers and fault
> tolerance you can bump up the reliability.

Agreed.  Which is why in multi-node ANY system can handle ANY zone at ANY time.
The processing nodes are stateless.  If any one of them blows, the
others just take the users who fail over to them.  Outside of a
latnecy spike for the reconnection time, its invisible to your code or
the player.

>
> The second part is performance. I leave this part as a exercise for
> someone else.

The development stack is not tuned yet.  Again., its for development.
Once we have multi-node up and running tuning is the next priority

>And finally, the evolution to P2P network with the key
> ideas of area of interest and the use of some sort of proximity servers
> and some distributed hash table (DHT) mumble jumbo.

Clients are inherently insecure.  P2P is inherently uncontrollable in
terms of how packets fly around the net. Direct client to client
communication exposes IPs that people maliciously use against each
other.

Distributed  client-based MMO systems have been "just around the
corner" for about 8 years now.  I'm not worried about the competition
any time soon.

>
> If Darkstar is not the right platform for us low-budget, no-budget
> operators, then what is - i mean something that is not 5 - never years
> away?

You have to make your choice of platform based on your needs, but we
are working on hard on making Darkstar a good platform for the small
guy.

Ironically, we have quite a number of customers right now and very few
of them are  the small independent guys we initially expected.
Instead, we had some very big people come knocking right away.

You'll see and hear more at AGC next wekk.

Jeff Kesselman

Chief Instigator, Project Darkstar