[vworld-tech] DirectPlay for MMOG's

[Brian Popp]

I have to disagree. I'm not using DirectPlay and don't really plan to,
but as someone who is currently writing a UDP manager, all of those
features sound VERY useful. While they may sound trivial, keep in mind
that UDP is a brain-dead protocol and I haven't found many free
alternatives that handle its shortcomings. Everything I've found is
either a very simple sockets wrapper (SDLNet) or a very high level
system with lots of complicated overhead. If anyone knows of something,
I'd love to hear about it. 

bpopp - bpopp.net

On Fri, 2004-01-16 at 03:16, ceo wrote:
> dienw wrote:
> > ----- Original Message -----
> > From: "Phil Taylor" <ptaylor at ATI.COM>
> > 
> > 
> >>during DX8 and DX9 betas the DPlay Maze sample was used to hook up >1000
> >>simultaneous users.
> >>
> >>        1)Reliable and unreliable delivery of messages. Reliable messages
> >>will be resent until the target application receives them. You can assign
> >>the delivery type on a message-by-message basis.
> >>        2)Sequential and non-sequential delivery of messages. Sequential
> >>messages will be passed to the target application in the order they were
> >>sent.
> >>        3)Message fragmentation and reassembly. If message size exceeds
> > 
> > the
> > 
> >>capacity of a particular network, DirectPlay automatically fragments and
> >>reassembles the message.
> >>        4)Congestion control. DirectPlay automatically throttles your
> >>        5)Send prioritization. To ensure that the most important messages
> >>        6)Message timeouts. To prevent the outgoing message queue from
> > 
> 
> Hmm. Obviously, I'm not criticising dienw, but that sounds like a 
> completely useless answer to me.
> 
> 1000 players is an irrelevant statistic without some basic info like:
>    - latency
>    - topology
>    - traffic pattern
> 
> If that's for client/server it's also deeply unimpressive (if he'd said 
> "> 10,000" at least we'd be talking about a serious server; > 20k and it 
> would be impressive as a stat alone (although less so if this was with 
> 1% bandiwdth usage and 300ms latency ;)).
> 
> Most of the "benefits" sound like basic IP / UDP / TCP functionality. 
> Automatic fragmentation? WOW!
> 
> "Congestion Control" ... who cares? What really matters is the algo... 
> If it's equivalent to early TCP, it's worse than useless; if it's an 
> implementation of the latest modern TCP CC algos, it's to-die-for.
> 
> Adam
> 
> 
> ______________________________________________________________________
> _______________________________________________
> vworld-tech mailing list
> vworld-tech at cubik.org
> http://lists.cubik.org/cgi-bin/mailman/listinfo/vworld-tech