Re: Majik.


Subject: Re: Majik.
phirate@ethos.co.nz
Date: Fri Jun 04 1999 - 00:42:13 EEST


> > 
> > First you guys need to decide whether you're going to do TCP or UDP. UDP
> > is preferable for realtime. I can write a connection layer for either. I
> > don't write code for Win, but so long as the class-abstraction is
> > maintained, if there is a problem, its easy to solve portably.
> 
> TCP. UDP would need much more work (major rewrites on server-side) and I
> don't think it is too good for firewalls. TCP is good enough, UDP would
> be yes, a little bit faster, but it would have a little bit more problems
> too :)

Well, you'll learn :) I've done realtime before, and TCP is useless for
anything much slower than a LAN. If you expect to have realtime fighting
over modem/TCP, you're in trouble :) I've seen a difference of 50ms on a
Quake connection result in a ratio of 12/1 frags, and they use optimised
UDP and movement prediction. And the server only had 3 players on it.
 
> We had previously used a system that if the object is visible or
> within certain range (on a map piece, thus not needing to add the object
> again when it becomes visible) the server would send its data when
> something happened with it (it moved or so). There was no other checks
> and it worked quite nicely. I don't see if this "syncing" thing is a
> necessity, they sync when something happens.

Ahh, so you don't expect 50 players on-screen simultaneously :)

> They are actually "automagically" prioritised. Human-controlled ojects
> does do more traffic than the others and thus the client does process
> them more :) Heh.

Well, you're the boss. But in that case you don't need my help, the TCP
module you have there, while it could be slightly improved in places, is
plenty good enough to handle lan/low-latency connections to your server.
Keep up the good work, feel free to gimme a buzz if you get stuck on
anything. I'll be watching :) And I'll send in patches if I happen to fix
something.

Phi.



This archive was generated by hypermail 2b25 : Tue Feb 12 2002 - 00:03:51 EET