[CF-Devel] Future crossfire changes/projects

[Mark Wedel]

Mike Ponicki wrote:

>
     
      The only real purpose to improving performance in crossfire is so that
     
     >
     
      we'll be able to have more than 4-5 people on a server at once without
     
     >
     
      getting serious lag. In everquest, there are between 1500 and 2000 people
     
     >
     
      on a server at once. I think our goal at first should be to get 100 people
     
     >
     
      on a single crossfire server with lag that doesn't make the game
     
     >
     
      unplayable
     
     
 There is certainly no douubt that crossfire is far from ideal in performance
right now.

 anyone know how much bandwidth a typical player in crossfire needs?  What about
for the commercial games (everquests/uol?)  At some point, I would guess
crossfre may see more of a problem in that most crossfire servers may not have
the same amount of raw bandwidth that the commercial servers are also putting
out. So at some level, while 100 players may get to be feasible in terms of cpu
performance, I wonder how many sites out there will have the bandwidth to
support that.

 Maybe someone wants to write a perl script that goes through the server log
file output, grabbing the CSSTAT values and plotting approximate bandwidth
usage?  May provide some relevant data.

 At many levels, crossfire is just not really efficient (sends image data, sends
what is likely to be a lot of repetitive naming information for objects, etc). 
Many of these are actually very hard to fix if you still want the game
extensible (for example, if you know the data was static, the client could
already have all the images, and the server could just send 2 byte identifies
for object names, and the client looks it up in its database, since it knows the
server hasn't added anything, etc).  Thats one advantage the commercial games
can offer.  While crossfire could do automatic updating of that informatiion, it
gets much trickier because the servers may noot be in sync (metalforge may have
added new names and csua has added other new ones, etc).  once again, with
commercial entities, they can be pretty sure all the servers are talking the
same thing.


>
     
      This way it would be much more like everquest, where the world is vast and
     
     >
     
      can't be traveled in mere seconds. Then, all of the performance gains in
     
     >
     
      the client/server architecture won't be without purpose.
     
     
 The other advantage of that is that there can be places which are discovered by
accident (ie, someone decides to explore the jungle).  With the size of the
current world continent, a player could basically examine most every square in
not a lot of time.

 Of course, in crossfire, it will always be difficult to hide things, because
the maps are available to anyone.  OTOH, you could very easily have some set of
maps that each server admin has added to some different spots on the continent.