>>If, for example, there is a 1/10 second delay in the transfer between
>>computers, that won't matter in a game like Doom. On the other hand, a formula
>>one car doing 300 km/h (185mph) will move about 8m (27ft) in that time. That
>>kind of delay makes it quite difficult to drive close to another car, as it
>>might cause problems like in your computer it seems like you've completed a
>>pass, but in the other computer the cars are still side by side.
>>Note that I pulled the 1/10 second out of a hat, but I think it is quite close
>>to a delay caused by a network (correct me on this if I am wrong).
>If you have 100 millisecond delay you're on a very slow network. 100 ms
>is a WHOLE lot of bit times, even on ethernet...
The amount of data needed to transfer for one car is probably something
like 8 bytes for position (4 bytes x, 4 bytes y), 2 bytes for heading,
2 bytes for engine rpm (to generate correct sounds), 2 bytes for speed and
maybe 2-4 bytes for other information. Now, speed could probably be dropped
out or maybe transferred about once a second, as can engine rpm. That still
leaves 14-16 bytes for each frame. Also, one master computer must collect the
data from the other players computers, combine them, and send out data for all
26 cars on the track.
Now, for the game, it is necessary to:
1) all other players send their 16 byte packets to the master.
2) the master has to combine the data and possibly handle collisions.
3) the master must now send 26*16 = 416 bytes to each of the other players.
4) each computer must calculate the screen for the players.
These four steps must happen about 20 times per second for the game to run
smoothly. Also, there must be time to simulate the player cars, and handle
the AI cars. This will total to about 10kb per second.
Will a network respond rapidly enough to allow the computers a reasonable time
to calculate the screen? (I'm not trying to convince that a network wouldn't
be fast enough. I only have had experience with ftp through a network, and
have reached transfer rates of few hundred kbytes per second, which would
indicate there shouldn't be any problems).
The 100 microseconds I used in my previous post came from the assumption that
the network doesn't respond immediately, but there is a slight delay.
If you Brian or somebody else know that a network (ethernet) can handle the
data fast enough, I don't see any reason not to include network support in
F1GP 2. Maybe they're just too lazy to develop the routines for network (but
if they already have serial line 2 player support, this wouldn't cause too
much work, would it?).
Samppa
--
| Kannaksenkatu 13 |----------------------------------------------------------|
| 40600 JKL | Check out the NHSound homepage at |
| FINLAND | http://www.jyu.fi/~tamasa/nhsound.html |