Hi.

Tonight, Jonahdude, Eggnogg and I played several RWR games. REEEEEEAAAAAAAAAALLLLLLLLLYYYYYYYY fun.

But, we encountered some issues....
- Lag. We live in UK, Canada and Australia, which isn't much good for real-time online gaming. We started with 300 bots, hosting in UK, but switched to host in Canada. Gameplay was a bit more fluid, but still unplayable. I then removed all bots and we setted up a 2v1 game. The host has a great advantage against other players, so there wasn't any other option than him being solo.
- AI chilling in the villa. They're defending bases which will never get attacked.
- Changing map is long. I guess there's a lot to load, but isn't there any way to get it faster?
- No log on crash. The game crashed five times in total for the three of us, for no apparent reason. Once twice in a row. We'd be glad to post the crash reports on the forum, but... there's none.

@pasik : If you need more informations on one of these aspects or about our experience, just ask. And get on IRC.

Thanks,
Paf.

Yup, it sure was fun

Basicly the same issues as Paf, a few movement and AI bugs, probably due to the lag.

so... THE ONLINE WORKS AND IS FUN AS HELL but you can really tell its a beta when you play it...

Thanks for trying such an extreme network conditions for RWR. I'm not particularly surprised you had issues there

The implementation currently depends on low lag, so LAN gaming and regional gaming should be pretty ok, I'm sure you might get hiccups there too. You can use DELETE key to show the statistics, it also shows some info about lag and the amount of data sent/received. There are things here that can be implemented to allow fluent gaming with higher lag too.

Most of the hiccups I've seen when playing online seem to be related to climbing over walls, I know that particular event needs better sync.

AI tries to defend all bases, but concentrates defense on bases near borders. Thus, some smaller portion of soldiers is left behind at the outer bases as well.

Map loading times have increased a great deal lately, but there are a lot of possibilities to optimize it, so it will get better in the future.

I'm mostly worried about the crashes; so did you check the log what was the last output? It can be found at C:\ProgramData\Running with rifles\rwr_game.log on Windows 7/Vista. Did it crash on the server each time?

Ho, I didn't look at the path correctly on the crash message box. I was looking in the installation directory. The rwr_game.log file doesn't look like explaining the problem. Maybe that's because the last time I played it didn't crash. Maybe you should create a new file each time the game crashes instead of overwriting the same over and over again.

Here's a pastebin anyway : http://pastebin.com/raw.php?i=LHLJ8WRM

I was one of the players. It was quite fun, but the game was virtually unplayable for me, because I had such horrendous lag (400 ping).

The biggest problem to do with latency was input lag. My character's actions in relation to my input had about a 1/1.5 second delay. I'm not too sure on this, but I think this would be because player movement (or everything for that matter) is calculated server-side, meaning that I have to wait for the server to respond before my character's actions are performed. If I'm correct on this, then I assume simply making movement calculated client-side would solve this.

You're basically right. The way it works currently is that the client sends its input events to the server, which applies them in the "real" game and then sends the inputs gathered from all the clients and the server itself to all clients, and the clients only now apply the whole set of inputs in their locally simulated games. Essentially, clients' inputs make a round trip on the server which is experienced as input latency on the client side.

In addition to this, the server sends sync packets of the game state meaningful to each client periodically, to keep the client-local games in sync with the server. The clients adapt to the sync data by smoothing out local errors of soldier/grenade/bullet/etc positioning, or, if badly out of sync, snapping the object to its correct place and motion.

RWR isn't far away of getting a proper client side prediction in place. With any luck, it's only short of applying the client's own inputs in the client-local game immediately they are inputted, basically killing the input latency, and estimating other objects' positions ahead of time. In practice, I might have to debug it quite a while to get it right. Anyway, it's just a matter of prioritisation when it will get implemented.

While CSP will fix the input latency, I'm not sure if even that can give a good gaming experience with 400ms lag. The lag will be then experienced as other soldiers being positioned in somewhat incorrectly in client-local games, making aiming and shooting quite hard on the client side. In the end, it's the server with the "real" game state that decides if a bullet hits someone, so with a big lag, a client might see that he hit someone locally, but in the real game, the bullet missed the target.