Synchronisation between server and clients

[JaBoG32_Herby]

Hi there,

 

I`m not sure whether this has been asked before but couldn`t find anything searching the forum.

 

While playing a couple of MP missions throughout the last days some squadmates and myself observed AI units that on the client sides were damaged beyond usability (e.g. Apache without tailrotor, B-52 with both wings lost, An-26 wit one wing lost) but still flying around nicely in the virtual world. The Tacview files however showed these units removed from the battlefield for the clients.

 

My question is now, how do server and clients synchronize the damage state of any unit during the mission and is every single bullet, missile, rocket or whatever is exchanged between units synchronized in terms of motion vector and impact result?

Will the skill level for AI if set to random in the ME in game be the same on the server as on any participating client?

 

 

Thanks for any help.

 

Regards

 

Herby

[Maverick-90]

Moin Herby,

 

Wir fliegen unsere Trainingsmissionen immer mit allen gegnern auf Random, und uns ist sowas noch nicht aufgefallen, was der eine abschießt ist bei uns dann auch immer weg ;)

 

I have never noticed such strange behavior, my Squad and I are always flying with all AI set to random, and never had problems

 

Maverick

[Panzertard]

Herby, from my own observations:

- All mission data, including the current skill level is decided by the server.

 

- The clients "recalculate" bullet trajectories for visuals only. It may look like it misses sometimes, but server decision on damage is the key.

Server will decide how much damage recieved and send this info to all the clients.

Ex: On clients you may see a missile fly by and miss. Then you experience a sudden death. You just recieved the updated server data regarding damage-state.

 

- IMO thise kind of data transaction are sent by UDP (not TCP). Packets can be lost over internet. So for some situations you may miss a "update this object -> new state".

IMO, client death is handled correctly in all cases - it seems this is updated differently.

 

You may have observed a few bugs where the AI should have been dead, but the server decided it wasnt damaged enough.

 

Bottom line: I think the visual damage model vs the virtual damage may not be proberly designed in some sitautions. And this is what you observed.

[JaBoG32_Herby]

@Maverick: thanks for your remarks.

 

@Panzer: Thanks as well for your thoughts.

 

I did a little testing this morning in order to identify the problem in more detail.

 

For that I`ve set up a mission that couldn`t be more simple.

An AI B-52 gets engaged by an AI MiG-29 and that`s it.

 

The result is that the Server, running BlackShark 1.0.2 without any mods on an XP platform btw, allows the B-52 to sustain 1 missile hit without getting destroyed. Then follow 4 more missiles that all get spoofed by flare with the last of these 4 however causing enough damage for the B-52 to crash.

The client however, having tried with either FC2 and BlackShark, has the B-52 killed after the first hit. Which looks more realistic in my opinion.

 

I also reviewed Tacfiles from an earlier mission where the named Apaches survived not only 40+ Shilka hits but also 2 direct SA-10 missile hits both without being wiped out from the virtual world.

That also led to the situation where the AAA on the server continued to fire on the Apaches while the same AAA on the client didn`t care about the flying "zombies" any more at the same time.

 

So this might show that it is not only the visual representation, as I understand you suggested, Panzer, but some kind of wrong set-up on the servers damage modelling.

Is there any file that I could look up to compare the damage definition on the two computers?

 

 

Thanks and regards

 

Herby

[Panzertard]

- On the server-side (in the spectator) you could see the AI-flight survive? Keep on flying?

- On the clientside the AI "died", at least visually?

 

Please, tacview may report things incorrectly - so if you reffer to what lives and not, try to make that decision on what happens in the game. :)

[JaBoG32_Herby]

When I refer to the Apaches I just saw both choppers fly without a tail and with burning engines on the clientside.

 

What I wrote about the testsetup with the B-52 and the MiG I observed the behaviour on both server and client simultaniously.

[KungFuCharlie]

Not sure on the AI side of things but I had both wings knocked off the A-10 a few days ago and stayed in the air for another 5 minutes. Not enough control to land it that way though. So be it a visualization thing or a flaw in the flight models, something weird is present under certain conditions.