Faith Black
Minmatar
Rolls Roids
   |
Posted - 2007.11.27 10:19:00 -
[1]
Edited by: Faith Black on 27/11/2007 10:24:25
 Originally by: kanojo1969
IMHO (as an old-as-the-hills software engineer), trying to locate the cause of the desync issue by casual invites is never going to work. I am disappointed that CCP don't have a system to let them replay these fights on demand.
One of the reasons for desynchs seemed to be the collision detection that gives different results for the client and the server.
On the server you were stuck in something, but on your client not. This is a scenario that needs to be taken care off, because server and client will naturally be out of synch by some amount of time. And positions of moving objects will therefore always be more or less out of synch on client and server too, which can mean that two objects collide and block eachother on someones's screen/client, but not on someone elses client or the server.
If you engage warp in a situation, where the server sees you hindered by an object and the client not, the client warps you normally away and you come out of warped, while the server still thinks you are stuck in your position with an active warp drive. If you click a button in that situation, like after having warped to a pos, you get the message: 'You can not do that, while you are in warp.'.
I assume the problem hasn't been fixed yet. Don't know, because I don't do fleetbattles at the moment.
Anyway there should be some means to approach the problem. I mean if the packets have time-stamps and are numbered, you should be able to track what's going on and find out, where information gets delayed or lost. Then the clients could do checks too, request resynchs, when things seem too odd (at a limited rate ofc, so that the clients don't hammer the server dead) and also send detailed reports back like noticed desynch conditions, packet-loss, latency, stress on the client machine to help analysing the problem.
Ok, this stuff needs also to be programmed and can introduce more bugs of course. I didn't say that it's easy. 
P.S.: I'm also not a software engineer and don't own a cluster. Got only some experience with bugs and doing the dirty work like supervising/helping students with multithreaded programming and implementing an OS in a lab or reverse-engineering, C++ template programming...
Can't wait for C++0x tbh., I've just seen a few code examples and read the feature list and think it will be cool. Probably becomes ISO standard 2009. Hope the compilers don't take so long to catch up. With ISO C++98 it was a pain. I better stop now. What was the topic again ?
|