America's Army Forums
http://archive2.forum.americasarmy.com/

Net-whatever
http://archive2.forum.americasarmy.com/viewtopic.php?f=45&t=13581
Page 17 of 19

Author:  [Dev]Grubber [ Sat Jun 21, 2014 11:10 pm ]
Post subject:  Re: Net-whatever

Don Quixote wrote:
Ok, thanks.
In this particular situation, with AAPG hit detection is in client-side or in server-side?

Bullet hit detection is client-side with a server verification.

Author:  Don Quixote [ Sat Jun 21, 2014 11:13 pm ]
Post subject:  Re: Net-whatever

[Dev]Grubber wrote:
Don Quixote wrote:
Ok, thanks.
In this particular situation, with AAPG hit detection is in client-side or in server-side?

Bullet hit detection is client-side with a server verification.

I see.
Would be to much trouble switching it to server-side so we could give it a try?

Author:  Keebler_750 [ Sat Jun 21, 2014 11:15 pm ]
Post subject:  Re: Net-whatever

Are there ANY tools that can be used to track the flow of data in such a way that we can start to pinpoint discrepancies?

I have used two computers side by side on a table connected via LAN (haven't done WAN yet) and video'd with an actual video camera to analyze what lag may exist. This merely shows the character of the lag, not what causes it, and bypasses the spectator problems.

Is there anything? How do you parse the data stream in real time? Seems unlikely.....

:(

Author:  [Dev]Grubber [ Sat Jun 21, 2014 11:30 pm ]
Post subject:  Re: Net-whatever

Keebler_750 wrote:
Are there ANY tools that can be used to track the flow of data in such a way that we can start to pinpoint discrepancies?

I have used two computers side by side on a table connected via LAN (haven't done WAN yet) and video'd with an actual video camera to analyze what lag may exist. This merely shows the character of the lag, not what causes it, and bypasses the spectator problems.

Is there anything? How do you parse the data stream in real time? Seems unlikely.....

:(

I'd like to see that video :).

There are tools out there to watch network traffic and even catch it. You can probably google and find some. Points of interest are missing packets and such. Watching network traffic can be pretty much gibberish to some.

Author:  Keebler_750 [ Sat Jun 21, 2014 11:35 pm ]
Post subject:  Re: Net-whatever

[Dev]Grubber wrote:
I'd like to see that video :).

There are tools out there to watch network traffic and even catch it. You can probably google and find some. Points of interest are missing packets and such. Watching network traffic can be pretty much gibberish to some.


What is really of interest is what we SEE vs what happens in the code, as this is obviously what we base our reactions on.

I think a lot of people are coming at this with little actual fact, and a lot of impressions of something not right and I'd really like to see this studied properly somehow.

Author:  [Dev]Grubber [ Sat Jun 21, 2014 11:59 pm ]
Post subject:  Re: Net-whatever

Keebler_750 wrote:
What is really of interest is what we SEE vs what happens in the code, as this is obviously what we base our reactions on.

Great question! Sadly, I have to give you a terrible answer. The important things vary from situation to situation. For many of the issues discussed in this thread, the important things to see are what you saw and what they saw. So if you saw them warp they saw smooth play that tells me a very different thing then if you saw them warp and they saw a jump as well. The most helpful information is exactly what happened for both.

Keebler_750 wrote:
I think a lot of people are coming at this with little actual fact, and a lot of impressions of something not right and I'd really like to see this studied properly somehow.

Indeed! As issues come in we take a look at relevant code and asses the issue, the better the information that comes in the better we can look at the relevant code for potential issues. Don't assume what the issue is, just tell us and let us analyze it.

Great points Keebler! Can I have a cookie? Fudge Stripes?

Author:  Don Quixote [ Sun Jun 22, 2014 12:31 am ]
Post subject:  Re: Net-whatever

Keebler_750 wrote:
I think a lot of people are coming at this with little actual fact, and a lot of impressions of something not right and I'd really like to see this studied properly somehow.

Indeed.
And I will take the chance to say that this one is the best netcode ever.

Author:  Keebler_750 [ Sun Jun 22, 2014 8:27 am ]
Post subject:  Re: Net-whatever

[Dev]Grubber wrote:
Keebler_750 wrote:
What is really of interest is what we SEE vs what happens in the code, as this is obviously what we base our reactions on.

Great question! Sadly, I have to give you a terrible answer. The important things vary from situation to situation. For many of the issues discussed in this thread, the important things to see are what you saw and what they saw. So if you saw them warp they saw smooth play that tells me a very different thing then if you saw them warp and they saw a jump as well. The most helpful information is exactly what happened for both.

Keebler_750 wrote:
I think a lot of people are coming at this with little actual fact, and a lot of impressions of something not right and I'd really like to see this studied properly somehow.

Indeed! As issues come in we take a look at relevant code and asses the issue, the better the information that comes in the better we can look at the relevant code for potential issues. Don't assume what the issue is, just tell us and let us analyze it.

Great points Keebler! Can I have a cookie? Fudge Stripes?


This why I started doing LAN tests...because you could see both computers at once, side by side in real time. You can also add WAN lag by both computers connecting to high or low ping internet servers.

Still needs to be done with someone that can REALLY parse the data stream though...not just the network packets, but the CPU and GPU data.

Author:  Don Quixote [ Sun Jun 22, 2014 10:42 pm ]
Post subject:  Re: Net-whatever

This thread mainly due to the sarcastic approach it has become a bit ridiculous, however I feel that the following needs to be said.
What happens with netcode of this version is due to a bad implementation of it. Dont know if it was due to some inexperience about Unreal or about how a tactical first person shooter should be made or even both, the fact is that the netcode is not properly implemented.

Answering to Kleeber question, being sure that I already said this to him, I am going to repeat.
With current netcode architecture where hit detection is on client-side, is not relevant if you are connected in LAN or in a open network environment, the discrepancies are the same in both circumstances. In fact the current netcode method is not good for LAN and cant be used in LAN. If you do not understand what I am saying and wanting to know more I suggest you to open a support ticket with Unreal.

Speakin now about the netcode.
The Unreal network is in fact based on client-side prediction and (like i said several times) is one of the best when we speak about game engines, it is highly flexible and allow several approaches with its implementation in different platforms or game types.
I will focus only the two main factors with its implementation because these ones are the decisives in matters of output and also because all the rest is just a matter of parameters adjustment.
The factors are, Client-side hit detection and Server-side hit detection (both with Unreal network client-side prediction).
(Please note that with server-side hit detection I am not referring the "pure" server-side netcode of AA3 nor I am going to speak about it, that's a different subject)

Server-side hit detection.
This should have been the method chosen for this version, this is the method that fits better in a online tactical first person shooter. Having in consideration the game type and platform, this should have been the method chosen by an experienced programmer.
This method has a high rate of reliability in open network environment, even having a latency of 150/180 ms (ping), in matters of hit detection.
In a open network environment with latency below 100 ms the reliability is close to 100 % and in LAN environment the reliability is 100%, in matters of hit detection.
With this method the reliability with hit detection and also gameplay experience is highly dependent from the distance to the server (latency/ping), players with high latency 300/400 ms ping cannot expect to have a reliable gameplay experience, mainly in matters of hit detection.
In short words we can say that this is the Unreal network methodology without "lag compensation".

Client-side hit detection (the current in game)
The programmer wanting to allow the players with higher latency (which i have nothing against) to have a smooth gameplay, this method should have been made available as optional for server administrators to enable.
This method is much less reliable in matters of hit detection for everyone (low or high latency players), but that's not important for the purposes, is not physically possible for a player with a latency of 400 ms to have any kind of reliability on his actions, also is not good and cant be used for LAN.
Still, this method (as matter of preference or even because lack of options) allows the players with high latency to play the game and have a smooth gameplay experience.
In short words we can say that this is the Unreal network methodology with "lag compensation".

This is what should have been done.

Author:  [Dev]Grubber [ Mon Jun 23, 2014 2:09 am ]
Post subject:  Re: Net-whatever

Don Quixote wrote:
Client-side hit detection (the current in game)
The programmer wanting to allow the players with higher latency (which i have nothing against) to have a smooth gameplay, this method should have been made available as optional for server administrators to enable.
This method is much less reliable in matters of hit detection for everyone (low or high latency players), but that's not important for the purposes, is not physically possible for a player with a latency of 400 ms to have any kind of reliability on his actions, also is not good and cant be used for LAN.
Still, this method (as matter of preference or even because lack of options) allows the players with high latency to play the game and have a smooth gameplay experience.
In short words we can say that this is the Unreal network methodology with "lag compensation".

I have no idea where you go this from but there is nothing unreliable about the client-side hit detection of LAN. In fact there is little about this implementation that has unreliable hit-detection in terms of what you see is what you can hit. Server-side would generate unreliable visuals and some slight guess work in leads. So I guess it depends on your definition of "reliable hit-detection".

Page 17 of 19 All times are UTC - 5 hours [ DST ]
Powered by phpBB © 2000, 2002, 2005, 2007 phpBB Group
http://www.phpbb.com/