Technical solution to eliminate desync in single-player sessions

if anyone is interested in some desync testing I posted this:

http://www.pathofexile.com/forum/view-thread/598924

in the bug forum a while ago.

The video I made: http://tinypic.com/player.php?v=347e0s9&s=5#.Uo0kTsSfjZQ

It's a video of me finding a barrel that very reliably desyncs me when I pass it, I encourage you to watch 30 seconds of it to see what I'm talking about.

I did some other testing so the vid is like 2 min, but you'll get the jist of it a lot faster :)

"

LogoOnPoE wrote:

... I would at least like to see that GGG, specifically Rhys fully understands what is proposed, before deciding whether it's flaws are too severe to implement or not. So far there seemed to be some miscommunication.

Whatever "miscommunication" you think may have occurred is easy to resolve. Simply go back to the sub-forum thread list, and click on the "GGG" icon to the left of the thread title to read all developer replies. Rhys replied in detail to the OP's hypothetical pipedreams, and hasn't seen fit to make any further comments over the last two dozen pages.

"

HellGauss wrote:

Come on guys, do not troll....

There is no need of a REALTIME representation of gamestatus on the server. There is only need of a not-so-much-delayed check of the integrity of commands sent by the client to the server.

Yes, and I'm CERTAIN that GGG job offer is going to arrive any moment now...

"

RogueMage wrote:

"

LogoOnPoE wrote:

... I would at least like to see that GGG, specifically Rhys fully understands what is proposed, before deciding whether it's flaws are too severe to implement or not. So far there seemed to be some miscommunication.

Whatever "miscommunication" you think may have occurred is easy to resolve. Simply go back to the sub-forum thread list, and click on the "GGG" icon to the left of the thread title to read all developer replies. Rhys replied in detail to the OP's hypothetical pipedreams, and hasn't seen fit to make any further comments over the last two dozen pages.

In qwaves final plan security is completely destroyed. There is no need for the server to predict anything, and only needs to hold onto 1 saved gamestate.
So the vast majority of GGG's comments and criticisms do not apply. Only those relating to security(such as immortality through never sending the 'i died' packet) and the one about validating snapshots requiring more processing(and code) than stepping through a seperate game.
Im confident that such a plan could be made to work efficiently. But as iv said about 2 dozens times already in this thread, security is nearly abandoned.

I feel fine as a math and informatics professor in my university.

Well... explain. Why do you think that there should be a realtime simulation of the gameplay on the server?

While you are answering, i'm going to get some popcorn.

"

SkyCore wrote:

In qwaves final plan security is completely destroyed. There is no need for the server to predict anything, and only needs to hold onto 1 saved gamestate.
So the vast majority of GGG's comments and criticisms do not apply. Only those relating to security(such as immortality through never sending the 'i died' packet) and the one about validating snapshots requiring more processing(and code) than stepping through a seperate game.
Im confident that such a plan could be made to work efficiently. But as iv said about 2 dozens times already in this thread, security is nearly abandoned.

It's not as bad regarding security as you make it sound. As I previously said, the not sending 'I died' packet, and just not sending anything can be resolved by server actually running the simulation for some defined time as if you did nothing. It that case if you died on the client, you most probably also will die on the server.

Regarding validating snapshots requiring more processing, I think Rhys was talking the the context of qwaves proposed leniency regarding fp rounding (which imho wouldn't work), not in the context of server load.

I wouldn't say that security is nearly abandoned. Some limited possibility to change the outcome in the predefined limitation of seed exists, but it's far from dropping security.

"

HellGauss wrote:

I feel fine as a math and informatics professor in my university.

Well... explain. Why do you think that there should be a realtime simulation of the gameplay on the server?

Right, professor, and as an engineer, I'm fine with your academic pursuit of a perfectly deterministic simulation running in a "cristalline a-temporal universe". In your own ivory tower, naturally.

And the answer to your request for an explanation is just as simple:

GGG runs a real-time game simulation on its own server because they want to insure security against cheats and hacks and chose not to trust the client in that regard. They made an architectural decision based on their own engineering priorities and business model. Given their self-imposed security requirements, that remains the most practical solution available.

"

qwave wrote:

..

I dont know if you are wrong or right, but I just want to thank you for trying to help and fix the problems PoE is having, it shows that you really care about the game even if in the end you are not right, discussions like this can make new ides and solutions.

"

Rhys wrote:

..

Thank you for participating with this, it shows how good devs you are. Keep a open mind, and continue to read suggestions of other people, it may help in some way.

This thread has gotten quite silly.

I can't bother to check... but has anyone compared GGG to Nazi Germany yet?

"

qwave wrote:

Calculating damage is not.

It is if you use fixed point arithmetic, which is the only way to deterministically crate values.

The reason why other games don't have this problem is because they don't do it, mainly because they dont use your crummy design