Thread: .3 Second Positional Update Delay (Requesting Dev Response)

Originally Posted by Soukyuu

I don't think throwing spite around will make the more motivated to reply.

yes a couple of months mentioning it in many threads as well as a few hundred pages of asking sure got an answer faster.

Replying for support. I also have this issue.

Originally Posted by Soukyuu

So let's just call devs names and get the thread closed! Gotcha.

It was a harmless joke and hardly something that would result in the whole thread getting locked. Think befor you post.

Please don't bicker guys/gals. Don't want the thread actually closed because people started carrying on arguments. It should also be said that I'm not expecting an immediate response, especially not on the weekends. However, the more replies we get, and the more likes the thread gets would increase our chances of getting something (theoretically).

Originally Posted by Althea

So I talked to a buddy of mine who is a game developer. Apparently, with MMOs the polling rate between client and server can't necessarily go much faster than 0.3 seconds.

Modern games with good network code actually use timestamps and even employ some predictional methods to figure out if it was actually possible for you to get out of an AOE. I'll explain:

Current FFXIV Netcode
An AOE pops up under your feet. So you run out of it. Your client sends your position to the server a couple times while you're in it, but ultimately you can't get in another positional check in time before the AOE actually hits. This is what's causing people to take damage even though they got out of it. I'll show this with an image.


Here you can see each positional update (in blue). In this case, the size of the circle is the length of time you have to get out of it (0.7 seconds). Well, if you manage to get in two positional updates, you're going to get hit by it because the 3rd positional check is going to be AFTER the AOE hits.

So how do you get around this?
You attach a time stamp to each positional check, so the server can negate latency, and it has much more information for it to make some predictions. You also will need to send some extra events to the server.

So, for example, if I send a time stamp to the server along with my position twice like in the picture above, then send a special update to the server saying "on my screen I JUST got out of it, and here's the time stamp" the server can check each timestamp and say "okay .. they moved at a speed that is indeed possible (they're not botting or hacking), it was a constant speed, then it looks like everything checks out. You have successfully avoided it! And you take no damage.

TL;DR
The client needs to send time stamps with it's position updates so that the client can make predictions on whether or not you actually could have avoided an AOE.

Thanks for that explanation. I'm not in the know as far as game development actually goes, so all I see from my end is the issue and that I haven't ever had the issue before in my 10+ years of MMO gaming, even with less then optimal internet. I don't know how easy or difficult it is to deal with time stamps and all that to fix the issue, I mainly would just like to know whether it's something they're even looking at. I know they use to see it as an issue, because they lowered it from .5 to .3 seconds, however I don't know if that was their fix and it's done, or if they're continuing to follow it.

I would wager most people fall into the low-latency category, where you need to be out by 85-90%.. which is why knowing Titan's rotation and anticipating casts/pre-moving is such a popular and working strat for the fight (I'm unaware of people that have completed it without using that strat at least).

Originally Posted by Sunarie

I would wager most people fall into the low-latency category, where you need to be out by 85-90%.. which is why knowing Titan's rotation and anticipating casts/pre-moving is such a popular and working strat for the fight (I'm unaware of people that have completed it without using that strat at least).

Exactly correct. The chart I threw together doesn't even take latency into account. Each of those positions that get sent to the server will be where you were 100-200 milliseconds ago, furthering the problem.

Using time stamps would require a pretty good deal of effort on SE's part; but it's kind of a necessary evil here. Honestly, I turned off my auto renewal. While I like this game and feel like it has a lot of potential, things like this and more quality of life things just aren't there to keep me going. Holding out until 2.1 currently.

Probably get no response because Yoshi has known about this ever since Ifrit was implemented and now they say its fine and working as intended. Just shoddy netcode, I wish they would hire someone who can do a good job and get this fixed -.- but I doubt we'll ever see that happening.