Why the netcode issue exists, and why it cannot be fixed.

Makes sense as well while casting my shields as a BLM will let the first hit get in before blocking or lowering the damage. Hope they fix without breaking the game.

Quote:

---

Originally Posted by APilgrim

To Billie21 - Eh, a CCNA background has nothing to do with (MMO) game development, programming, or management - so it's a rather innocuous thing to bring up in this topic. It was only brought up as a very direct response to someone. I fully admit (again) to trying to deduce things second-hand along with everyone else since none of us work for SE and SE isn't responding or talking. And with that said, I'm bowing out of this interesting topic with the hope that it is eventually addressed further.

---

What does CCNA mean?
I am not disputing your answer, actually I was agreeing with it. In reaction to the same mesage that you answered.The one quoted just above, weighting your opinion against someone else's opinion based on how knowledgable you and the other guy appeared, rather than the sole content of your messages.

But that's off-topic, I didn't mean to start a discussion on this subject, I just couldn't help but react to this way of thinking, because I've seen what it can lead to.

As for deducing second-hand, as long as SE remains silent, that's the only thing anyone can do.

Awesome post, OP. I feel like I have a much better understanding of the situation now, so thanks. However, this really bothers me:

Quote:

---

Originally Posted by Sinth

-Server infrastructure is more costly (servers must do much more data processing)

---

Strange that Squeenix would be so fastidious about cost in pretty much every other area but this. :/ I don't really know anything useful about server stuff, but would the added processing from having to do everything server-side mean that the servers themselves also have less capacity to process players and such (perhaps why the servers get overloaded so easily)? I could just be pulling stuff out of my ass, but it makes me kind of boggle at why they'd choose to do it this way. Especially when I'm pretty sure the client live-state is the standard system for other MMOs.

And on top of that, ps3 users like me.... oh boy.

Quote:

---

Originally Posted by Blarp

If this is true, can you explain why some people say they do not have this issue? Is it a perception issue?

---

Can you prove that God doesn't exist? Can you prove that there is no dragon on my garage?

Argument from ignorance fucking everywhere....

so basicly what your saying is what we have always known se doesn't care about its over seas market

For the 'Esuna test', I've finished casting mount chocobo while autorun is on, but I do not actually get on the mount even if I finish the spell. This suggests the server decides whether you get on the mount or not, and even if its response is late (or not at all), you still don't get on the mount. Note that this also works in reverse. You can definitely move with about 0.3s left on any spell and it will almost always finish casting. Your client can easily tell you this action is impossible, but it doesn't matter because it's not the one making the decision. As long as the server says it's okay (and it's okay because to the server you never moved, due to lag), it's also okay to the client.

in titan you really only die if you get in stacked plumes . land slide is a joke to avoid , just side strafe out and back in to avoid . You can even run into landslide when it appears and avoid it . Only way to get hit is for it to appear on you and you either don't move or you run out and get hit . So remember run out and back into it = 100% avoidance .

The game reminds me of the matrix , the game world has rules but you can bend these rules using the lag, like casting spells and moving while still casting and not getting interrupted etc ,

Quote:

---

Originally Posted by Grailer

in titan you really only die if you get in stacked plumes . land slide is a joke to avoid , just side strafe out and back in to avoid . You can even run into landslide when it appears and avoid it . Only way to get hit is for it to appear on you and you either don't move or you run out and get hit . So remember run out and back into it = 100% avoidance .

---

The 'run out and then in' sort of works because most of the problematic abilities must be preempted. That is, you must be moving before it happens to reliably avoid it. Therefore, 'run out and then in' is not preempting the action, and is almost certainly too slow for you to get the intended effect (run back in and get hit by AE), so you actually end up avoiding the AE. That said, if you can run out of something, you can just stay out of it too. It's most certainly not the 'run back in' part that makes you avoid the AE.

I did the 'Esuna test' a bit more, but I used mount chocobo because it has a very distinct result when the spell succeeded. Although this game is laggy, it is not common to have an entire second of lag, so in the interest of saving time here's what I did:

1. Cast mount chocobo.
2. Auto-run on with 0.5s left (this basically simulates a 0.5s casting).

Now there are obviously some room for error based on my reflexes, but what usually happens is that I start running at 0.5s, the spell castbar finishes, but I do not get on the mount.

This means despite the fact that the 'spell interrupted' message never came, the server did get the fact that my position has changed and thus did not allow the spell to complete. The client was not told of this, but the outcome is consistent with what you'd expect (shouldn't be able to cast a spell while running). The server doesn't really care that it interrupted my spell without ever telling my client that it was interrupted, which is consistent with the assumption that the client has basically no power to do anything to begin with. In this case, it isn't even being told the spell was interrupted.

The OP makes a good point, but its not all doom and gloom and wouldn't take millions to fix. What could be done is a hybridization of the two systems. The server will check your movements over a larger area and let the client select where it is on a smaller scale. This is how other MMOs actually do it. If you ever played WoW, you'll notice that some people who are lagging are teleporting a bit in a predictable pattern or chunks. They kinda stay sill then move and repeat.

Althea's suggestion on pg 10 is also a simple way to fix the issue. Let the clients tell the server they exited or re-entered the AOE. Yes it will make it easier to hack to be immune to AOE, but the impact is alot less than the overworld mining teleports already in effect.

It's because the fix is easy (have client handle it) is why this problem is extremely hard. The problem didn't come due to lack of understanding of how fundamental MMORPG concepts work. It's because someone up there chose to go against the norm and until that guy changes his mind you can't possibly fix it. It's not like someone just facerolled and accidentally produced this server-based code that is nothing like any other MMORPG. It's obviously very deliberate which is why it is very difficult to convince the guy who thought this is a good idea to change it.

Just curious: how was FF11s netcode? Is it similar to other mmorpgs or FF14?

Quote:

---

Originally Posted by Sinth

The reasons why SE would choose server dictated live-state:
-Prevents desync bugs (on the server)
-Prevents countless potential duping bugs
-Prevents (in theory) teleportation and other unauthorized client commands from being accepted

---

Unfortunately not a single one of these have been prevented.

There are lots of desync bugs, hence the issue.
There have been a couple duping bugs and even worse than that... as the game is extremely hackable
lol Teleport, look at all the mining bots.

Quote:

---

Originally Posted by Dyrstahl

Just curious: how was FF11s netcode? Is it similar to other mmorpgs or FF14?

---

It was equally terrible, but the gameplay was so slow, you could play with like 500ms and never notice.

There has to be some thing done soon many people i know have or are quiting when subs run out due to not been able to enjoy it due to this delay.

Avoid it player is happy server kills you any way not going to make people mad at all and quit.

Quote:

---

Originally Posted by Blarp

It was equally terrible, but the gameplay was so slow, you could play with like 500ms and never notice.

---

Playing white mage on 11, even with 80% fastcast, I don't remember frequently seeing a cure land on a dead player -- the server cancels the spell before it completes if the target dies while casting (even if the spell takes less than ~100ms to cast). It seemed that any amount cured showed on the client did accurately reflect the server's state. Really the only questionable syncrhonization issue I've seen is something that SE ninja-patched in a few years ago involving zoning: previously a warp or zone change would be successful even if you died slightly before the warp resolves, and you'd end up dead at your destination. Now if you die even while warp takes effect, it's canceled.

I hate to dig on a group trying to make a living but...

The server guys can't even get the "Server Live-State" correctly because the client can, apparently, still send whatever commands it wants to the server and have it implemented (Teleport, Item Conversion, Level up, system queries, packet intercepts, etc).

So basically we are getting all the drawbacks without any of the benefits. Year/months to change the code? Millions? What is the code right now? I bet it was outsourced. I can't image that the people who coded and worked on such a great game coded the servers to allow such blatant vulnerabilities. I'm not even talking about the art.

The game mechanics are smooth and polished to almost OCD status. Do you believe the same group who changed Scholar pet control mechanics (that you can't issue commands when casting) is the same group who does the netcode? Those guys give a shit. I mean, come on. Can't issue pet commands while casting? How minor is that?

The server guys so far, just deal with problems as they come up. Slowly. Without any planing. And many emergency maintenance.

Also I'm not sure how they made the decision, but server side architecture, how they've implemented it, is just really really bad. You trust and verify. That's the golden rule. Do it any other way and you're going to have a bad time. Right now, how I see it, they don't trust, and don't verify.

Even in a server based game somethings have to be done by the client. The client always issues teleport commands to the server because pathing is a very difficult problem for the computer to figure out. Instead the client just sends a position update with timestamp, and the server treats your movement as teleport. Therefore it is easy to have teleport hacks since all you're changing is how much you've moved.

At any rate usually the most common 'hacks' are really exploits. That is, doing stuff that is legally acceptable in the game in an unintended way. It's very difficult to hack your client to say "I have no GCD", but perhaps you can find a tricky combination of abilities that somehow reset your GCD. It's hard to say "I have Godmode", but perhaps you can find a clever way to stack certain buffs/items to have 100% chance avoidance which is usually good enough. Exploits cannot be stopped architecture design, because what you're doing is allowed by the game.

Quote:

---

Originally Posted by Astarica

At any rate usually the most common 'hacks' are really exploits. That is, doing stuff that is legally acceptable in the game in an unintended way. It's very difficult to hack your client to say "I have no GCD", but perhaps you can find a tricky combination of abilities that somehow reset your GCD. It's hard to say "I have Godmode", but perhaps you can find a clever way to stack certain buffs/items to have 100% chance avoidance which is usually good enough. Exploits cannot be stopped architecture design, because what you're doing is allowed by the game.

---

How do hackers steal gil when an online player goes near a market board? I've never heard of anything so crazy even in the dime-a-dozen Korean F2P MMO's. There's some pretty serious issues/vulnerabilities with this game. There's days I think SE took the 10 years of accumulated knowledge and experience they have in this genre and flushed it down the toilet before they set out to make this game.

Quote:

---

Originally Posted by Blarp

How do hackers steal gil when an online player goes near a market board? I've never heard of anything so crazy even in the dime-a-dozen Korean F2P MMO's. There's some pretty serious issues/vulnerabilities with this game. There's days I think SE took the 10 years of accumulated knowledge and experience they have in this genre and flushed it down the toilet before they set out to make this game.

---

I don't know the details of that particular exploit but I thought they found a way to get other people to buy stuff from them without that guy even being there. It'd still fall under the 'exploit' category. A hack would be me taking over your account and grab all your money, and it's clearly not what's happening as the accounts themselves are not compromised. And it also highlights the real serious problems with MMORPG tend to be exploits, not hacks, so gutting your netcode to stop hacks isn't really that helpful.

You say "poor network connection" in your OP however I feel that statement is a little inaccurate.
I live in Australia, which is technically a "supported" country, however due to my location, I ping to NA servers at 350ms (without VPN) or 240ms (with VPN). If I were to ping any other server in Montreal I will get 240ms..
In addition to this I have 0% packet loss and my local speed is 100Mbps down/2.5Mbps up.
From a technical standpoint, my service is not poor, but simply my geographical location.

My connection is not poor, however without a VPN the game is impossible to play at an endgame level..

I think it is worth noting that a big downside with the server side option is that someone who lives in NA, Canada or EU will have a pretty large advantage over someone who lives far away, like me.

Client side however, this advantage is greatly reduced.

This will become an issue once PvP is released next patch.

Quote:

---

Originally Posted by Altena

I think it is worth noting that a big downside with the server side option is that someone who lives in NA, Canada or EU will have a pretty large advantage over someone who lives far away, like me.

Client side however, this advantage is greatly reduced.

---

It's more like the other way around. In the server model everyone is limited by the 0.3s pooling frequency, as regardless of how close you're to the server, you can't get key information any faster than every 0.3s (like when AEs fire off). Everyone is roughly equally screwed under FF14's model. If your client could make its own decision, the guys living closer to the server would have an advantage. It does create an equal playing field, but in the wrong way (by crippling everyone equally).

Quote:

---

Originally Posted by Astarica

It's more like the other way around. In the server model everyone is limited by the 0.3s pooling frequency, as regardless of how close you're to the server, you can't get key information any faster than every 0.3s (like when AEs fire off). Everyone is roughly equally screwed under FF14's model. If your client could make its own decision, the guys living closer to the server would have an advantage. It does create an equal playing field, but in the wrong way (by crippling everyone equally).

---

I see; that's pretty interesting to note.

So to make everyone equal, they basically reduce the performance of the "optimal" network/geo location to a more standardized level... Hah, seems pretty poor design choice in my opinion :/

Really confused why they decided server side was a better choice..

Just a question about VPN. How is it supposed to make things better? I don't know much about networking and such, and I know nothing about VPN. But I read up some on it, and the way I understand is that VPN mask your IP but in the process it lowers your internet speed. If that is so, I don't understand how it would help.

My lag issues aren't really what has been discussed in this thread though. I would be extremely happy if I had 0.3 s lag. The worst days I can't even craft. You press the actionbar and then have to wait 5 seconds before the action is performed. It's been like this on and off the past week. Although late at night when not as many people are online, all lag issues goes away. Is SE squeezing more people on each server than they are able to handle?

Quote:

---

Originally Posted by Sinth

Close proximity (or low amount of hops) to the data servers. In low-latency scenarios, the server live-state model is superior in all regards to client live-state.

---

So JPs are suffering this problem in large numbers as well because ... ?

Quote:

---

Originally Posted by Altena

Really confused why they decided server side was a better choice..

---

Because SE's design decisions were based on their paranoia about 'hacking' and RMT, technical issues were entirely secondary. Which is why V1 was a dog performance/UI wise (hacking/botting) and no mailboxes (RMT) and why V2 still has fundamentally broken mechanics like this.

you all should keep something in mind. their polling is an issue because NONE of the content is optimized for it. the content is optimized for realtime. which is why you often still get hit regardless of dodging. if they were to just optimize the content for their polling speed it would be a non issue.

those of you saying you are not getting this. well actually you are. for whatever reason you are not noticing it, well good on you then.

the polling server side effects everyone. period. they need to do something about it and unless they plan to restructure the whole game I suspect the only solution would be to optimize the content for the polling

It's a really dumb down explanation, that is almost technically invalidated by elegant engineering code. It is not an impossible issue, not for the past 5yrs or so, since online gaming started to become big.

Let's face it, FF series and SE in general was never very good at technical prowess, and they continue to keep that name.

There are lots of ways to fix(technically "mitigate") this on pure tweaks alone, but they don't have the brain power to do a "tech surge" on it.

Realistically they should have thought this though about this before making the content, but that's another one of SE's weaknesses, that is nothing new (like decade of, this is so "SE").

I have a question for the OP.

I'm not nearly as knowledgeable on this subject as the OP, but I do understand it.

Here's my question: If the server is dictating live-state, why do buffs sometimes go off but have no effect? I don't mean like HG/Bene deaths. I mean like sometimes Rampart, Foresight, Bloodbath, and even HG are consumed but the buff never registers (the user still lives) and the buff has no effect, yet the ability goes on cooldown.

In a server live-state, the thing I would expect to see is perhaps that you don't appear to get the buff (no status icon) but you actually do. Packet corruption of the client saying "I get Bloodbath, the server saying "You can has bloodbath, we agree it should be off cooldown", and the client not properly rec'ing/handling the recognition packet.

This is easiest to see when it happens to Bloodbath, because scrolling combat text shows us all the minimal heals. It's harder with rampart to say "Oh, well is he hitting me 20% weaker or not?", but Bloodbath or HG are pretty easy to see. (HG is not immune to all damage though, so some people may falsely observe this)

Hmm.. Perhaps it could be the sending packet that is corrupted/failing? Maybe the server tracks whats on cooldown just to prevent hacking, but the client has some basic protection. Maybe the client throws it on cooldown, notifies the server, and here's where the break happens?

Not everything is jumbled up into the server, and should not be. There is a polling list, and several key factors. Some of it is slow down by anti-hack checks, but not all of it.
Some of it is very low risk, and "even if a hacker tries, he won't be able to do much", successive checks that can flag a maligned code.

for instance position checks (the part that is the center of attention). You can actually have a predictive analysis going on based on the ping time of a player. Say you have:

player A, he has a poll of 35ms on his flag, the server does a rough mathematical equation, that predicatively gives him 2% "fail safe" zone for the AoE circle.
player B, he is polling of 180ms on his flag, the fail safe is expanded to 10%.

To prevent "lag hacking" that's common in some games, you resolve it with successive checks on other factors that "time out" or "impossible limits" making a high risk into a low risk problem.

It basically all boils down to the brain power of the programer(s) ability to make mathematical equations that can predict player data, and then have a proper "fail safe" for situations that are clearly outside the norm.

It's elevator theory, where if you find someone to stare at the lobby elevator for long enough time, you can program a sequence that can do the exact same thing using if/or/and

Quote:

---

Originally Posted by Steeled

I have a question for the OP.

I'm not nearly as knowledgeable on this subject as the OP, but I do understand it.

Here's my question: If the server is dictating live-state, why do buffs sometimes go off but have no effect? I don't mean like HG/Bene deaths. I mean like sometimes Rampart, Foresight, Bloodbath, and even HG are consumed but the buff never registers (the user still lives) and the buff has no effect, yet the ability goes on cooldown.

In a server live-state, the thing I would expect to see is perhaps that you don't appear to get the buff (no status icon) but you actually do. Packet corruption of the client saying "I get Bloodbath, the server saying "You can has bloodbath, we agree it should be off cooldown", and the client not properly rec'ing/handling the recognition packet.

This is easiest to see when it happens to Bloodbath, because scrolling combat text shows us all the minimal heals. It's harder with rampart to say "Oh, well is he hitting me 20% weaker or not?", but Bloodbath or HG are pretty easy to see. (HG is not immune to all damage though, so some people may falsely observe this)

Hmm.. Perhaps it could be the sending packet that is corrupted/failing? Maybe the server tracks whats on cooldown just to prevent hacking, but the client has some basic protection. Maybe the client throws it on cooldown, notifies the server, and here's where the break happens?

---

On that specific issue its impossible to say without inside knowledge of the source code. It could possibly be a result of packet loss in addition to a server/client error checking issue. It could possibly be a bug in the client. It could possibly be a bug on the server.

OK so let's take OP's post into light.

This won't be fixed! Therefore I put the next question at hand.

Why design content to rely so heavily on this? There are plenty of mechanically fun fights that are nearly unaffected by this...

Garuda and Cadeuceus come to mind.

You give us a bazillion things to dodge and a system that doesn't perform well in that situation, we're going to complain... Since the system can't be changed, let's be honest, we'd prefer content that performs well within the system.

Quote:

---

Originally Posted by Jonny_Tapas

OK so let's take OP's post into light.

This won't be fixed! Therefore I put the next question at hand.

Why design content to rely so heavily on this? There are plenty of mechanically fun fights that are nearly unaffected by this...

Garuda and Cadeuceus come to mind.

You give us a bazillion things to dodge and a system that doesn't perform well in that situation, we're going to complain... Since the system can't be changed, let's be honest, we'd prefer content that performs well within the system.

---

Simple answer: Left hand doesn't know what right hand is doing, SE tends to be a little more idiot that it doesn't even remember which is left and which is right.
Complex answer: It can be fix, they don't have the brain power to do it easily, and the decision falls into limbo.

Quote:

---

Originally Posted by kukurumei

Simple answer: Left hand doesn't know what right hand is doing, SE tends to be a little more idiot that it doesn't even remember which is left and which is right.
Complex answer: It can be fix, they don't have the brain power to do it easily, and the decision falls into limbo.

---

Except that isn't true at all. If you are to believe the OP, you can't just change the server architecture on a whim. I find that explanation to make a lot of sense.

I think they do design around the positional limitation, because for people that have a good connection to the game there's no problem. Remember how they increased the dodge time for Twintania hurricanes? Someone is looking at that balance.

The thing they likely did not account for is the the extent of how bad some infrastructure is and how poor some people's connections are. That isn't a problem of integration at SE, that was a problem with assumptions very early in the developmental cycle.

Quote:

---

Originally Posted by FinagleABagel

Except that isn't true at all. If you are to believe the OP, you can't just change the server architecture on a whim. I find that explanation to make a lot of sense.

I think they do design around the positional limitation, because for people that have a good connection to the game there's no problem. Remember how they increased the dodge time for Twintania hurricanes? Someone is looking at that balance.

The thing they likely did not account for is the the extent of how bad some infrastructure is and how poor some people's connections are. That isn't a problem of integration at SE, that was a problem with assumptions very early in the developmental cycle.

---

Because the OP is telling us 2nd grade math on a calculus equation. It's true only so far as stripping away 90% of the "exceptions".

Ever hear of Battlefield 4...it's been...in the news for a bit. They are actively fixing and improving their network as we speak. It is possible, it isn't even the hardest thing to do. If it has to be done even in late stages, it can be done.

SE is still 5yr behind the curve, they always have been and always will be at this rate. Bad netcode doesn't make it impossible netcode.

There is this ideal call engineering. Brute force is just...brute force, if you believe that's the only option, you really need to expand more.

Quote:

---

Originally Posted by FinagleABagel

The thing they likely did not account for is the the extent of how bad some infrastructure is and how poor some people's connections are. That isn't a problem of integration at SE, that was a problem with assumptions very early in the developmental cycle.

---

I don't know a lot about network things and whatnot, but I do know that places like Korea and Japan have internet that is leaps and bounds above anything available in the US. To me, a great deal of this problem reeks of SE's inability to see beyond the borders of Japan, for the same reason they gave the entire western hemisphere the same number of servers (in a single datacenter) as one island nation.

It simply never once crossed their mind that what works in Japan wouldn't work for the rest of the world.

Quote:

---

Originally Posted by Naunet

I don't know a lot about network things and whatnot, but I do know that places like Korea and Japan have internet that is leaps and bounds above anything available in the US. To me, a great deal of this problem reeks of SE's inability to see beyond the borders of Japan, for the same reason they gave the entire western hemisphere the same number of servers (in a single datacenter) as one island nation.

It simply never once crossed their mind that what works in Japan wouldn't work for the rest of the world.

---

Sorta sorta, there are very fast connections in japan, there are very slow connections, and there are connections that don't even work for 20% of the year (islands for example shut down on bad weathers).

No different then NA, if not a bit better.

I ping the NA datacenter at 35ms, and I still get stupidly bad titan run lags. Their netcode is subpar, their hardware is probably subpar, and their own estimates were probably subpar.

It's never clear cut issue except the fact that everything needs to be upto 2013 standards. Things like this is expected to be solved, because they have been solved(or mitigated to an extent).

Ever heard of GGPO, one hobby programer show the big corporations, how backwater in their thinking they were, fast forward half a decade, netcode has substantially been improved because one person did the obvious, "use their brains".

The glaring problem really is that these aren't even up to 2010 standards. Seriously, decade old games have adapted better than this.

I grow tired of all the people growing super defensive about this, there's nothing to defend. More often than not, the people that defend it are buying the company line, literally.

Quote:

---

Originally Posted by Steeled

The glaring problem really is that these aren't even up to 2010 standards. Seriously, decade old games have adapted better than this.

I grow tired of all the people growing super defensive about this, there's nothing to defend. More often than not, the people that defend it are buying the company line, literally.

---

The problem is that the game is coded to future standards that our current network infrastructure is not yet (obviously) capable of handling in order to provide a good experience. The server running the show is the way of the future. I expect to see many more games which run the majority of their code executions in a server live-state environment to arrive in the next 5-10 years. However, SE may have been a bit too ambitious in their efforts to lead with this technology given today's typical networking speeds.

Quote:

---

Originally Posted by Sinth

The problem is that the game is coded to future standards that our current network infrastructure is not yet (obviously) capable of handling in order to provide a good experience. The server running the show is the way of the future. I expect to see many more games which run the majority of their code executions in a server live-state environment to arrive in the next 5-10 years. However, SE may have been a bit too ambitious in their efforts to lead with this technology given today's typical networking speeds.

---

Then how do you explain the complete lack of predictive coding both on the client side and the server side, that has predominated the better half of post Y2K online gaming.

Or the backward failsafes for sticking with TCP, that has lead to massive ISP p2p toggling flags.

Or the FF11 like animation sync problems with network congestion instead of rollback syncing

Or how well stupidly bad it is in comparison to many MMOs that run with less hardware, less intrusions, and generally better all around "brain power".

It's not 5-10yrs ahead, that's pure BS. It's just inferior coding and planning.

Just like 2013 computers are far more energy efficient while being more powerful and perform better optimzation then 2000 era models. There is never a time that "well in the future we can brute force it".

No there is never a future that will brute force it. That's not engineering.

This isn't even about SE or FF14 anymore, this is about pure technology and theory. There's no white knighting here.