why is the tcp timeout still 6 sec after all these times?

"

Char1983 wrote:

You are not going to destroy any gameplay if the person has lag spikes over 2000ms on a regular basis anyways. Have you ever played on a connection that has lag spikes of >2000ms on a regular basis? There is no gameplay to speak of to be destroyed then. You just log out anyways.

sure, with lockstep enabled, you clearly notice them now.
in predictive mode i'm not really sure you can differ 2s hiccups from failures in the predictive engine.

You can, trust me.

EDIT: Even if you can't, you don't want to play in such situations. So no gameplay will be destroyed.

If 2 seconds isn't long enough, then 2.5. This isn't something you just armchair, you actually collect data and test it. But the core idea is: as low a duration as possible, without having "lagspike logouts" ruining gameplay.

Then you set the logout delay for the same duration.

Then you rebalance monster damage around the idea that logout isn't instant, but on a short delay. Not a crazy 10 second delay, but maybe two seconds? So spike damage becomes something you react to.

Then rebalance player healing rates. In particular by removing the instant ones.

It's all connected.

"

ScrotieMcB wrote:

If 2 seconds isn't long enough, then 2.5. This isn't something you just armchair, you actually collect data and test it. But the core idea is: as low a duration as possible, without having "lagspike logouts" ruining gameplay.

Then you set the logout delay for the same duration.

Then you rebalance monster damage around the idea that logout isn't instant, but on a short delay. Not a crazy 10 second delay, but maybe two seconds? So spike damage becomes something you react to.

Then rebalance player healing rates. In particular by removing the instant ones.

It's all connected.

Its all connected and it all doesn't change the fact that in PoE you need the ability to instantly logout, especially if you play HC, this is a carryover mechanic left from d2 on purpose. Not sure why your pushing your "bullshit" agenda on this thread when it really has nothing to do with this specific topic, its just minorly related because of "logout"


As for OP perhaps they could make this customizable via a command and tie it to your account. Like set_tcplogout_120 (with the number being in milliseconds of time it takes before it logs you off when it stops receiving packets) Obviously if you set this too low your connection will drop out quite often.

It would be really funny though if I go to Europe and set it to 50ms, and then can never reset once I go back to South America, cause I never have less than 50ms latency here ;).

Goetz, I'm not trying to say that characters (HC or not) shouldn't be logged out in a timely manner after logout is initiated. All I'm saying is that it should be the exact same delay regardless of method...
...if you have a power outage
...if your ISP takes a shit on you
...if you pull out your power/ethernet cord
...if you Alt+F4
...if you logout manually

Instant isn't feasible, but as short as reasonably possible. Perhaps even 2 seconds is too much (1.8?); like I said, it should be tested.

I don't understand how anyone could argue for people who can choose to logout should have their characters remain vulnerable for less time than those of people who involuntarily are put into that situation.

"

ScrotieMcB wrote:

Goetz, I'm not trying to say that characters (HC or not) shouldn't be logged out in a timely manner after logout is initiated. All I'm saying is that it should be the exact same delay regardless of method...
...if you have a power outage
...if your ISP takes a shit on you
...if you pull out your power/ethernet cord
...if you Alt+F4
...if you logout manually

Instant isn't feasible, but as short as reasonably possible. Perhaps even 2 seconds is too much (1.8?); like I said, it should be tested.

I don't understand how anyone could argue for people who can choose to logout should have their characters remain vulnerable for less time than those of people who involuntarily are put into that situation.

And i'm saying you aren't covering every single instance that a player might want to alt+f4 or get disconnected from the game.

As far as "instant" not being feasible its quite quick now unless your having some of the weird connection issues that GGG cannot possible consider all of those possibilities.


Vulnerable for less time is because players chose to disconnect at that time, even in D2 its basically instantaneous when you logout. This isn't about voluntary disconnect vs involuntarily disconnect the issues cannot be clumped together into 1 little bundle, not with the way the game is designed and not with GGG allowing this behavior to exists ever since day 1.

So because all connections are not equal, all PCs are not equal, basically the only factor that is the same is ultimately the same game, we should bring down the people that have the ability to disconnect quickly and punish them for what they can achieve because it isn't "fair" that some have better performance then others?

No, just absolutely not. This isn't even the same TOPIC we are discussing in this thread, again you really are trying to push this idea what somehow you shouldn't be able to alt+f4 if you properly identify a threat you cannot face or an instance in which you would die, that its "not right" you can do that. Its like that by design this topic will go on for pages and pages regarding this which isn't relevant to the main focus of this thread if you want to make your own topic on this feel free.

But poe has too many factors to bring all disconnects to the same time:

spike damage is too high currently to support your suggestion
its an online game, people have issues
its following the same principle\design of d2
PEOPLE HAVE GOTTEN USED TO THE ABILITY TO DO THIS FOR OVER 3 YEARS


You aren't suggesting a small change here, which is why it deserves its own topic, many things have to go into consideration for a FORCED differential of disconnect, which is an entirely different topic.

"

goetzjam wrote:

As far as "instant" not being feasible its quite quick now unless your having some of the weird connection issues that GGG cannot possible consider all of those possibilities.

I mean, there's a lower bound.

The client can't tell the server "I'm done" faster than the distance between client and server over the speed of light at best. Usually longer than that due to switching costs.

For some people, this is 25ms. For others, it's 500ms. The message itself is either the realization of the connection being terminated (aforementioned keepalive) or an explicit logout message.

Living close to a datacenter in the current world gives undeniably massive gameplay advantage, but it's never really been talked about (much like having certain computer hardware).

Forcing all logout paths to go through the same path (i.e. 'logout' from client is disconnecting from the server) and then using an aggressively lower timer (i.e. 2s, or whatever GGG determines minimizes the number of false positives for the userbase) on the connection from the server's POV would remove this discrepancy.

The lower bound and upper bound with those changes would be exactly the same and GGG would have a hard number to start balancing the game against.

Monster rebalancing would need to be packaged with the change, and monster damage over time would need to be balanced against player recuperation (regen/leech/etc) over time, and pack size would probably need to be reduced to reduce a massive generic monster multiplier, and pack size reduction means that AoE skills need to be rebalanced, and...

---

I feel like it's not a problem they can solve with the devs that they have now. Not enough people, not enough time. If it was just a Jonathan problem (like implementing lockstep was), that would be one thing, but this would touch almost every facet of the game.

And it's something that they probably could have set on day 1 but did not possess the foresight to know how big of a choice it would eventually become. That early decision made a bunch of other decisions for them, and undoing it is almost as hard as rewriting the entire game all over again.

---

A strict decrease of the upperbound (from 6s to lower) isn't quite the same as the above suggestion to "increase the lowerbound (from 25ms to 2s) and make the upper and lowerbound the same", so as far as this topic is concerned, there's still goodness to be had.

I like the idea of having lower and upper bound be the same. I really like it, for a variety of reasons.

By the way, when I say "logout timer" I mean "from the immediate previous communication," not from the logout command itself.

Examples with 2 second timeout:
* Player A has decent ping and sends a logout command. When the server received the logout command, the immediate previous communication was received 80ms ago. The server waits 1920ms then logs out the character.
* Player B is having issues with lag spikes. When the server receives B's logout command, the immediate previous communication was received 1100ms ago. The server waits 900ms then logs out the character.

In short, when I say "2 second logout delay," I really mean "2 seconds minus ping."