Concrete proof finally provided that the 2002 error has NOTHING is server side issue

[Larirawiel]

There are more propabilities:

There are not aware of this behavior because that person who implemented it is no longer in the company and this behavior is not documented.

They know this but maybe this behavior mitigates another bug. Even if it is hard imaginable.

It was a quick & dirty hack and now they have not the ressources to fix this.


Cheers

[Packetdancer]

They know this but maybe this behavior mitigates another bug. Even if it is hard imaginable.

It's very easy for connections passing through NAT to end up silently dropping; the client side doesn't see that the connection is dead until it tries to send something, but the server side sees that the client has disconnected. It's due to how a number of home routers/modems will handle NAT.

If they saw an issue back in ARR with the connections from behind a NAT setup dropping, I can easily see adding a safety measure where it refreshes the connection every so often to make sure the connection doesn't silently go 'poof'. I'm not sure it's the approach I'd personally have used in a scenario like that, though FFXIV's network code may have made it the most appropriate one for whatever reason, but it's an approach to things like that... and one I've seen used by other software elsewhere. So if the game is doing that, it absolutely could be an attempt to fix that issue.

After all, the 2002 errors are frustrating, but it'd be even worse if your connection got dropped silently and you sat there thinking you were still in queue, only to get unexpectedly disconnected the next time the game tried to refresh the connection. Holding your spot in queue and manually disconnecting you to make the connection get refreshed strikes me as the equivalent of going after a particularly persistent mosquito with a howitzer, but it will (generally) get rid of the mosquito.

And even if that isn't the reasoning, it could still be addressing something else. it could be that it's a load-balancing scenario, where when it bounces you it has the potential to ask you to reconnect to a different login server instead, and that provides a method to shuffle people around to keep any one login server in a pool from being too overloaded, or whatever.

But if that's the situation, I'd argue that still falls into 'archaic netcode' as listed by Catwho, or a closely-adjacent area. And is probably something where reworking the entire login system (and making sure the change it doesn't re-introduce whatever original quirk prompted the scenario) would likely be a not-small undertaking.