Re:Netflix might get into cloud gaming, forms another new ga...
by andrewmcwatters on Wednesday October 19, 2022 @07:38PM
(#33269172)
In software engineering, you should account for your failure modes that occur the most frequently. In cloud gaming, the default is to run overbudget for every single frame of gameplay, 100%.
The entire business model is a failure mode. Let's ignore that for a moment.
Assuming gameplay and network conditions are perfect, you're still over frame budget. So out of the box, you're paying for a substandard experience. Let's say you accept that because you're a poor college student (most likely, I'm guessing), or insert lifestyle and customer purchasing decision making here.
So, what happens next? Well, if we start thinking about the ideal customer profile for these things (people not wanting to spend basically any money on games, while getting as much per dollar experience they want), you're probably also talking about people who don't have gigabit Internet connections.
What happens in a real-time environment when your feed drops? Just guess for a moment. It's like putting the controller down. You straight up weren't playing the entire time your packets weren't being exchanged.
Except, wait, for the past 23 years, real-time video games have been built with specific network logic to accommodate those same consumers and the realities of poor networking.
But wait again! Those clients are sending commands to a DUMB TERMINAL. This was the exact reason client-side prediction was invented. And your packets dropped. Your video feed cut out. Input wasn't being sent. There is no way to reconcile that from a cloud gaming service. You can't rewind video. You can't send input late.
So, 23 years of enjoying late 90's innovation was just flushed down the toilet. Because there is nothing about cloud gaming's architecture that allows for prediction when the entire service is a dumb, high-definition terminal.
The entire business model is a failure mode. Let's ignore that for a moment.
Assuming gameplay and network conditions are perfect, you're still over frame budget. So out of the box, you're paying for a substandard experience. Let's say you accept that because you're a poor college student (most likely, I'm guessing), or insert lifestyle and customer purchasing decision making here.
So, what happens next? Well, if we start thinking about the ideal customer profile for these things (people not wanting to spend basically any money on games, while getting as much per dollar experience they want), you're probably also talking about people who don't have gigabit Internet connections.
What happens in a real-time environment when your feed drops? Just guess for a moment. It's like putting the controller down. You straight up weren't playing the entire time your packets weren't being exchanged.
Except, wait, for the past 23 years, real-time video games have been built with specific network logic to accommodate those same consumers and the realities of poor networking.
But wait again! Those clients are sending commands to a DUMB TERMINAL. This was the exact reason client-side prediction was invented. And your packets dropped. Your video feed cut out. Input wasn't being sent. There is no way to reconcile that from a cloud gaming service. You can't rewind video. You can't send input late.
So, 23 years of enjoying late 90's innovation was just flushed down the toilet. Because there is nothing about cloud gaming's architecture that allows for prediction when the entire service is a dumb, high-definition terminal.