The success hidden in Stadia’s failure

This article isn't about the Stadia game platform itself, but rather the technical capability it represents. In brief, Stadia technology is misapplied. It should be an application platform and not a game platform. I am not optimistic that Google will figure this out, but someone else should.

Stadia, Google's new streaming gaming service is here and the response is generally negative. Technology personalities, such as those on Youtube, are declaring it dead on arrival.

Stadia may be failing as a game platform, but it has successfully demonstrated that hardware and network capacity is sufficient to create a computing model I call Personal Application Omnipresence (PAO). PAO would provide omnipresent and cross-device access to personal applications running expressly for the user. In many ways, PAO is basically the same notion as Stadia. But PAO is for any application—not just games.

PAO is what someone, likely someone other than Google, should be building.

You can read the full description of PAO if you're interested, but it's fairly easy to describe by listing the ways it is similar to Stadia and the ways it differs.

Stadia is like PAO...

  • Personal applications are running on a compute server. Essentially, a compute server is dedicating an application session and process instances to each user (or "player").
  • Applications are omnipresent in that the singular running instance of an application for a user—that is, the singular process instance—is reachable from anywhere, from any of that user's devices.
  • Applications adapt to device capabilities. The same game can be played on all of the user's devices, even though each device has different capabilities. Whether this is done by reducing the application's inputs and outputs to a lowest common denominator or by adapting to the capabilities of the device(s) dynamically is left to the application (game) developers. It's basically "responsive design" in the modern web-dev sense, but applied to a native application. Similar to Microsoft's UWP.
  • The technical means underlying the whole experience is user interface streaming, specifically video streaming.

Stadia is not like PAO...

  • Stadia only exists as a third-party service. PAO would be an operating system and application platform that individuals would be able to run themselves on their own compute servers or on server capacity they lease. In modern parlance, PAO would be "self-hostable."
  • Stadia does not provide session multiplexing. That is, Stadia does not allow multiple client devices to be viewing and sending inputs to the same game session concurrently. Obviously that makes sense (to a degree) for a gaming platform since without expressly designing a game to work with multiple concurrent inputs, providing input from two devices simultaneously would be like connecting two mice to a PC and moving both with your two hands at the same time. However, for an application server environment, allowing multiple concurrent sessions provides a lot of fringe benefits such as immediate and seamless transitioning between multiple devices, concurrent consumption of media, an implicit ability to present, and so on.
  • Most significantly, Stadia is only a gaming platform. PAO would be for all application types, which could include games, but also email clients, messenger clients, browsers, office suites, graphic design applications, CAD, video editing applications, etc.
  • A PAO compute server would also host the user's data and files (or at least see those files as local to it). A singular file system, potentially with federated backup, would provide omnipresent access to files across all devices, without synchronization (it's singular).
  • PAO would probably use something more like remote desktop protocol as this is optimized for user interface streaming, as compared to general-purpose video streaming. But perhaps it would adapt to the application, using RDP-like UI streaming for applications and video streaming for games and media in general.

The hidden “success” of Stadia

When I described PAO in 2012, I acknowledged that a challenge is network availability, bandwidth, and latency. Snark about "negative latency" aside, Stadia has proven that the network challenge is surmountable. This is especially true when you consider that most applications are not as bandwidth hungry as video games.

I doubt Google will be the company to realize this vision. So this hidden success is not likely to be reaped by Google. PAO is a decentralizing force, meaning it's not aligned with Google's goals and revenue streams. But some company should realize that these game streaming platforms are just the first use-case for a holistic new computing model. They can call it PAO or whatever they want. I just want it, ASAP.
About this blog