And here's the other thing. I do not want my PC to be turned into a crumbly version of a Playstation with a halfassed interface and tiny text. I don't want it to be turned into a good version of the Playstation. I don't want it to be a Playstation at all, meaning I don't want forums, voice, text chat, friend groups, the controller, or anything else to be owned by a single company with vendor lock-in. That's why it's a PC, not a console. I don't need their launcher, it's called "Putting your game shortcuts in a folder called "Games."
Steam has features like big picture mode because not everyone is playing at a desk. I alternate between my desk and couch depending on the game, Steams interface is quite nice for this. Having a company with heavy influence on this that actually gives a shit is very nice for actually getting things done. Biggest missed opportunity is a lack of a plugin system. People have done great work on the Steam Deck and I wish I had that experience on the desktop.
Windows is around half of the global server market share, and yes, there are some VERY large deployments out there.
Windows server primarily exists in a specific market that is better served by and currently being moved over to SaaS.
Its difficult to actually compare numbers due to radically different deployment methods. (How are we counting a large scale k8s deployment? Methodology will heavily skew mumbers in one direction or another) And I find it makes much more sense to compare where they are deployed and why.
Windows server is used to host internal applications, Active Directory, and Exchange in most cases. Its popular in SMB, with large enterprises also using it for these specific functions. However, most of those are in a slow process of moving to SaaS, as its easier for traditional enterprises to account for in their financials, and its more profitable for Microsoft to sell Entra and Exchange Online licenses. Same for many of the traditional vendors.
The actual software companies serving applications to massive userbases, both B2C and B2B, are all using Linux for this. It scales better and they require real orchestration. Windows is simply way too stateful in its design to accomodate those workflows.
Updating a Linux major release (e.g. RHEL 9 - > RHEL 10) is basically guaranteed to break your entire stack. Windows isn't perfect, but it isn't even in the same universe of badness as Linux
Have not had issues with this personally, but I also tend to avoid traditional "pet" servers. I could definitely seeing this being an issue for traditional "run it on a box" enterprise software, which needs to seriously update its deployment model.
We have, and the sheer volume of resources consumed to keep Linux from shitting itself makes me wonder if we're really saving any money at all over the Solaris days. How many resources are consumed by the OpenShift ecosystem, which is little more than a workaround to Linux?
Interesting question that really depends on the deployment imo. The traditional model used back in the day doesn't scale and is less stable than the k8s baseline we have now. If you have a well designed architecture, properly configure your scaling parameters, and actually rightsize your workloads so you arent wasting overhead, you can absolutely pay less and do so with far, far fewer people.
With that said, it assumes alot of proper housekeeping that large businesses completely fail to do. They're filled with devs who will overequest resources for their application (which is amplified the more replicas it had), stateful apps that arent properly configured to scale down, and manual processes to solve problems that were addressed by existing primitives because the person who designed it didn't actually understand how this technology works.
Solaris and BSD were realistically better successors to Unix with the traditional server model, but I think the faster moving pace of Linux ultimately won because it enabled the development of more novel systems, which were necessary for hyperscaling. If we were moving at a slower speed, it would've been more difficult for Google to develop Borg, which was fundamental to their growth. We probably wouldnt see other new solutions that have been built upon Linux, such as NixOS (fully reproducible, violates certain Unix principles to address dependency management issues), Talos Linux (minimalist, secure distribution that can only run Kubernetes), or general purpose immutable OS's. The traditional method of server hosting was insufficient for the direction things were headed, and Linux happened to be the most capable of adapting so something new could be built.
