Theory: in a world of SaaS/public cloud, “open source,” is an implemented standard, instructions for plumbing, not the actual product sold.
The actually product (the thing sold, paid for) includes running the commercial software, storing the data, managing how the software is used. It can also include doing the unseen, tedious Morlockian work of making it “enterprise grade,” even “cloud grade.”
Giving away the open source stuff lets the commercial company (the SaaS) define and control its 3rd party dependencies – here, how New Relic’s data is collected and modeled. That is the blood and oxygen of a monitoring company: no monitoring data, no company.
As a strategic bonus: open sourcing things allows the SaaS to kill off (perhaps “limit,” to be less hyperbolic) areas of the market that rivals could own, differentiate, and monetize.
Kubernetes does this (IaaS becomes low value, table stakes, undifferentiated), not sure about Linux. Linux is something different: it is the product, not the standard.
There is a major flaw in my reasoning, viz. Kubernetes: all of the code to make it actually work below and around kubernetes can be proprietary, usually. The software defined networking (despite years of trying to open source this awhile back, it remains, largely, inaccessible alchemy), the security integration, the PaaS layer (though, actually, here, another instructive confounder to my present theory), etc.
But, perhaps, this illustrates my theory: those parts of the market have been defined as the place that competition can occur.
You would then expect rivals to open source those things to damage (pardon, again: “limit”) their rival’s competitive advantage, while keeping their own stuff closed. (“Closed,” of course can also – usually in public cloud – mean “only available as a running service in one of the big three clouds or a the SaaS in question.”)
Of course, all of the rivals can agree to give up monetizing parts of the market and all open source that part, all giving up on making money off it. This is usually done somewhat by accident, or with one large competitor reluctantly throwing in the towel and giving up after a long fight, as with HTML, RIA’s (Flash vs. HTML5 vs. Androind/iOS – not only open source here, but very much industry de facto standards),
Docker, and, of course, kubernetes. OpenStack was a failed instance of this theory: the public cloud companies would not see one of their core assets open sourced and made non-differentiating. Also, the whispers are that the OpenStack community has its own problems, not least of which the inability (and the lack of will) for its primaries to spend billions on building their own Datacenters and global networks out.
Build tools are an odd river in this terrain that I don’t know well: has Jenkins being open source really limited the commercial opportunity of commercial build tools and CI/CD?
And security – always security: somehow impossible to fully open source as in drive out all the closed source and monetizing possibilities.
(Also, Hyperic tried this open sourcing the systems management stuff. Perhaps too early for it to be successful, perhaps getting acquired just halted the corporate enthusiasm needed to pursue that course. That was too long ago to remember, recount, and reminisce. [Nagios was that deafening dull-thumb of the GPL death-cult, where “life” is a vendor’s ability to make closed source profit margins and scale top-line revenue to billions.])