On 24 Sep 2025 at 10:34p, hyjinx pondered and said...
Now it's a sea of beige, and a sea of beige operating systems. Even
Linux has ceded ownership largely to a few big corps when it comes down
to it, excluding the kernel, for the most part anyway. But even that has parts coded by IBM, Microsoft, Google and others. Nothing is free any more. No more cathedral and the bazaar.
I think that's a bit of a simplification. IBM, MSFT, and Google
contribute to Linux because it's in their strategic best interests
to do so, but that doesn't mean that Linux is any less "free" than
it was before.
What _has_ changed is that the barrier to entry, and overall cost
(in terms of time, energy, politics, etc, but not necessarily money)
has increased dramatically. BigCo contributions to Linux seem to
dominate so much because those companies have the resources to
sponsor engineers shepherding their changes through to integration;
but a lot of other folks do not. So the biggies can afford to do
it, and passionate individuals do it just because, while a lot of
smaller organizations cannot justify the expense.
But even within the ranks of the big boys, there is discontent:
for a working engineer, time and energy (and political capital) have
a very direct relationship with money, so often it's easier to just
float a patch in your local repo than upsteam a change. And that
causes real problems: when I was on Google's kernel team, we had an
_enormous_ patch set and it took a very long time to rebase onto an
upstream release.
Why not upstream all of that? A great question, with a few different
answers. One is that some of it couldn't; some stuff had been done
in collaboration with a vendor, under NDA, and Google was legally
barred from sending that code upstream. Some was because, even
though there was no significant intellectual property concerns, code
might be so Google-specific that it didn't make sense to send upstream;
much of that is historical baggage, but getting rid of it takes time.
But probably the biggest reason was that it wasn't economically viable
for a lot of stuff. Google might make a change that was a win, but
for a specific, constrained use-case. It may be cool to upstream, but
when it's sent someone looks at it and says, "yeah, this is neat, but
it only works for n=1; you should generalize it for any n." Except
that doing that generalization might be 10x the work of the current
patch: the engineer can't justify the investment because it provides
no additional value to Google, so it's easier to just float the patch.
Of course, over time, that decision is more expensive than doing the
work and getting the thing upstreamed, but we're talking about a 5-10
year timeline here.
Anyway, yeah, Linux is as free as ever, but the days of Torvalds
taking any random code and integrating it are certainly over. The
bar is much higher, and that cost can be borne by the Googles,
IBMs, Amazons, Metas, and Facebooks of the world, but not so much
by the smaller players, let alone individuals who are more
interested in casual contribution.
--- Mystic BBS v1.12 A48 (Linux/64)
* Origin: Agency BBS | Dunedin, New Zealand | agency.bbs.nz (21:1/101)