[OpenIndiana-discuss] What belongs in a general purpose O/S?

[Joshua M. Clulow]

On Wed, 20 Aug 2025 at 00:22, Carl Brewer via openindiana-discuss
<openindiana-discuss at openindiana.org> wrote:
> What actually belongs in an operating system?
>
> I see, or at least I think I see, a lot of effort on various platforms,
> to maintain applications (gimp? really? bundled?!), and when humans are
> scarce, is this at the expense of device drivers, installation systems
> and so on?
>
> For a little while, I recall one of the *BSD ports systems

Ultimately ports is just a separate consolidation of software with its
own build system.  This isn't really better than OmniOS or OpenIndiana
are doing with installable binary packages of software, it's just
different.  There's even a similar split between core pieces, which
mostly come from illumos-gate and are generally common between
distributions, and other software like web servers or VirtualBox,
which tend to come from other consolidations.

> good pathway to take, but now we seem to have this weird IPS thingo and
> all the barriers to entry that it introduces.

It's not really clear what barriers you're talking about.  Building
IPS packages and running repository servers is not actually that
complicated, and there are several mature existing frameworks for
doing it en masse; e.g., oi-userland and omnios-build and
omnios-extra.  It's true that performance could be better and memory
usage could be lower, but those are questions of degree, not kind.  It
is otherwise a robust, modern packaging system that integrates well
with ZFS-based boot environments for atomic updates.  It has first
class support for managing various core illumos facilities like the
driver binding database and SMF services.  It continues to be
maintained and improved by folks in our community.

> "we" spend a lot of time farting around with applications that really,
> aren't up to a niche community to support. I use VirtualBox on my O/I
> servers, but should it be bundled, or something that we get with source
> and compile ourselves?

Asking users to compile their own software doesn't make a lot of
sense.  It feels like the core question you're asking is about if we
should be maintaining a bunch of packages with the scarce resources of
a volunteer community like the one surrounding OpenIndiana -- but
surely if everybody has to spend more time compiling and patching and
maintaining their own software from scratch, the community will have
even less time and energy to do anything else?

If somebody is willing to compile some software that's personally
important to them, it's really not more work for them to just write
the build recipe and publish the package.  The community benefits,
obviously, but _that person_ also benefits the moment they have to
install that software on a second computer, because the work is, in
every sense, already done.

It's also easy to see everybody agrees that having more drivers would
be good!  The thing is, the only way we will get drivers is if you
pick a piece of hardware you'd like to use, and _write a driver_.  If
you have hardware for which a driver is not currently available, the
best thing to do is to start working on that driver right now.  The
kernel is not magic; it's just a big C program.  Anybody who can
compile random bits of software on a UNIX system can surely learn how
to do it, especially with help from other resources in the community.
The most important question to ask is: if not you, then who?


Cheers.

-- 
Joshua M. Clulow
http://blog.sysmgr.org