[TUHS] Re: AIX moved into maintainance mode

[Kevin Bowling <kevin.bowling@kev009.com>]

On Thu, Jan 19, 2023 at 11:23 AM segaloco via TUHS <tuhs@tuhs.org> wrote:
>
> Bringing it back around to AIX, this may be a bit of a leading question, but for those who are more in the know than not on how AIX works, is there any chance there are little design nuggets hidden down in there that now, not being critical to an active IBM project, may find their way out there into the world?
>
> This is kinda my latent curiosity with any of these commercial systems, if there's something absolutely amazing hiding down in one of the codebases just waiting to see the light of day in some post-commerical source release that might improve the situation out there in open source UNIX-like land.  Some ideal SMP scheduler, quality drivers, etc.

There's definitely a lot of nice stuff in there, most "reviews" or
recollections of AIX distill down to "OMG ODM" and completely miss the
forest from the trees.  I've seen some data and I think AIX scales, at
least out of the box, much higher than Linux on massive CPU and NUMA
systems.  That may be a bit of a cause and effect, there's a
likelihood of someone buying a top range POWER system to run as a
single system image.

However OpenSolaris provides a sober analysis of what you are after:
1) how much work is involved in open sourcing a commercial UNIX.  The
then CEO blogged about how he had personal involvement to get it done,
and the high hurdle was lawyer work getting rights and approvals from
third parties to re-license everything.  AIX has, in addition to
whatever vestiges of Bell/AT&T code, Bull and Motorola code and
probably a lot of others (OSF, HP, Sun, etc are mentioned in the
copyrights on install) whose rights may not exist in any recognizable
form after 40 years.

2) how little is generally applicable outside a native environment.
In terms of code, the main contribution from Solaris today outside of
it is ZFS.  The ports are all a bit of a side car (even in Illumos),
Larry has pointed out on this list how hard they worked to get unified
memory in SunOS only to have that lost again by ZFS.

As some random anecdote, in FreeBSD there is a reference to Solaris
prior to OpenSolaris here
https://cgit.freebsd.org/src/commit/sys/kern/subr_turnstile.c?id=961a7b244dbfc467c112b7a594825da8c0a41acf
and FreeBSD and macOS also use OpenBSM inspired by Sun.  The BSDs
occasionally share some code (usually drivers or higher level
subsystems like ufs or pf) but as time goes on it seems more
ideological than code since they have all drifted sufficiently from
one another.  So close study of any interesting AIX bits would be
about as useful as code for cross pollination.

On the other hand, source dumps are great for historical records and
study.  For instance, the heirloom tools
(https://heirloom.sourceforge.net/doctools.html) came from the work
done to open Solaris.  If you get source control repos, it also gives
you a nice whodunit.

> Of course, the usefulness of any such thing would depend on any theoretical eventual license applied to a source code release.  Something restrictive would prevent proliferation of a good idea, but in any case, there are so many lineages just ripe for plundering, and as time goes on, it becomes more likely those source codes will actually be accessible and licensed to allow that.  Who knows though...
>
> - Matt G.
> ------- Original Message -------
> On Thursday, January 19th, 2023 at 9:19 AM, Adam Thornton <athornton@gmail.com> wrote:
>
> The era of general-purpose computers won't end.
>
> The problem is that a great many single-purpose items are (and increasingly will be), for reasons of scale/developer availability/familiarity, general-purpose computers that come from the factory supposedly packaged to do only one thing.
>
> But all of them will have brains that will let them do arbitrary things. Some of these things will be done at the behest of the organizations controlling the society where the developers come from. Some of them will be done at the behest of transnational organized crime rings. Some will be done by enthusiasts. But I don't think we are too far from the world where you can't trust your toothbrush unless you carved it yourself from a stick with a knife that's been in your family for generations.
>
> But really, this is all just "Reflections on Trusting Trust," which was, what, 1984?
>
>