[illumos-Developer] ON is bloated and a bunch of stuff that doesn't need to be there should be ripped out (was Re: Heads up: perl 5.8.4 removal)

[Aram Hăvărneanu]

On Fri, Nov 12, 2010 at 7:33 PM, Garrett D'Amore <garrett at damore.org> wrote:
> That said, I will point out that we probably *could* remove this from
> the x86 tree without contention... I don't think *anyone* has x86 a.out
> binaries that they need.

I should also remind everyone that Sun386i binaries don't work in
current Solaris or illumos. The only a.out binaries that work in
Solaris on x86 (if any!) are binaries that were compiled in the ELF
era, so in a way, they were always obsolescent. No a.out x86 software
should exist, really. SPARC is a different matter, though.

>
> Upshot: those of you still needing this, be prepared to offer compelling
> evidence supporting your case, or to modernize, or to maintain older
> systems with older operating systems to keep your legacy on life
> support.  It won't go on indefinitely in Solaris, nor in illumos.
>

Without trying to be rude, I'd recommend Solaris 10, Solaris 8 or
NetBSD with compat_sunos(8). It's not like illumos doesn't care about
your programs and sends you away, it's that we are a project with
limited resources, and at the moment this resources are required for
crucial things. NetBSD has a very different goal than us and it is
also reached a certain level of maturity and it allocates more
resources for compat_sunos(8) then we could ATM.

SunOS 4.x compatibility is not just a.out. That's just the top of the
iceberg. During SunOS history there were many incompatible standards,
many formal, like XPG, many informal, de facto standards -- things
that people just did in a certain way and SunOS did best (IMHO better
then other vendors) to implement all this standards. Because the
standards are either incompatible or mostly compatible but with
crucial, very subtle differences it was very hard to achieve this
level of support. The cost of maintaining all this legacy can only
increase in the future as we need to support newer, better stuff and
we need to be careful not to break the old stuff. Legacy stuff stands
in the path of innovation as new stuff has to conform to certain
complicated ways of doing things and constraints in order not to break
the old stuff. While a.out is not much code and doesn't have
complicated interactions with other code, SunOS 4.x compatibility as a
whole, does. SunOS 4.x is vastly different from current reality and
requires a lot of effort to maintain. Probably about the same effort
as the effort required to support Solaris 8 software and there's
probably 10,000 more Solaris 8 software in existence then SunOS 4
software. Paying this price doesn't make sense for a project with our
resources.

SunOS 4 is 22 years old. I understand that there are people who still
want to run software this old, but there are alternatives for it.
Developer time seems to be better spent doing things that benefit more
people.

-- 
Aram Hăvărneanu