[indiana-discuss] OpenSolaris on really old Sun hardware?

[James Cornell]

przemolicc at poczta.fm wrote:
> I don't understand this approach - the slowest PC-like tower/desktop box
> can also have OpenSolaris installed on it but will be much, much faster then the old Sun workstations.
>
> Regards
> Przemyslaw Bak (przemol)
>   
It has to do with culture and knowledge of being able to write programs
for SPARC and Solaris, which is held quite high to some.  RISC vs CISC,
and nostalgic reasons also, in addition to just simply knowing the
differences in implementation for the tool chains, from Sun Studio to
GCC and to other software as well.

Anything actually modern is out of reach by regular income people.  I've
owned a few secondhand SPARC machines which were suitable at the time,
but since Sun hasn't updated since then the gap is closed where you
could get away with using them.  I'm specifically talking about Ultra-2,
Ultra-10, Ultra-80, which were suitable granted you have 1GB of ram when
Solaris 10 came out in '05, but now things have flipped and those
machines are the last now low-end to work with SXCE, with anything less
being a non-option, and the Ultra-2/10 being questionable.  The Ultra-80
with quad 450MHz USII is workable still, but is heavy on power needs
with its 670w PSU allowing it to use that much under load, plus very
slow when stacked against an Intel x64 processor with only two cores.

Nonetheless, Sun systems still in some cases have more threads going for
them on the low-end, and some people want to implement their software on
them.  It's still possible with those mentioned, but expect support to
disappear around the end of Sun's Solaris 8 legacy extension program, or
by 2012, especially if they end up laying off more people.  I question
my ability to get my DVDRW repaired in my Ultra 20 right now with how
things are going, so it's a good thing I didn't opt for Solaris support,
it doesn't seem to be well-backed, and probably never was.  Most of us
here support ourselves of course, so it doesn't matter.  Thankfully
OpenSolaris will live on, as long as past contributers can work out the
blobs in the build system... but only ISC would host such a thing as
they are one of the more open mirrors versus university networks which
only have some of a clue about OpenSolaris, let alone SPARC, if at all.

I'd bet you that without Sun in the loop, we could bring back legacy
support for these machines, and spin off an optimized version... the
problem right now is that this doesn't mesh well with Sun's intentions
to kill off needing to support these machines, so maybe if past SPARC
engineers working for OpenSolaris on their spare time can make a case
for such a thing to set investors and managers at ease.

The boundary to draw is what is too slow, what is too old, and what can
be supported which is in common use, and is still available to buy
through alternate channels, regardless of Sun's intentions.  This
requires a fork of OpenSolaris because they won't let it happen, and I'd
suspect MartUX (Martin Bochnig) is the only one that qualifies to take
over and absorb such a duty.  With back ports from Sun's continuation of
support for SPARC on servers, which are compatible, but specifically
intent on disabling our ability to use older machines, simple
modifications, and further work could be done.  As for myself, there's
no way I can afford to keep a SPARC machine powered here unless it's
less than 220W nominal usage, which is hardly ever the case, bills just
accumulate to an extent which is not possible when dealing with more
modern SPARC machines with more processors.

Sorry for such a long reply, but I'm simply stating the fact that Sun
cutting the workstations makes sense, but it doesn't mean the end, and
it certainly is not correct to say such a machine is unsuitable for use,
let alone for OpenSolaris, which I find to be not the case when
interacting with such machines.  For a desktop, it's certainly a hard
sell, but they are very much usable for workgroup and small frontend
business servers or for databases.

Some people just get enough CISC during the day they want to puke, and
one should not discount that some would rather spend their cash on a
SPARC machine for the same price, even if it's only half the performance
according to benchmarks such as SPEC, as there are other merits within,
including historic and inquisitive necessity to understand something
which many feel is a niche, granted it is.

Indiana is less of a monster in resource usage, minus the ZFS overhead,
so it's suitable for these machines compared to SXCE.  If one could
still use UFS on Indiana, it would be more than workable on them.

- James