Dennis Clarke writes:
> I am glad this has been said. I have a whole slew of old hardware all the
> way back to the Sun SPARC 670MP and a big ol' DEC AlphaServer from 1996.
> They make sort of cool fridges if you buy a bar fridge and skin it. Other
> than that I have no idea why anyone would care anymore. My quad Ross Sparc
> 20 serves a purpose on very rare moments when I want to test something
> obscure. Then I power it off again. Even with a vintage Solaris 8 support
> contract it can not be patched anymore .. sun4m just isn't supported
> anywhere anymore. Why would it?
Please don't blow the US-I revival out of proportion: I've never suggested
the resurrection of several kernel architectures that (apart from the
short-lived Solaris 8 Foundation Sources) have never been opensourced.
Contrary to that, the effort is minimal: just add a couple of symlinks and
remove the code in {ufs, inet, wan}boot to reject US-I machines as
unsupported. If you do this, they work just fine in both Solaris 10 and 11
to date.
If it were only the symlinks, a separate package with them could easily be
hosted elsewhere. But since shared code is affected, we run into a
problem: I don't think IPS can properly deliver the same file from
different packages. Even this trivial case suggests that support for older
hardware might require a full-fledged fork of ON (and maybe even a
different build of other consolidations; I don't know for sure). All this
just to remove a single error and add some symlinks which Sun doesn't like?
I fear Jim Carlson is right in this case: just a repository for removed
drivers won't be enough to handle legacy systems.
> Put NetBSD or OpenBSD/FreeBSD on them or even run Solaris 8.
It seems ridiculous having to move to a different OS if the native one
would work just fine *with minimal effort*.
> But Solaris Nevada/OpenSolaris ? Not even in a drunken moment in some
> geeky bar in Silicon Valley would anyone scrawl that crazy idea on a
> napkin.
Quite the contrary: those machines make excellent timekeepers. Just attach
e.g. a GPS receiver to them with current ntpd. Newer machines are
increasingly problematic since they have serial chips with large fifos
which make for terrible timekeeping. On se(7D), it would be possible to
turn them off according to the specs, but unfortunately, those sources are
closed.
Rainer
-----------------------------------------------------------------------------
Rainer Orth, Center for Biotechnology, Bielefeld University
