Re: Dropping sparc32 for lenny
Martin wrote:
> 1. There are still sparc32 systems in production and in use; but what
> proportion of them are 1. publically available (as in COTS rather than
> Verliog like Leon or the UltraSPARC T1), 2. run Linux, 3. run Debian?
> If they care about having a Debian/sparc32 port then where are they?
> And why aren't they offering some support in keeping it running? For
> developers with experience of custom designed embeded systems, helping
> maintain kernel code shouldn't be too difficult.
We also have to ask ourselves what Sun's position is on this. Granted that I
only have tenuous contact with them but it has been put to me (a) that they see
no advantage in supporting or even providing documentation for older
architectures and (b) that their contractual relationship with SPARC
International prevents them from from making any information available or
offering detailed support. To quote an exchange elsewhere:
"From a blunt point of view, why would any hardware seller be interested in
getting anything working on anything that isn't in the current product list?"
"Because most people, particularly if starting with a totally unfamiliar
architecture, are going to start off with something as cheap as they can
possibly manage. Sun can either look at the substantial amount of second-user
kit as a blessing- giving people a chance to take a look at the architecture for
minimum investment- or a curse, but they /either/ have to keep first-time users
happy irrespective of whether their machines are new or second-hand, /or/ buy
back every scrap of kit that's made redundant to make sure that the fact that it
doesn't work reliably with one of the OSes they promote doesn't besmirch their
reputation. That's why."
My own feeling is that their attitude is a phenomenal mistake: Sun's making sure
that the very latest devices are properly supported isn't going to do them a
blind bit of good if 95% of the machines out there aren't reliable.
With the best will in the World I want to see SPARC survive as an alternative to
x86- there has to be /one/ and PPC appears to be on its way out now that Apple
has jumped ship. But if people can't get an arbitrary machine, stick in a
mainstream distro CD and know it will work what chance does it have? How come
PCs- with their poorly-defined architecture and lack of agreement across the
industry- can be made to work reliably but Suns can't?
Sun are currently in the position that DEC were when the PC came out. They're
producing kit which, all else being equal, would blow away the competition, but
they're ignoring the fact that most projects start with somebody getting his
hands on something he can afford out of petty cash (or at least out of his own
budget) which in practice these days means either a PC, a second-user SPARC or
possibly a Mac. If second-user SPARC systems simply don't work reliably they'll
get thrown away and replaced by PCs, and at that point Sun lose friends.
/Why/ are there problems with sparc32? Because nobody really understands the
architecture except Sun. Why are there MMU miss issues on sparc64? Same reason.
What chance do we have of a fix? Slim since Sun says the hardware's obsolete.
Who does this affect? First time users, and casual developers who want to test
that their code works on SPARC. What will the result be? Development tools
languish, Linux on SPARC dies.
Maybe I'm being pessimistic, and in a few months time everybody finds that
they've got enough cash in their pockets to go out and buy T1 systems.
--
Mark Morgan Lloyd
markMLl .AT. telemetry.co .DOT. uk
[Opinions above are the author's, not those of his employers or colleagues]
Reply to: