Jonathan Schwartz replies to Linus regarding ZFS and GPLv3 [LWN.net]

uncooperative on specs

Posted Jun 13, 2007 16:53 UTC (Wed) by qu1j0t3 (guest, #25786) [Link]

Yeah, Theo describes a deplorable lack of cooperation. Apple is no better when it comes to their hardware.

On the plus side: Sun and Apple offer superb O/S and machine integration.

On the minus: Despite being fundamentally hardware companies, they seem paranoid about spilling secret sauce. Releasing specs for your proprietary hardware can only improve interest in it, you'd think, because it would mean much more solid support from Linux, *BSD, etc.?!

Sigh. I hope this public exchange helps thaw things - for everyone.

you forget ...

Posted Jun 13, 2007 17:38 UTC (Wed) by JoeBuck (subscriber, #2330) [Link]

That Sun is, as far as I can tell, the only major hardware producer to release the full synthesizable RTL code for a state-of-the-art processor (UltraSPARC T1) under the GPL.

http://opensparc-t1.sunsource.net/download_hw.html

you forget ...

Posted Jun 13, 2007 17:43 UTC (Wed) by ewan (subscriber, #5533) [Link]

It's not being forgotten - the Theo comment in question says:

"Sun gets great press out of UltraSPARC being all "open", but what use is supervisor-mode documentation when the rest of the chips that the supervisor-mode code has to communicate with are entirely undocumented???"

Which seems fair, and adds to the idea that this is all 'Freedomwash' - a thin skim of freedom intended to improve Sun's superficial appearance without actually helping anyone.

exactly

Posted Jun 13, 2007 17:48 UTC (Wed) by qu1j0t3 (guest, #25786) [Link]

It's completely unrelated to Theo's objection - which still stands.

totally unfounded, which is par for the course with Theo

Posted Jun 13, 2007 18:39 UTC (Wed) by davem (subscriber, #4154) [Link]

No, it's totally unfounded.

If you actually _LOOK_ the full hypervisor mode of the chip
is documented and Sun published the entire hypervisor source
tree as well as a full simulator as well as RTL for all of
the chips involved. What more do you want?

Theo is just jerking everyone's chain because he didn't get
his way in one situation with Sun a long ago in the past and
it's easier to make fan fare than to try instead to
build a healthy relationship with Sun. Which, in all honesty,
would be better for OpenBSD than what he does currently.

Theo is about manipulating people and making things seem the
way he wants them to seem.

totally unfounded, which is par for the course with Theo

Posted Jun 13, 2007 18:48 UTC (Wed) by davem (subscriber, #4154) [Link]

And having read his comments more carefully, he's even more full of
it.

Linux does support the v215 unlike his comments state, the PCI-E
controller that's he's huffing and puffing over the specs about
is a 20 or so line driver in the Linux kernel.

arch/sparc64/kernel/pci_fire.c

FreeBSD has a driver too, I think they beat me by like a day or
two :-)

Theo prefers to complain and be difficult, rather than deal with
truth and be cooperative.

well, somebody should inform Theo

Posted Jun 13, 2007 19:06 UTC (Wed) by qu1j0t3 (guest, #25786) [Link]

Somebody should let Theo know then, that all the information he wants is readily available. An email should do it.

Since Jonathan's apparently taken to "open letters" to specific individuals, maybe his next blog post will address misunderstandings about Sun's total openness of documentation.

(I also want to see the one that begins, "Dear Steve, I am sorry for spilling the beans about ZFS in Leopard. I had no idea you were so secretive about product plans. I hope this hasn't soured a beautiful friendship, love, Jon.")

well, somebody should inform Theo

Posted Jun 13, 2007 19:31 UTC (Wed) by davem (subscriber, #4154) [Link]

The hardware docs for the PCI-E chip aren't available, but the
programming is pretty damn clear from the Niagara hypervisor
sources, the T1000 and T2000 Niagara systems have the same
exact PCI-E controller as the UltraSPARC-IIIi systems Theo
is moaning about (v215, Ultra45, etc.)

And he could easily look at the Linux or FreeBSD drivers, both
work and are tested. Or is 20 lines of code too much for Theo
to understand? :-)

I don't have the docs either, yet I was able to write a driver.
I didn't ask Sun for the docs, because I didn't need them.

Instead of complaining, I used the information available, did
a little testing to figure out the missing gaps of information,
and wrote a driver. If I really needed to I could have looked
at the Niagara chip RTL, the PCI controller hardware is described
in there too.

The main point is that Theo could be quiet and add support to
OpenBSD if he wanted to, but it's his choice to be abrasive.

And it is this abrasiveness that backfires on him and his projects.

Companies are reluctant to cooperate with him, because all he
will do is turn around and crap all over that company if they
don't give him exactly what he wants.

well, somebody should inform Theo

Posted Jun 14, 2007 0:18 UTC (Thu) by xoddam (subscriber, #2322) [Link]

> The main point is that Theo could be quiet and add support to
> OpenBSD if he wanted to, but it's his choice to be abrasive.

Theo isn't that badly informed. He says OpenBSD *has* support, and he
says it was reverse-engineered.

For all I know Theo's "reverse engineering" is the same process as your "a
little testing to figure out the missing gaps of information" :-)

The only factually incorrect claim he made in his comment to Schwarz' blog
is in claiming that OpenBSD and Solaris are the only OSes with support for
the machines in question -- and if Linux and FreeBSD added support only in
the last month, he might be forgiven for his ignorance (if not for his
posturing).

posturing?

Posted Jun 14, 2007 12:24 UTC (Thu) by qu1j0t3 (guest, #25786) [Link]

What's wrong with the posture that hardware vendors should publish interoperability information if they want complete and timely operating system support?

Furthermore, nobody else is applying such pressure, and chipmakers appear too dim (monopoly effect, I guess) to induce this simple fact alone. We should praise Theo, not bury him.

posturing?

Posted Jun 15, 2007 0:06 UTC (Fri) by xoddam (subscriber, #2322) [Link]

> We should praise Theo, not bury him.

Theo has a very good argument to make in favour of hardware vendors
opening their specifications. I'm more than happy to praise him for his
general stance on this issue.

*However*, in some instances (here with Sun, and a few months ago with the
OLPC hardware) Theo doth protest too much.

Complete, detailed and accurate hardware documentation may be a
*sufficient* condition for someone to write a driver for OpenBSD, but it
is clearly not a *necessary* condition, and providing it is simply not a
priority for most hardware vendors. In cases where openness *is*
something the vendor cares about, there's nothing wrong with them choosing
to satisfy that goal by some other means.

Content-Type: text/plain,format=flowed

Posted Jun 15, 2007 0:10 UTC (Fri) by xoddam (subscriber, #2322) [Link]

Ok, format=flowed does have its uses sometimes.

Jonathan Schwartz replies to Linus regarding ZFS and GPLv3

Posted Jun 14, 2007 14:07 UTC (Thu) by davecb (subscriber, #1574) [Link]

Er, Theo is famously contentious. A grain of salt is recommended (;-))

--dave