Slashdot Mirror
Sun vs. OpenBSD?
An anonymous reader writes "CNet has an article up about OpenBSD trying to get documentation for Sun's UltraSparc-III processor. Basically Sun is giving them a bit of run around....There is some documentation available for the processor, but not enough to get things to boot."
much like this article from the other day.
man
No manual entry for
For more info see this previous /. article from the BSD section that didn't make it to the main page here:9 13211&mode=flat&tid=122
http://bsd.slashdot.org/article.pl?sid=02/11/26/1
But Sun didn't put its open-source community liaison officer, Danese Cooper, in touch with de Raadt until after CNET News.com informed the company of his dissatisfaction. Cooper is "already well-known in the open-source software community," Sun representatives said. But the company acknowledged it needs to improve its work with open-source groups, saying the task would be addressed "within the next few weeks. Cooper has been responsive and is pressing the OpenBSD case within Sun, said de Raadt, but he's reserving judgment until he gets what he needs. "
It seems they are now working very actively to solve the problem.
I agree completely with Sun on this one. They have to make a buck, and when a free OS comes along wanting to utilize its systems and take away from its revenue. I work with Sparc 10 systems everyday, and I have come to love the Solaris system, it is pretty flawless and is specifically designed for the Ultra Sparc architecture. Plus, many people the utilize the Sparc Systems with Solaris use specifically written programs for the Solaris systems and are designed in house, this OpenBSD system could be hell on those programmers and Sys Admin.
But Sun don't many any money from Solaris, it's just the stuff you need to run your apps on Sun hardware. Sun are a hardware company first and foremost.
Secondly, Sun don't even own the UltraSPARC, certainly not in the same way that Intel own the Xeon - see the SPARC International web site. SPARC is about as close as you can get to an Open Source processor.
From reading the article, it seems that Sun simply didn't have anyone looking after the BSD community like they had looking after the Linux community, and when the matter was brought to their attention, they assigned someone to do the job.
Of course, anyone who pays $$$ for modern Sun kit is an idiot if they want to run anything other than Solaris on it, because Solaris, as well as being a solid and powerful Unix implementation in its own right, is designed from the ground up for SPARC hardware, it doesn't have to make any compromises for compatibility's sake. That's why NetBSD is generally slower than FreeBSD, it deliberately avoids anything too platform specific, and performance suffers as a result.
Over the last few weeks this argument has been raging. I've lost count of the number of IRC conversations I've had on ircnet #openbsd, and here's my take on the options presented to the OBSD developers.
1:If the OpenBSD crowd want the docs, sign the NDA. Linux developers did this. It's not that big a deal.
2: Look at the Linux source for hints. This surely isn't too difficult.
Why are Sun not willing to make the relevant docs fully-disclosed to anyone who wants them, sans NDA? In part, the answer is simple: The USIII / III* proc is still pretty new. Solaris doesn't yet fully implement all the chipset features, but will do in future releases. Is it a good idea for Sun to open the proc docs to any Tom, Dick or Harry, including other chip manufacturers, at this stage? Probably not.
There's been a lot of negative talk propagated by the OBSD community regarding this issue. Classic "blame the faceless multinational" diatribe that most of us grew out of in our teenage years.
Access to the information the OBSD developers have requested is a privilege, not a right. They want to build a kernel around the USIII, which is great, but the rules have been set by Sun, and are quite clear. Deal.
(Yes, I do work for Sun btw)
Super Awesome Broadband
Nice troll. From the SPARC International FAQ:
And
Given that IBM dont care what OS you run as long as its there hardware
Not strictly true (or perhaps that should read not *historically* true).
Sure, these days an IBM boxen can run pretty much any OS you care to name. But not too long ago, IBM shipped all their machines with OS/2 WARP - their own attempt at a multi-tasking environment.
After OS/2 died like the lame dog it was (multitasking? It hardly even unitasked!), IBM were forced to give up on their abortive attempt to take over the software market and ship Microsoft Windows as the default OS.
Personally, if I want a BSD-based datacenter server, I'd run it on an Alpha system, not an UltraSparc. But that's just me.
-Blacklaw
OpenBSD has support for UltraSparc I's and II's (as well as the non-UltraSparc's). There was significant enough change in the way that the UltraSparc III's work that the OpenBSD team can't figure out how to get it to boot. The reason Mandrake works is because Sun already released the documentation to the linux guys under a Non Disclosure Agreement of some sort (The legality of of Open Source and a NDA makes me laugh). I don't know BeOS well enough to say much there, although they weren't mentioned in the article, so I doubt they can use the Ultra III's. I would wager the same holds true for DOS. Anyone know enough about BeOS?
You also might be interested in this OpenBSD Journal article about the same subject. Usually the discussions about the articles are very insightful on deadly.org
the problem is sun professes to be open. http://www.sparc.org/faq.html
..."
"Myth: Sparc is a closed proprietry architecture
if sun wants to be able to call itself open, its up to us to ensure they back up the marketing fluff.
Its not about source code, as if it were, OpenBSD couldn't use it anyhow, a la the Sun ECC code for OpenSSL, which carries some patent limitations that makes in incompatible with a strict constructionist BSD license interpretation (a la Open, and not Free or Net). Its about documentation, which they asked for repeatedly. The *reason* this is a story is because Sun runs around claiming openness. If this were about an IBM or HP chip, it wouldn't be a real story, as Sun is the leading drum beater about how open they are in comparison to MS.
ostiguy
Please read the article. Theo says they've tried, can't really deal with it. But, before you make any comment about the abilities of the OpenBSD programmers, you might want to check if you're using OpenSSH...
Score:-1, Funny
It exists only in a non-open form - via an NDA.
This is why it can not be used by OpenBSD.
"Myth: Sparc is a closed proprietry architecture ..."
Interesting that you should quote this bit, but not the rest. SPARC is open. Anyone can build a sparc chip, and modify and customise the specs to suit their needs. Sun do that, and one of the chips they've made to suit their needs is called the UltraSPARC-III. Nothing is stopping you producing a WienerSPARC-IV and running *BSD on it. Nothing about SPARC is "viral" in the way that the GPL is - ie products derived from the specs do not have to be open in the way the initial specs are. This is Sun's case. It's fair - think of SPARC as a sort of BSD licenced hardware spec. You can take the spec and produce "free hardware" or you can take the specs and produce "proprietary hardware" - true freedom in the way the BSD licence views it. How is this out of line with Theo's policy? Are they requesting any closed source projects based on BSD code return the specs so they can make OpenBSD work better with it? No. Then why should they stamp their feet about closed hardware implementations based on an open standard?
Try NetBSD... safe,straightforward,useful.
If Sun made it policy that it required an NDA to get Sparc 3 Theo would go away. That makes Sparc 3 a closed architecture.
:-) Sparc != UltraSPARC-III. SPARC is an open architecture. UltraSPARC-III is a proprietary, closed chip. What is difficult about that? TCP/IP is an open protocol, OpenBSD implements it in an open way, and MS does so in a closed way. One closed implementation of an IEEE spec does *not* make that spec closed and proprietary.
Aaaaargh! Stop it!
Try NetBSD... safe,straightforward,useful.
First of all, as much as I admire his work it must be said: Theo is an abrasive asshole. He is idealistic to the point of being utterly inflexible and insulting. He'd get farther if he behaved like a grown-up.
Secondly, to all the people who are accusing Sun of having no strategy, no plan, no policy, no hope, etc.. Just because YOU haven't bothered to find it out doesn't mean that it doesn't exist! There are a LOT of posts here that berate Sun for doing something that they clearly are not, or failing to explain something that they make perfectly plain. Go do your damned research!
"People who do stupid things with hazardous materials often die." -- Jim Davidson on alt.folklore.urban
With regards to taking the code from the Linux folks, it's not only a matter of respecting the GPL (if that was even a factor) but more a matter of correct coding. How can the OpenBSD folks be sure Linux code is implemented correctly, per the specs of the documents Sun gave them? They can't, without the documents.
This was discussed in some OpenBSD mailing lists. It's not only a matter of making the code 'work' on UltraSPARC III systems, but a matter of making it work _correctly_ and robustly. To do that, they need to read the specs.
If the architecture was too specific then you could not differentiate implementations. There are SPARC implementations for embedded markets, business markets, HPC markets, etc. that all make trade-offs.
Example: UltraSPARC III contains an integrated memory controller. The SPARC architecture cannot specify the type and programming model of the memory controller or the architecture would have no longevity. In order to know how to use the memory controller you would need UltraSPARC III-specific information.
It's been said before: UltraSPARC III is a proprietary implementation of an open standard.
The "open standard" part means Theo is able to design his own 64-bit SPARC, not that Sun has to provide the details of its own proprietary implementation. The rationale is preventing lock-in. You can buy a competing 64-bit from, say, Fujitsu if you don't like Sun's implementation or performance or pricing or support costs or features. Theo is twisting the marketing slogan for his own purposes by claiming it should be his own definition of "open".
DaveM smart. Theo whiner.