Submitted by Simon Phipps of Sun has responded to the recent criticism of Sun’s openness, pointing out that even releasing information that they may already have costs a lot of money. “Jonathan asked me to look into this, to ensure we’re pursuing an open path across all of Sun, not simply the software group. We take all input seriously, and we can’t solve all problems for all parties, but we’re committed to doing our best to faithfully engage with all the communities we serve, in the same spirit as the existing Open Source Ombudsman Scheme. With the support of my team and others in the community I’ll try to build a new scheme that is fair and transparent.”
Seems like a fair answer to me.
Now, move on.
“Seems like a fair answer to me.”
Pretty much.
There is nothing they can do if no documentation exists, or they’re legally disallowed to distribute it.
But nowadays open source projects are better sources for hardware documentation, anyways. I remember programming an ethernet driver once, and I had to replicate all the functionality of the Linux code even if they did not make much sense (apparently there are many undocumented bugs, which they’ve already fixed).
Bugs? I thought they were, “features”? 😉
“Seems like a fair answer to me.
Now, move on.”
I agree 1000%.
Recent criticisms from Theo and Linus of Sun were completely ridiculous. But, both of those individuals are well known to be extremely arrogant, with their heads so far up their rear-ends they need plexi-glass belly buttons to see where they’re going.
Really, Sun has become one of the most open source friendly of all major corporations, and has made huge contributions and donations to the open source community.
And it’s Sun’s right, as well as any company’s right, to release what they want to open source, or to publish or not publish documentation. And they have legal obligations with their business partners.
So if Theo wants to have his temper tantrum about getting all docs on all Sun chips, well, too bad. Sun will release what they can, within legal and cost limits.
And Linus can have his tizzy-fit over the fact Sun wants Linux drivers for OpenSolaris (who wouldn’t), and he wants ZFS fully open sourced. But, again, Sun will release ZFS for open source when it makes sense for them.
But it would make no difference; the whole driver would need to be virtually re-written again from scratch due to incompatibilities. I mean, if anything, the best source for drivers is *BSD – and that is where they’ve been getting them from. Case in point, the 3945 wpi driver for solaris is based on OpenBSD/NetBSD/FreeBSD code.
What OpenSolaris needs is more mindshare, not more ‘code’ from other operating systems. The issues that related to this dilemma have nothing to do with licences but the relationship that Sun employee’s have with the opensource community. Remember folks, there are around 32 thousand Sun employees – the head may say one thing but it would be naive to assume every one of the 32 thousand are lock step with ‘the leader’.
ZFS was released open source into the OpenSolaris community before it was in Sun’s Solaris. http://www.opensolaris.org/os/community/zfs/
What Linus said was that it isn’t available in a way in which it could be ported to Linux.
I don’t know you could truly call it open if it’s tied down by licensing. To me, it’ll only be truly open when it’s allowed to be freely ported and modified.
“I don’t know you could truly call it open if it’s tied down by licensing. To me, it’ll only be truly open when it’s allowed to be freely ported and modified.”
http://mail.opensolaris.org/pipermail/zfs-discuss/2007-April/026922…
I would consider that “freely ported and modified”.
Yes, but as I understand the situation it cannot be included in the Linux kernel due to licensing.
Code from the Linux kernel cannot be freely ported to the FreeBSD kernel. Does this make Linux not open to you?
I’m talking about including ZFS in the kernel.
Yes I realize that, I’m just extrapolating. If ZFS is not free enough just because it can’t be used in Linux, then surely much of Linux should be not free enough because it can’t be used in FreeBSD.
By that logic, the Linux kernel cannot be “freely ported and modified” since the FreeBSD project cannot use it due to the GPL license within their kernel.
Its all a matter of perspective.
“And it’s Sun’s right, as well as any company’s right, to release what they want to open source, or to publish or not publish documentation”
It’s also Theo’s (and Linus etc) right as a customer of said company to ask them for documentation of the hardware they have so that they can make use of it.
“So if Theo wants to have his temper tantrum”
Having an opinion and calmly expressing it does not equal throwing a temper tantrum. Seriously, if you thought this was Theo having a tantrum you clearly haven’t been around long.
I think it’s interesting how some people just can’t admit it when Theo’s right , no matter what, and instead opt to root for the big companies who really don’t give a fsck about OSS.
You think this is all about Theo and his ego but it’s really about the usability of hardware for the users.
Open documentation results in better hardware support which in turn makes happier users. In the long run this is better for all OSS operating systems.
“I think it’s interesting how some people just can’t admit it when Theo’s right , no matter what, and instead opt to root for the big companies who really don’t give a fsck about OSS.”
But Theo is completely wrong. He’s assuming that all the documentation is already there, and that Sun has complete legal clearance to release it to the public. As Simon Phipps clearly explained (and you might want to actually read his blog linked in the subject of this thread), there are other considerations. But since you evidently have not read what he had to say, here it is in summation:
1. The product in question may be past it’s end-of-life date and there’s actually no team anywhere in Sun that can answer the question.
2. The documentation may not exist. Writing it would require costly reverse engineering (even if people with the skills to do so are still at Sun).
3. It may exist for internal use, but releasing it outside Sun would need legal review to check for ‘trade secrets’ belonging to others. That legal review is time consuming and costs real money. And the answer may turn out to be “can’t release, can’t name the company that’s obstructing us”.
4. It may exist and it may be possible to release it. Doing that, though, is actually a non-zero cost since usually the documentation is in an old format of some kind.
5. It may actually exist in an online-ready form.
So, clearly, simply releasing requested documentation is more complicated than you or Theo obviously assume. Now, if Theo said he would pay for Sun’s costs (whether that’s legal, or reverse engineering, if need be), then he’d have more of a point.
Otherwise, I strongly believe Sun’s intentions are reasonable. And, as I’ve said, Sun has increasingly become a major contributor and friend of open source, especially under Jonathan Schwartz’s leadership.
So it seems ridiculous to me for open source leaders or advocates to turn around and slap Sun in the face for giving absolutely everything they want, right away. They just comes off as being obstinate little children. Meanwhile, in their responses in their blogs, both Jonathan Schwartz and Simon Phipps have come off as mature, professional, pleasant adults doing what they can to appease the ever fickle open source community.
And I’m not rooting for “the big corporations”. On the contrary, I’m always extremely distrustful of all major corporations, and always favor the the little guy. That’s why I never shop at WalMart and favor small Mom & Pop shops instead. That’s one of the reasons I have a strong dislike for Microsoft. And I hate PG&E. And I think Larry Ellison is a snake. Need I go on? Clearly I’m not a corporate fanboy.
And Sun does “give a fsck about OSS”. Under Schwartz, it has staked it’s present and future on open source, systematically releasing most of it’s software portfolio to open source, not the least of which was GPLing Java. What the hell else do you want from them?
I just believe in giving credit where credit is due. And Sun is earning major kudos from me. Just because they don’t have all documentation immediately available for all of their chips doesn’t mean they’re not trying. Accusing them of “duplicity” is a low blow and uncalled for.
Just give them a friggin’ break, for goodness sake. And grow up while you’re at it.
It’s also Theo’s (and Linus etc) right as a customer of said company to ask them for documentation of the hardware they have so that they can make use of it.
I seriously doubt they are Sun customers. Most customers would make an informed purchase and know what OS the hardware supports. Theo and Linus are makers of OS that compete with the OS that comes on said hardware. I seriously doubt Theo or Linus “use” the hardware in question.
“I think it’s interesting how some people just can’t admit it when Theo’s right , no matter what, and instead opt to root for the big companies who really don’t give a fsck about OSS. ”
This is nonsense. Instead of getting the facts straight, some people just support whatever rubbish their OSS leaders spout.
Theo is wrong about OpenSolaris not supporting the PCI-E bridge.
Here is the email regarding OpenBSD and the V215/V245:
http://undeadly.org/cgi?action=article&sid=20070407155547
“I just committed the final bits that make the Sun Fire V215 and V245
fully supported in 4.1-current. These machines have a PCIe host
bridge, supported by the new pyro(4) driver. Writing this driver was
the biggest challange, since Sun doesn’t provide any documentation for
it. So it took some extensive digging through the OpenSolaris code to
figure out how the hardware was supposed to work.”
So Sun has open sourced the driver that makes the PCI-E host bridge work and the code is in OpenSolaris.
Infact here are the files.
http://cvs.opensolaris.org/source/search?n=25&start=50&path=px
OpenSolaris doesn’t contain source for some of the other drivers for the system most probably because the manufacturers of those devices don’t want Sun to open the code. As others have pointed out those are available as binary drivers.
http://www.sun.com/servers/entry/v445/arch-wp.pdf
According to this whitepaper all the other I/O devices are non Sun made, industry standard devices. So all of Sun’s IP in that system has been opensourced and is in OpenSolaris. Fire is the host bridge in question and the OpenBSD Pyro driver now works with it.
Theo’s entire rant seems to be a huge publicity stunt. Simon Phipps response seems more truthful than Theo’s meaning less rant in the end.
Edited 2007-06-21 02:21
Sun is becoming a model open source citizen! I like the direction the company has taken since Schwartz has taken over. Good stuff.
One thing Sun needs to do is come clean about SCO and their involvement with them.
Pardon, what relationship with SCO? all I remember was the paying for some IP (probably relating to opensourcing Solaris) – I would hardly call giving SCO money in the single digits figures providing money for their war chest.
Contra to the conspiracy theories that exist; there is no covert relationship between SCO-Sun-Microsoft; The only relationship Microsoft has with Sun is the $2billion settlement which included $2billion cash plus access to Microsoft IP (as well as collaboration).
Edited 2007-06-19 14:36
http://www.groklaw.net/article.php?story=20070602183551120
I am a SUN fan. If you go back through my posts you will read I wanted SUN to be the OSS darling, and not IBM.
But SUN’s invlovement with SCO needs to explained. If it looks, walks, and quacks like a duck, more than likely, it’s a duck. I’m too cynical to believe this is all coincidence. McNealy is gone now, so it’s shouldn’t be anything for Mr. Schwartz to just come out and be open about what happened. Otherwise it’s a stumbling block. Ask yourself how can you fully trust someone who may have had involvement in a plot to subverse you in the past but won’t talk about it when asked why? How can you? The obvious answer is you can’t.
All they need to do is come clean about it. They don’t even need to apologize, althought that would be nice, but they do need to come clean about it.
In the end, however, this is something that bolsters Theo’s argument that SUN needs to do more to show they are really open. Because as it’s stands, I don’t trust SUN and I can’t advise anyone interested in OSS to trust them either.
Oh please, did you actually *read* the link – do you know what it is actually *ALL* about? what SCO says and what the relity is, as you and I well and truely bloody know, can be as far apart as New Zealand from the United States.
Microsoft licenced code from SCO to improve their services for UNIX offering; and Sun needed to purchase components that were still owned by SCO. Please, if Sun and Microsoft were *really* going to bank roll a litigation suite, wouldn’t you think that firstly they would do it in an alot more concealed manner, and secondly, wouldn’t you also conceed that they would need alot more money?
Sco will say any damn thing; of course they would say that “this is an IP purchase for UnixWare AND Linux” to make their case seem more legitimate – make it appear that two of the IT world largest companies are purchasing IP related to UnixWare AND Linux – when in reality, it is just UnixWare. Sun and Microsoft see the IP purchased as purely from UnixWare, SCO on the other hand seen that purchasing of UnixWare IP means they’re also purchasing Linux IP.
Again, do a little reading analysing before screaming to the top of your lungs whilst claiming that there is some sort of grand conspiracy against Linux – if Linux isn’t successful, it will have nothing to do with law suites or so-called ‘voodoo’ happening behind the scenes. It’ll be something of its own making.
Yes, I did read the link, and have since the beginning of the SCO trial. I’ve read every stinking last piece of it. Briefs, motions, filings, exhibits, financial statements, and so on.
If you want to wave SUN’s flag and hold them harmless, so be it. I still cannot, and will not, say SUN is trustworthy until they clear the air about SCO. If you don’t like that, then I can’t help you.
And I never screamed at the top of my lungs. And you’re the one saying grand conspiracy. All I see is business manuvering, which is why I said they didn’t even need to apologize.
Good for you, but it seems that all the reading doesn’t seem to actually have informed you.
Hey, I’m not the one with the chip on the shoulder; it was you screaming and yelling over how ‘evil’ and ‘underhanded’ Sun was.
I mean, geeze, Sun does some stupid things, and if you’ve ever tuned into me on my blog or other forums, I’m going to be the last one to appologise for Sun – but if you do want to attack Sun for flaws, the SCO conspiracy is nothing more than that, conspiracy theories.
Excuse me, but you’re the one jumping up and down over non-existant conspiracy’s. Sun has already ‘cleared the air’ – they bought some IP – end of story, full stop.
You have a bad habit of putting your words into other people’s mouths.
Sun has not cleared the air. You may be pacified, I am not.
Whoever ends up owning SCO’s dead corpse after the litigation, will have an interest in recouping their legal costs. A way to do that may be to go after anyone who was behind SCO’s deal.
It’s a lot more likely that was Microsoft (and they’d present a very attractive litigation target) than Sun.
In particular, even if Sun actually was complicit in that game, it would make more sense to split the defense, and let Sun sell out Microsoft to the court, *once the litigation starts*, since if there was indeed a conspiracy against Linux that Sun was part of, then they’d have some great things to use against Microsoft in court.
So, I’d suggest patience as the best strategy: if Sun wasn’t part of the SCO game plan, giving them the benefit of doubt is a good thing, and if they were part of the game, it would be strategically stupid to make it attractive for Sun’s new leadership to have to defend themselves along with Microsoft, so giving them the benefit of doubt does not hurt.
Either way, we’ll find out in a year or two.
I doubt either were part of the game; Microsoft for example, most of their UNIX compatibility stack was taken off from a variety of *BSD sources, now they needed SRV5 compatibility – although I don’t believe the whole SCO conspiracy, I wonder why they didn’t approach Sun instead of SCO, given how comprehensive Sun’s own UNIX licence is.
As for Sun, I think the err of concern comes from the fact that it was very generic in their reply – ‘we purchase IP’ – IP related to what what product, what specific IP was purchased? I’m sure if they said “some drivers, some of the system level libraries etc’, there would be alot less concern. But given how companies in the past use the mearky term IP to hide their underhanded business dealings.
As for SCO, I have a feeling it isn’t SCO that is pushing for litigation but Connical; SCO actually had a working plan when it was selling desktop Linux; Caldera desktop linux (which I bought a copy) was very good for its time, and the attention to detail was awesome.
I have a feeling that Connical started to demand instant results, they wanted more money, and they wanted it now – Ransom Love (Who now has left) probably didn’t want to have any part of it, so Connical found some patsey would be willing to roll over and play dead.
As for purchasing IP, just because a company is dying, doesn’t mean that their IP is of any less value.
Oh, I can certainly picture Microsoft and Sun buying licenses from SCO without the necessity for a conspiracy theory to explain that behavior. Sun could be precluded by the non-disclosure section of their contract with SCO from giving further details on their purchase from SCO, for example.
I’m just saying that even in the worst case scenario, giving Sun the benefit of doubt is not a bad strategy.
The response addresses the lack of documentation for older Sun hardware.
Unfortunately, Theo’s posting was not about old Sun hardware. Theo specifically mentioned two current Sun products (V215/V245).
The new scheme I envisage is not just about old hardware. The same constraints continue to affect newer systems too – see cases 2-5 that I list in my blog posting.
No, its the same situation (most likely). Good ol’ citizen Theo didn’t explain to you *why* OpenSolaris doesn’t have the sources necessary for the v215/v245. The problem is the mpt(7D) SCSI controller is still closed and part of the “ON-closed-bins”. This is a tarball that contains binaries of various drivers and executables that haven’t been opened. Now because mpt(7D) hasn’t been opened yet, does not mean OpenSolaris won’t *run* on a v215 or v245, even if you don’t want to use the currently closed mpt(7D) driver. You *could* netboot it (Solaris is great for such things) of throw a different controller in there, say a fibre HBA and hook it up to some external storage.
Now why Sun hasn’t opened mpt(7D) is a bit of a mystery. mpt(7D) is used for various LSI SAS disk controllers. Should LSI not want the code released, Sun would be bound be legally bound by that decision until a suitable “open” replacement for mpt(7D) comes around.
No, Theo didn’t tell you this, because in this case generalization was better for him in this situation. I hope you see what he’s done.
Edited 2007-06-19 12:45
> I hope you see what he’s done.
Yep, Theo did, what he wanted to: To show us that Sun is much less Open than it wants to be. Thanks for confirming.
So you’re saying that Sun should risk litigation to open up the mpt(7D) code that might possibly limit their ability or interest in opening up farther aspects of their software and hardware?
I hope not all of Theo’s followers are as short-sighted as you are.
Edited 2007-06-19 13:54
I think one of Theo’s main points is that his correspondence with Sun (and other hardware vendors) often results in a lot of hand-waiving and promises that never amount to anything.
Companies love to play PR spin games where they say a lot of big things but then nothing comes of it — just like politicians before election day.
I’ve experienced this firsthand — I recall submitting a patch to VIA for a broken FreeBSD NIC driver they were distributing on one of their motherboard CDs — it was a four line patch, but without it the driver wouldn’t compile. They danced around telling me that it was all taken care of, made me talk to a bunch of different people, and then finally I was told that they couldn’t update the sources because some other company owned them. And you know what? They just kept on distributing the broken driver. The grand irony of this is that I later found out from one of the FreeBSD dev’s that the VIA driver was a ripoff of the original FreeBSD driver! (modified slightly to work with this specific onboard NIC implementation).
The big problem with some of these companies is that they have little or no consideration for your time. You talk to about five different people who all tell you it’s going to happen, and in the end you get shrugged off by an outright lie.
Sometimes they might actually have legitimate reasons for not taking action, but more often than not they feed you a line of bull and hope that you’ll just forget about the whole thing. They don’t want to give you the truth and say “we can’t give you x_thing because of y_reason” as that makes them look bad. It’s easier to tell you that they’re going to help you and then hope you’ll give up and go away on your own.
Except that it’s not mere “handwaving” this time on Sun’s part. There are real issues at stake but Theo just doesn’t seem to have a problem with making baseless accusations against Sun.
According to his own words webmink thinks I/O chip manufacturers have a “moral duty” to provide complete documentation.
“Unlike those companies which sell I/O chips for a living, and thus have a moral duty to provide complete, externally presentable documentation for their chips,”
If it is a case of LSI blocking release of the documentation, then I would encourage Sun in the future to consider non availability of “complete, externally presentable documentation” as a dis qualifier when choosing either 3rd party components or licensing technology for Sun produced components.
Sorry, but I’m not sure I understand how this situation is different from any other peecee vendor that chooses to use an LSI SAS disk controller. I’m sure IBM, Dell, and HP, and so fourth all sell systems that at the very least have LSI products. You aren’t bugging them about it, and most importantly (and ironically), neither is Theo, why?
You should be going to the root of the problem, and that’s simply not Sun.
One of the things I noticed about his postings is his emotion behind the posts, which is what you’d typically see more on Usenet as new posters get flamed and respond accordingly.
As for what he’s asking for, I myself know it’s not as easy to disclose trade secret information without checking to see if it would be in violation of contract. It’s something that Theo doesn’t really appear to understand very much from what I had read.
I don’t fault Sun at this time for anything – they’re doing quite well and I see them bouncing back harder from the low point they were at now that they’re opening up and providing moral, ethical and technical support to the community.
From here on out, patience is the ultimate virtue and all we can do, unfortunately, is wait with our mouths dropped, drooling over what’s to come next.
De Raat is demanding that SUN release the docs for a specific part that they don’t manufacture. Glad to see he hasn’t changed since his NetBSD days, his temper tantrums are always entertaining… how some people still take him seriously is beyond me.
I wish Mr. Phipps the very best but from my own experience dealing with him I cannot tell the public he succeded in improving much over the last 2 years. Maybe he will do better in the future. Yes, his job is difficult.
Going back to Theo and his comments on Sun: He is right on the money as far as I can tell…
We will wait and see what happens over time.
Sun needs to have a fundamental directive to force its openness on its vendors. Don’t want to share your documentation with our customers? Then we don’t want your chips.
This spirit openness needs to be a pervasive driving force, transforming the Sun work culture at all levels. Sound radical? Of course. Sun has been hurt, badly, by Linux and missed the boat 10 years ago when things were really getting underway. Now that Open Source is a mainstream concept, Sun really has to work 5x as hard to make up for lost time. Or die. It’s that simple.
The other thing they need to do is get their lawyers in line. The lawyers at these big companies are like back seat drivers. Sometimes they just need to be told “find a way to make this happen”.
Edited 2007-06-19 13:41
aparently you havent read the article.
Simon clearly, factually and briefly articulates the complex barriers to easy “quick-fix” solutions for the documentation issue. There is no silver bullet, is there?
Isn’t the nasty issue here about the tendency to skip the hard work of civil discourse on the real issues in favor of negative spinning and characterization of Sun’s intent? What a boring waste of time and energy. We should probably leave those thin “discussions” to mainstream commercial media.
Thanks to Simon for trying to keep this discussion real.
I come from the openness debate from a pragmatic perspective. I like open source systems for the visibility it gives me. What I don’t like are religious warriors claiming X is not open enough, for that Y shouldn’t be used because of the proprietary drivers. What I really, really hate are “open source” companies that keep all the interesting keys to the kingdome to themselves. I don’t think (deep-down) Theo is the former and I don’t think Sun is the latter. I also recognize that a system like Solaris, or the SunFire platform is not the intellectual property of one company and that it takes time to 1) convince other vendors to open up and 2) work around any pieces that other vendors refuse to open.
I think Sun has done an admirable job, and I’m looking at deploying our next applications on Solaris 10 for x64. (Something I would not have considered under closed source Solaris). I even think it’s okay for Sun to bundle closed source pieces as I would rather have good quality drivers for the hardware than often slower performing (lowest common denominator) pure open-source drivers. (For example, some closed source ATI drivers versus open source ATI drivers). However, this is a short term solution to the current hardware available and Sun should understand STOP BUYING HARDWARE FROM VENDORS THAT REFUSE TO OPEN THEIR DRIVERS. If all major vendors (IBM, HP, SUN – Dell is happy to live in shit and sell Windows) started making openness as a requisite, then Broadcom, ATI, NVidia, etc. would deliver and the driver/hardware issue would be a mute point.
Edited 2007-06-19 16:36
No, Sun should buy the best hardware that meets the requirements of their customers at a price and performance level that meets their needs.
Sometimes this means choosing hardware that doesn’t have open specifications but is significantly cheaper or better than available alternatives.
Forcing the choice of open hardware is only practical when that qualifier is more important to customers than any other.
If I have to choose between:
* $500 for a slow video card but has open source drivers
* $100 for an ultra fast video card but no open source drivers
I’m going to choose the second option every time because “perceived freedom” is not worth the difference in price.
Agreed. But SUN should include openess regarding documentation, as part of their needs when they evaluate their hardware partners. Why? Because that’s what is expected from an OSS friendly company.
P.S. I think SUN is a great company and deserves their ‘OSS supporter’ badge. It’s just that things take time. OSS people have every right to make complaints to get problems noticed. This will only help for the better.
Agreed. But SUN should include openess regarding documentation, as part of their needs when they evaluate their hardware partners. Why? Because that’s what is expected from an OSS friendly company
I’m certain they do at this point, and Simon, by posting this, has assured us that it will be a consideration going forward.
Edited 2007-06-19 17:07
Someone like Sun has the market power to go to a Broadcom and say “sorry, can’t use your RAID chips – you won’t open the drivers. However, Intel will.” In my personal experience the contrast has rarely been so stark as in your example. In many cases the performance is comparable but maybe the Broadcom chipset is a little cheaper. In some cases there is no real performance or cost difference, but rather what the engineers have been putting in there. It also encourages the selected vendor to continue opening their drivers. Nothing sucks worse than testing the waters with a competative, open-source product to find no market acceptance. I agree they should pick the “best” hardware but best is often viewed along many axis, which may include the degree to which the driver is open.
You seem to be merely repeating what I already stated.
The point is, even a few dollars difference can make a big difference. Remember that Sun has a responsibility to their shareholders *before* the community to bring the most return on their investment. Obviously there is some balance to be hand in that, but “freedom” and “altruistic goals” only go so far.
Finally, Sun actually doesn’t have the market position to be able to tell Broadcom “sorry can’t use your RAID chips – you won’t open your drivers.”
As much as I would like to believe Sun has the kind of market power, they don’t. Maybe Dell, but definitely not Sun.
I also don’t believe that any of Sun’s engineers would arbitrarily choose hardware are you seem to imply by saying there could be no cost or performance difference. Everything I have read or heard indicates otherwise.
Actually, you gave a product that performed poorly at 5 times the cost. That’s a strawman. What happens when the product is just as good at a 3% premium? What happens when the part is even cheaper but at 95% perofmance? I don’t want them to open source for a religious reason – read my original comment. I want them to open it up because it’s more appealing to customers. In particular it’s more appealing to this customer. I like ZFS and I like MDB/DTrace – great selling points – but not if the long run option is a 75% open kernel. In the short run, that’s fine, as they make their transition, but I think it will actually be in their shareholders best interest, in the long run, to do what customers want.
Actually, you gave a product that performed poorly at 5 times the cost. That’s a strawman.
I disagree. It doesn’t really matter if the difference is $3 or $500. Depending on the person the difference may be irrelevant as long as a less expensive option exists.
In addition, you seem to gloss over the fact that 3% on a single system may not big a big deal, but if someone was purchaing thousands of systems that 3% can add up very quickly.
I’m certain that if you asked the average IT department who is purchaing 1,000 machines if they wanted them with a video card with an open source driver but it would cost them 3% more for their total invoice than it would if they opted for a driver with no source code they would choose the cheaper option.
That’s my point about “perceived freedom”; it’s all a matter of perception.
In the short run, that’s fine, as they make their transition, but I think it will actually be in their shareholders best interest, in the long run, to do what customers want.
That’s the for the shareholders and Sun to decide, not for people who think that small increases in price are acceptable for “perceived freedom.” Remember that most of Sun’s customers could care less if the source code is available for a particular driver for their hardware or not, they just want a completely supported, working solution.
While Sun’s customers did want the source code for Solaris for other reasons, in this particular case, I do not believe the same necessarily applies.
Its also rather impolite, in my view, to force your particular view of “necessary freedoms” on a company who you do not represent as a shareholder or Customer (probably).
Personally, I could care less if the source code for a driver is available or not. Wherever possible, it is certainly nice to have that insurance, but it is not worth money to me.
In an ideal world, a Customer would be able to simply say “I would like to ensure my systems are fully open and am willing to pay a 3% premium.”
Its the same situation Dell is having right now with some of their systems.
Nevermind that in some industries, the full specifications of hardware will never be available due to government or other regulations.
Edited 2007-06-19 21:02
Actually, I do have a very modest 500 shares of Sun and am purchasing hardware for my business through their startup essentials program. I really like the turn Sun has made to open source. I am already paying a premium for some equipment through Sun, and the startup essentials discount brings it more in line with the rest of the market. (Their SATA drives are not cheap – almost $1 per Gig.)
One requirement I had for the platform we settled on (Solaris) is open source. I looked at FreeBSD and I already had a fair grounding in Linux. If Solaris were not open source, it would not be on the table. Part of the reason is we expect to host some of our contracting customer sites in house. ZFS and Zones are a large part of what drew us to Solaris, but it would have been a non-starter under closed source.
Good for you, but as I said, it is a matter of perspective.
There are more things to consider than perceived freedoms when choosing the right solution.
“Someone like Sun has the market power to go to a Broadcom and say “sorry, can’t use your RAID chips – you won’t open the drivers.”
Actually no. Many PC vendors, like Apple, Dell and HP, have far more volume than Sun does to dictate terms to vendors.
It depends on the market. For SATA adapters – that’s true. They sell many more to DELL and HP than they do to Sun (or Apple that only pushes about 1.5 million boxes a quarter). When you start looking at more specialized chips – like Fibre Channel chipsets for their Sparc line (endian difference) – the market often gets more rarified. And in some cases, the chip sets are customizations of their existing chips, so they are semi-custom or at least not-standard.
Let me rephrase myself. When Sun is choosing a vendor for chips, it should given preference to vendors that will open-source drivers. Thereby encouraging vendors to open-source their drivers.
When you start looking at more specialized chips – like Fibre Channel chipsets for their Sparc line (endian difference) – the market often gets more rarified.”
Most Sun servers use PCI or PCI Express for I/O. PCI is little endian. Devices don’t get specially made for endian differences on Sun boxes. The Sun I/O controller that bridges the CPU bus to the PCI bus does the conversion. So there are no special chips that 3rd parties make for Sun.
I don’t think most of the low-end (High volume) Sun systems use fibre channel controllers on board. The current generation uses SAS/SATA controllers. So the same market that caters to Dell, HP and Apple also caters to Sun’s I/O device needs.
” And in some cases, the chip sets are customizations of their existing chips, so they are semi-custom or at least not-standard.
Unless you have hard data to prove this, that is not the case.
“Let me rephrase myself. When Sun is choosing a vendor for chips, it should given preference to vendors that will open-source drivers. Thereby encouraging vendors to open-source their drivers.”
That can’t be the primary motivation in choosing a technology. Sun usually uses it’s own drivers for Solaris. There are very few binary only Solaris drivers for SPARC. The same reason for why Sun doesn’t have the Market clout to dictate terms also ensures that 3rd party vendors don’t write Solaris SPARC drivers.
Edited 2007-06-19 23:02
“I agree they should pick the “best” hardware but best is often viewed along many axis, which may include the degree to which the driver is open.”
Clearly you haven’t been involved in any large scale design project…
No, but I am a customer, and openness is important. An axis that is important to me, as a customer, is the openness of their drivers and kernel. If they want to pick the “best” chip that precludes them from opening their drivers and parts of their kernel, I might not be so interested in Sun as a vendor. After all, I have plenty of other closed-source Unix options. I also have several open-source Linux options.
Assuming there are other customers concerns like mine, and we have no prospect of getting a soup to nuts open stack sometime in the future (with the understanding that Sun is still in transition), is the “best” chip one that makes it harder for customers to buy your systems? I’m sure an engineer tucked away in a cubical, with no regard for how to market the product might pick chipset A – that is a binary only, closed source driver that prevents the release of portions of the kernel source. However, is that the “best” chipset for the product? If chipset B delivers 95% of the performance at 103% of the power consumption and is willing to release their drivers as open source, isn’t that a better chipset in some sense?
Just words, words, words and no SPECIFIC solution.
Yep which ones, which ones do NOT fall in there, etc? Can you give specifics?
This is beautiful but void of content:
Pedro.
I agree, Pedro. All words, and nothing solid to judge me by. I have to start somewhere though – Jonathan only asked me to do this on Sunday after all.
Yes, Theo’s rants are still entertaining, but honestly, they are distracting, too.
On the hardware front, remember that Sun sometimes *makes* support chips, and these may have IP (in the form of licensed cell libraries) which implement certain controllers, etc.
It is these that are likely the source of the issue about not being able to document the functionality.