With the recent push from OpenBSD to open firmwares to redistribution as well as obtaining new documentation for several wireless chipsets it would seem OpenBSD is pushing for other areas to open up as well.Theo de Raadt posts that for the past four months he and the OpenBSD developers have been seeking documentation for the AAC RAID controllers from Adaptec and have received to date nothing but air. The purpose of this drive is to add the desired functionality of being able to properly recover without dropping into the BIOS.
Theo has given an ultimatum that should the people at Adaptec not see his side of things by Monday, support for the card will be dropped for the OpenBSD 3.7 release, set to be tagged on Tuesday.
Well I like the idea and appreciate the work Theo has done but the timeline sounds rediculous. Corporations take forever to make decisions and there is no way that even the most willing one could get it done by next Thursday. Hell, it’ll probably take at least that long to “bubble up” through the ranks.
Well, this discussion has been going on for four months, so the ultimatum at the end is, if not justified, understandable to me.
Even though Theo might have the best of intentions for the community, I’ve been a product manager and his harrasment technique via the overly-zealous OpenBSD community is quite unprofessional and certainly can harm OpenBSD in the long run.
Personally I applaud their efforts but you really cannot come up deadlines and declare to stop support just because you can. This childish and rekless behaviour by Theo doesn’t justify increasing the enimosity between the project and different hardware vendors.
I had just pulled an all nighter when I posted the previous… The title should say “Tuesday”.
That said even ultimatums take time to get to someone who can respond. Reminds me of that scene in Austin Powers 2 where Dr. Evil asks for 1 Trillion dollars… it’s just beyond insensibile. That said, good luck.
I understand why ones do not release drivers for webcams and joystics for linux and other OSes.
But I met some RAIDs from Adaptec, for example, which was not supported with linux and not even planned.
How would one be crazy if he releases RAID that do not work with linux? 30% sales drop instantly, because most people do not use windows on severs, those who need rarely use RAIDs on them.
Dropping BSDs is another 5% sales drop.
I just plain do not understand why do they ever do that and what do they hope for.
but you really cannot come up deadlines and declare to stop support just because you can.
erm… yes, you can. from the email to misc@….
Some of you may remember how this affected back a few years ago Adaptec refused to give us documentation for their SCSI controllers, and this slowed the quality of our ahc(4) and ahd(4) support. We let our users know who to complain to, Adaptec listened, and we had two boxes of *complete* documentation a few days later.
….sometimes deadlines and cutting off support is the only way to get people to listen.
Personally I applaud their efforts but you really cannot come up deadlines and declare to stop support just because you can.
Why not?
My box is due for an upgrade. So earlier this week, I took great pains to ensure that the components I purchased where fully supported on Windows, Linux and Mac OS X. I will never spend a dime again on any commercial software and/or hardware vendor that fails to grasp the concept and importance of transparency, openness and collaboration in this industry.
If for several understandable constraints a vendor cannot provide drivers or support for a particular operating system, then the vendor should at least provide open documentation and specifications for those interested in doing so. What is so complex, economically disastrous, or mind boggling about the concept?
No matter what you think about Theo I think the whole silly part should be aimed at the HW industry. Why not let OpenBSD make drivers for their equipment in a descent manner? What are they so scared of? Why is this an issue at all, just give them the damn specs and let them write drivers. Why bring on that cost to the company of making closed source drivers when the OpenBSD community offers them for free…. just stupid if you ask me.
I think OpenBSD is, in contrast of many Linux distributions, really making moves for the ENTIRE open source community by doing these efforts.
I also think it`s childish. How is it possible for a company taking OpenBSD seriously when Theo is gonna pull stunts like these?
When you are running a production server and suddenly support is dropped for some hardware what then?
Go Theo, go ! this will benefit all OSS.
remember when Theo did the same thing for the wireless chips, some linux guys talket about wrappers and so on.
“When you are running a production server and suddenly support is dropped for some hardware what then?”
Its hardly Theo’s fault that Adaptec refuses to provide the necessary documentation. Polite mails are the best way to remind Adaptec that they’re losing potential business here.
Large vendors fear openness because they believe right after some piece of information is released, a competitor steals all their ideas, produces the same product without R&D expenses, and effectively puts them out of business.
This is the sad truth about the level of knowledge of an ordinary executive.
Since executives are most concerned about making money – the only thing someone wanting openness and freedom can do is to buy open products and avoid closed products.
I personally plan to support the new open graphics card scheduled for release this summer
I do not agree with the notion that providing open documentations and specifications leak out trade secrets to competitors. Research and development costs are hardly bound to writing drivers or interfaces to hardware. They are bound to the fabrication and manufacturing processes for the hardware.
Even if Nvidia released full open documentations and specifications for their various hardware components, it does not automatically suggest ATI can copy or clone Nvidia’s technology. Why? Well, because of their disparate fabrication and manufacturing technologies. Hardware engineering is hardly like coding where you can copy and paste code and reuse libraries.
Also bear in mind that it is only until fairly recently that computer hardware vendors stopped distributing specifications with their products. I do not understand where the flawed logic that providing specifications and documentation for your products automatically means a vendor’s fabrication processes will be leaked, and hence the vendor is out of business. It’s a totally ridiculous (PR) excuse put for noncomplying vendors.
Doesn’t Intel, AMD, SUN, to mention a few, release specifications for their processors? Why are they still in business then? And why isn’t Tom, Dick and Harry cloning their processors?
How is it childish for Theo to say “okay we’re not going to include support for your hardware” but it’s not childish for Adaptech to say “we’re not going to let you have any docs on our hardware”
???
My own dealings with big businesses have proven that sometimes they can be downright wishy washy. You can waste a lot of time achieving nothing.
Maybe one of Adaptech’s competitors will be more willing to provide documentation. I’m at the point right now where if a hardware manufacturer wants to dick around being wishy washy I’m going to take a hard look at their competitors.
You say it is hard to take OpenBSD seriously for this, yet is it really any easier to take someone seriously if they have no backbone what so ever? Is it better that they simply wait half a year for something they don’t want instead of just saying that they want something else?
The Adaptec guy seems to be off his rocker and thinks that OpenBSD wants the source to Adaptec’s GUI software, yet all they are asking for is documentation to right a driver.
And this conversation has been going back and forth for four months, how could Adaptec ont understand this idea after four months? That must mean they are trying to screw with OpenBSD, which would get be angry in such a situation.
If support is pulled for something then it is because the support that was there was not good enough to be worth having. If you ended up with that case, perhaps you should talk to Adaptec or not upgrade your operating system or maybe move to another.
even if he may be a difficult person. He seems to be the only one out there that has principles.
If a device is not acceptable using because it offers no documentation to work it under ANY OS, then the *only* thing that’s right to do is to not use it.
Linux in contrast try to get everything to work on it, and they end up with lots of badly-written, half-documented drivers, and lots of binary-only crap. The solution is to avoid non-standard crap that doesn’t work. I will never buy a Lexmark printer or a D-Link device again for that reason. If everyone followed their principles, companies would be forced to comply with that.
Yes, I concurr wih this.
I do not believe that providing info required to write a driver equals to publishing trade secrets.
Let’s take an uber-cool graphics card for example. It maps textures to 3D objects blazingly fast. It can do so, because it has the newest chips on it, designed to do this. However, to write a driver (I have never writen one, this is how I see it should work), I only need info like “to upload a new texture, flip this bit, set that bit, send texture size info onto the bus, wait for 0 to appear on the bus, then send the texture data”. How the chip works with texture later is of no interest to the driver. Therefore I fail to see, what problems do the HW manufacturers have, wih publishing the specs.
I might be wrong. If so, someone please correct me. And provide an example, how info required to write a driver for a chip equals to trade secrets about the chip design.
Hi
I thinks Theo has style / standards and should be consistant . Go for it
My2c
perhaps i’m wrong, but…
the adaptec driver (as it is now) would be removed from the generic kernel. i think, you could still compile the driver as it is now into a custom kernel by activating the option in the kernel-config.
so only difference to the situation now, is the extra effort for compiling a new kernel. oh, and “official” support is only for a generic kernel so you won’t get help with a buggy “adaptec-kernel” anymore until this situation is resolved.
I might not entirely agree how he stated his email, but what he is trying to do is a good thing for all open oource operating systems.
Quoted from Theo’s post: We’ve been trying to get more Adaptec AAC RAID documentation out of
Adaptec for nearly 4 months, so that we can add RAID management (ie.
the ability to recover a RAID array without rebooting and doing it in
the BIOS) and it is incredible how much we are being dragged around.
The ability to recover disks without having to reboot is something that servers require. It does no good to people depending on those servers to be down beause a disk failed when normally, the disk should have been rebuilt while the machine was still running doing it’s job. Adaptec is making it impossible for this to happen. If you can’t count on your RAID hardware then hell, by all means take it out. He’s right, there are alternatives.
This happened when Daren Reed altered his licese to ipf. The OpenBSD team took it out and now we have pf and pf kicks ass! Unfortunately, the OpenBSD doesn’t produce hardware so all it can do is go with alternatives.
His tone was a bit unprofessional, but I guess I can understand if he’s frustrated. Besides, Theo has a record for not having the greatest people skills. What else would you expect from him?
Anyway, the point is that his goals should be supported by all open source groups because this will help out all the BSD camps and the Linux camps as well. It would even help out those hobby OS’s that are regularly mentioned on this site. This quest is a Good Thing(TM).
nxt:
a prominent example for the problem you don’t understand is the intel centrino chipset.
they save money by implementing a lot of stuff in software. so if they want you to write your own driver you’d need to know those software tricks they have to get the hw working.
i can see why they don’t want to give their competitors that info, but hey information wants to be free
those vendors sell “windows-hardware” and should label them accordingly. read: they don’t want your money.
Yes, you could add the aac as it was prior to removal to the kernel, but at that point OpenBSD washes it’s hands of you.
Anything not using GENERIC is not the concern of OpenBSD.
but at that point OpenBSD washes it’s hands of you.
uh.. ya, that was noted in the post -> oh, and “official” support is only for a generic kernel
Did anyone else read the headline as: OpenBSD 3.7 to Drop Adaptec Advanced Audio Codec (AAC)?
I don’t know if it’s better to post this here or onto the openbsd-misc list, but anyways….
First, Theo is full of crap. I’ll say that again: Theo is full of crap. I don’t think that he’s actually interested in making the AAC cards work. Instead, I think that he’s interested in stirring controversy, petty bullying, and silly ‘freedom’ tripe.
I worked at Adaptec for almost five years, until last year. I worked on the FreeBSD (and Linux) AAC driver, and I ported the AAC management CLI to FreeBSD. It’s available right now in the FreeBSD ports tree. I also added the proper shims to the driver so that the Linux AACCLI would work under emulation. The fact that I did these things is pretty well known in the BSD community; several other projects have contacted me over the years for help and information about AAC. But during the time the Theo claims that he’s cared about AAC, he NEVER ONCE CONTACTED ME! If he had come to me before I left and asked for help on making all of this AAC stuff work on OpenBSD, I would have been happy to help him. Heck, I might have even ported the AACCLI for him on my own.
Unfortuntely, Theo chose to ignore resources that would have helped him, and instead chose his normal super-confrontational antics. I have to commend Doug Richardson (one of the nicest men I’ve ever worked with, BTW) for his very appropriate response. If Adaptec provides an open SDK later this year, good for them, but it certainly is not due to Theo.
Theo could have had AACCLI support years ago, but chose not to. I hope he removes the driver from the tree. That would really teach everyone how mature and ‘right’ he is.
Scott Long
You may not have even noticed it but your little tool is restricted so it cannot be redistributed and is only available as a binary. This goes completely against what OpenBSD looking for, therefore your tool was not a valid option.
OpenBSD is looking to make their own that they can redistribute freely under the terms of the ISCL (the two term BSD). They don’t need the code, just the documentation to make it themselves.
And this tool of yours, it isn’t actually that widely known in the OpenBSD community, I didn’t know and I consider myself part of it. Perhaps it is known in the FreeBSD community, to which you were a part.
Re: d0h @ Nate:
I was emphasising that you get no help, not just “official” help, they will ignore you unless feeling extremely nice.
The AACCLI is not ‘my little tool’, it’s been distributed for years by Adaptec on the CD that comes with the cards. It’s also available for download from the website. I’ll accept that you might not be familiar with the FreeBSD ports system, but surely you’ve peeked once or twice at the FreeBSD AAC source and have seen my name there. According to the OpenBSD CVS logs, apparently someone has looked at the FreeBSD source in the past few years.
As for your goals of Libre, I hope they go well. I’d love to have a completely open AAC tool for FreeBSD also. But until that’s possible, I think that the FreeBSD community appreciates that it’s possible to manage the hardware now using the existing resources.
I’ve been planning a migration of a modest number of servers and firewalls over to obsd over the past few months. Part of that has been reading mailing lists and archives to get a feel for things.
This whole adaptec thing has really sealed the deal for me. I think a flat our refusal to include the driver in the next obsd distrobution is 100% inline with the obsd mission statement of security and stability. Even if it was included with a warning, someone would use it, and then it would be history that obsd shipped a bad driver (even though it was warned to be bad).
I don’t personally have an adaptec raid controller on any of my machines, but will be in the market over the next year for a few new ones. Probably in the range of the 3.8 release timeframe so with any luck this will be resolved and they can be in the list to check out.
Henning Brauer just posted on Undeadly that LSI MegaRAID (ami) would be your best bet then, seems they’re likely to have management by November.
While Stallman just talks about it, while Linus accepts binary drivers in the Linux kernel, Theo is the one consistently fighting and denouncing the spin doctors and enterprises that will not truly cater to FLOSS community, and Theo is the only one coding!!
All hail Theo!!!!!!
This is Theo’s answer to Scott Long. It was posted on OpenBSD’s misc list. I believe it’s in the best interest of the readers to see who stands for Free Software.
“Thanks for going to a public forum and saying I am full of crap.
I really appreciate that. Boy, you sure do want to see all of
our projects do well, don’t you.
Apparently you have zero idea of where we are going.
While you are content with shipping binary stuff in your source tree
and in your ports tree, we are not. We do not ship binaries. We are
not interested in shipping a binary for some CLI. We actually do have
the Linux CLI working in emulation, but we will not supply it to our
user community. I have cancelled that effort by that developer. We
will not supply something to our user community that they cannot fix
and improve themselves.
We have been talking with Adaptec for 4 months. They have not
given us management information.
We have been talking to Adaptec for more than a year to get other RAID
controller information, as in, how to even get the mailbox stuff
fixed. They have not given that to us, either.
Noone thought to talk to you. You are, I am sure, under a
non-disclosure agreement with Adaptec, and I am sure you would
therefore not give us documentation. We are quite used to FreeBSD and
Linux people signing NDA’s by now. Yesterday on the phone Doug said
“But we did give OpenBSD documentation, we gave them to Scott Long”.
Thus, Doug mentioned that *you* had documentation, and thought that
was enough. Of course it is not. You do not help us, I told him.
That is not how it works. And so it stands — we still have no
documentation.
Did I get an offer from you for documentation before you went onto a
public site and said I was full of crap? No, I did not.
And I expect that now that you have said I am full of crap, we still
will get no documentation from you. Right?
We are working on a driver-independent raid management framework. One
command (perhaps called raidctl(4), we don’t know) that should work on
any controller from any vendor, which would do management, because the
management stuff would be abstracted in a driver-independent way into
each driver. Yes this is a difficult project. We have support for
AMI almost working. We will support some other product, as well, then
we’ll see where Adaptec stands.
I do a lot of work on OpenBSD. I am sure that you do a lot of work on
your stuff in FreeBSD too, so you know what it is to be a very busy
busy person.
When a vendor ignores me and the efforts of 4 other people trying to
get the vendor to listen — for that long, we have no choice.
Yet, you, Scott, you think that you are therefore able to slag us and
call us wrong, because YOU are in the loop and we are not? Because
you used to WORK at Adaptec, and we did not? That somehow makes us
full of crap?
I have been watching the mail going to Doug over the last 24 hours.
I have been counting controllers mentioned in mails and am now up to
over 1,800 Adaptec RAID controllers, with people from very large
commercial operations complaining that they have been switching to
other controllers (or, having now seen Adaptec’s failure in this
regard, that they will now actively not buy Adaptec again).
Those controllers will not be supported in OpenBSD 3.7 in May. If
Adaptec wishes them to be supported in a future release, they had
better come and make amends. We are sick of supporting the hardware
of vendors who shit on their customers via us. Maybe they can repair
this horrid situation enough that we will once again support their
controllers by the time OpenBSD 3.8 ships in November.
Quite frankly, you don’t understand what we are trying to do, and
Scott, this is just like the binary only Atheros driver that FreeBSD
ships.
I like it when all hardware is supported with source code, but just
because our methods for getting there are different than yours, Scott,
that gives you absolutely no right to go posting such a thing as you
did there.
Shame on you.”
Scott, considering you worked on the code for the thing I felt calling it “your” was valid, as it was reasonably small I felt that calling it “little” was valid amd as it is a tool and not a driver I felt that calling it a “tool” valid.
Sorry for any confusion.
You say that Theo didn’t contact you?
I’m not surprised: if I were working on a software device driver for a hardware component, the last thing that I would do is contact an ex-employee of the hardware vendor — it could cause all sorts of current and future complications; potentially the vendor could put OpenBSD in hot water for utilising an ex-employee.
You may have worked on the driver for FreeBSD/Linux/etc, but that doesn’t make you the “open source contact” for these devices.
It’s completely correct that Theo and the team approach Adaptec themselves. He wants the _offical_ documetation and support — not the documentation and support of an ex-employee hacker (in the positive sense).
Sorry, but you’re not (and shouldn’t be) in the picture for this.
Theo’s got the correct agenda: he’s one of the few people who have stood up and challenged vendors on their policy over open access to their hardware. I hope he and the OpenBSD team continue to put vocal pressure and use direct action, because it has been having an effect.
Quoted from Theo’s reply: “We are working on a driver-independent raid management framework. One
command (perhaps called raidctl(4), we don’t know) that should work on
any controller from any vendor, which would do management, because the
management stuff would be abstracted in a driver-independent way into
each driver. Yes this is a difficult project. We have support for
AMI almost working. We will support some other product, as well, then
we’ll see where Adaptec stands.”
Okay, now that’s pretty freaking cool. A device independent management tool for managing RAID controllers. That would truely be a pretty usefull thing to have. No matter what controller you have, the interface would probably pretty much be the same. The idea of it is pretty awesome. I hope the OpenBSD suceeds in this endeavour.
I have some RAID hardware at work that I’d like to try it on. I just have to find out what RAID chips they use. We mostly have HP severs. I know some people hate Compaq/HP servers, but they’ve been very reliable for us. Almost as reliable as our AS/400 if that’s ever possible because AS/400’s don’t break (ok, rarely break).
I guess we’ll be waiting until the end of the year though.
Good luck OpenBSD!
just want to add my vouce: Theo is right on this one.
I hadn’t actually realised that OpenBSD was against inary drivers – and I might reconsier using OpenBSD more often in the future.
I just want to voice my opinion too that Theo and OpenBSD are doing a really impressive work. OpenBSD’s upfront/righteous/honest demands on hardware vendors is a beacon of light in the darkness of short-sightedness (linux *cough*) that is so common these days.
And raidctl sounds awesome..
Scott, you seems to miss the point there: OpenBSD developpers surelly had seen your name on FreeBSD driver, but why should have they known that you’re the man with the documentation and the man who signed NDA with Adaptec ?
Why do you take everything as a personal offence (for them to not ask you for help, to ignore your knowledge upon Adaptec hardware, to say that the driver is broken etc.) ? Are you universaly famous ?
So what do you feel strange in they’re demarche ? they adress to Adaptec for more documentation. Not to an obscur FreeBSD developper that suddently and lately emerges.
And … Adaptec, it’s not the moon, it’s just documentation, why the hell are you denying this ? and only documentation about the protocol that permit to manage the RAID controler state …
Theo conducts himself shamelessly and his cadre laud him for his commitment to freedom. Scott even went so far as to just discuss the possibility of getting OpenBSD’s broken driver functioning at the level of the one in FreeBSD, only to be met by “ONLY DOCUMENTATION IS GOOD ENOUGH FOR ME!” from a user and strawmen about the status of the management binary.
All of this public drama really makes all of the parties involved appear unprofessional, and only serves to devalue Theo’s own ultimatum.
> Doesn’t Intel, AMD, SUN, to mention a few, release specifications for their processors? Why are they still in business then? And why isn’t Tom, Dick and Harry cloning their processors?
Why stop at register specification, how about asking for chip masks and verilog sims to be opened out. I’m sure you’ve heard that the Chinese were able to clone a GM car – cloning chips would be absolutely no problem for them at all.
Is there really a competative advantage in keeping technical documentation and implementations in a binary only form?
For some applications this is a certainty. This is where software patents are hoping to go. More and more hardware devices are using more or less generic processors, memory and parts. The actual way it works seems to be all done in software.
A good example is graphics cards. Even under Windows good/bad drivers make a massive difference in performance meaning that the actual function of the of the device is probably all in software. There have been efforts to harness all that GPU processing power for normal applications.
Why should a company care then if the documentation for just the hardware is available. Well probably because without all that convoluted software the device doesn’t work at all.
All that means is that the hardware company can’t protect its trade secrets or competative advantage by making the hardware obscure.
Is this a bad thing for Open Source. Undoubtably. However I think its a trend that is set to continue. I suspect that the recent surge in GPU performance is a temporary trend set to reverse itself as the focus switches back more generic, general use processors. Other hardware will likely tend to the way of the software modems.
Though I am in no position to comment one way or another, there is a rumbling that Adaptec has no docs to give on the lists.
The rumour that they only have a source implementation available and that’s it scares me. That would suck for any developer working there.
Though, Scott could definately answer that one straight.