Linked by Thom Holwerda on Fri 3rd Aug 2007 15:42 UTC
Talk, Rumors, X Versus Y "Last week at OSCON someone set up a whiteboard with the heading 'Tools We Wish We Had'. People added entries (wiki-style); this one in particular caught my eye: 'dtrace for Linux, or something similar'. So what exactly were they asking for? "
Order by: Score:
Sight
by diegocg on Fri 3rd Aug 2007 16:16 UTC
diegocg
Member since:
2005-07-08

It's a sad thing that kprobes has been available way before dtrace was available (a "Dynamic probes for linux" was presented in Freenix in _2001_) , still sun has managed to finish a complete solution before linux did. Developers doesn't seems to care about kprobes+systemtap, nobody seems to hack in it except IBM and redhat (and there're not lot of resources on it apparently)....when are they going to include the patches to probe userspace code, and so on???

Edited 2007-08-03 16:18

Reply Score: 5

strained logic
by butters on Fri 3rd Aug 2007 16:52 UTC
butters
Member since:
2005-07-08

This guy is trying to argue that Linux developers have something against DTrace that makes them want to copy it badly. There are only a few things that Linux devs don't like about DTrace: it's incompatible with the GPL, it requires tight kernel/userspace integration, and it has its own programming language.

But apparently the licensing issues are FUD, because porting DTrace to Linux wouldn't violate the CDDL. Yes, but it would violate the GPL. The solution presented in the comment thread is for the Linux project to treat kernel modules like userspace applications. Right... because a module linked into the kernel is the same thing as a user process trapping into the kernel.

But we already put nasty binary blobs in the kernel, right? Sometimes, yes, but these modules are never distributed with the kernel, and they're almost always device drivers. Only users with closed hardware need them. Here we're talking about loading non-GPL code into the kernel as a part of a pervasive tracing framework. We may sometimes use non-GPL code for hardware enablement, but not for core functionality.

Finally, the claim that SystemTap is a hopeless knockoff is a stretch. Yes, SystemTap works by generating and loading a kernel module, whereas DTrace activates code paths already in the kernel. This avoids evaluating a branch at every probe point to see if it's enabled, but apparently it's a security concern. What makes loading a module any less secure that activating dormant kernel code?

The other cheap knockoff claims are that SystemTap doesn't cover a lot of userspace yet, and it doesn't have a couple of DTrace's features. One of them, speculative tracing, sounds a lot like a way around the branch evaluation problem I mentioned before, which doesn't exist in the SystemTap implementation. So SystemTap isn't quite at feature-parity with DTrace yet, which is understandable given that Sun had a four-year headstart.

Nothing but a hopeless knockoff developed by people who refuse to adopt DTrace out of stubborn arrogance. Yeah, that's exactly what's going on...

More like this:

https://ols2006.108.redhat.com/2007/Reprints/keniston-Reprint.pdf

Edited 2007-08-03 16:55

Reply Score: 5

RE: strained logic
by bcantrill on Fri 3rd Aug 2007 17:14 UTC in reply to "strained logic"
bcantrill Member since:
2005-09-16

The safety issue has nothing to do with "activating dormant kernel code" (which DTrace does not do, by the way), but rather with the fact that the SystemTap user is implicitly generating native object files and loading them into the kernel, where the DTrace user is implicitly generating virtual machine instructions and sending them into the kernel to be (safely) executed. These are night-and-day different approaches, as I outlined extensively here:

http://blogs.sun.com/bmc/entry/dtrace_safety

As you seem to have little understanding of the mechanics of DTrace, I would also refer you to our 2004 USENIX paper:

http://www.sun.com/bigadmin/content/dtrace/dtrace_usenix.pdf

And to an article that I wrote last year in ACM Queue:

http://www.acmqueue.org/modules.php?name=Content&pa=showpage&pid=36...

In particular, if one reads the ACM Queue article with SystemTap in mind, one cannot help but see the stark contrast in the two approaches -- SystemTap is simply not endeavoring to solve the same problem that DTrace solves.

Reply Score: 22

RE[2]: strained logic
by milek on Fri 3rd Aug 2007 17:56 UTC in reply to "RE: strained logic"
milek Member since:
2006-02-20

I've been using DTrqace for years now in a production. I have never crashed system with dtrace, while with SystemTap it was very easy (no, I didn't wanted to crash a system with SystemTap - it just did).
And as someone else already mentioned - lot of people and companies are using dtrace and basicaly nobody is using SystemTap except for a quick 'adventure'.

Reply Score: 7

Palpable
by tony on Fri 3rd Aug 2007 18:39 UTC
tony
Member since:
2005-07-06

Wow, the animosity between Sun advocates and Linux advocates continues to escalate. I think that's unfortunate for both parties, especially for people out there who use and enjoy both.

From Sun's perspective, I wonder how wise this is. The users that they want are almost all in the Linux camp. Picking fights tends to embolden, and not convert.

It's interesting how the roles are reversed now. Linux was the scrappy underdog, taking on a powerful but complacent and overconfident foe. Now, Sun is going guerrilla, the scrappy underdog taking on the Linux institution. If the Linux community does become too complacent, that would be Sun's opportunity.

Reply Score: 2

RE: Palpable
by sbergman27 on Fri 3rd Aug 2007 23:05 UTC in reply to "Palpable"
sbergman27 Member since:
2005-07-24

"""
Wow, the animosity between Sun advocates and Linux advocates continues to escalate.
"""

Bzzzt! Thanks for playing, though. Carol will give you a month's supply of Rice-a-Roni on the way out.

I see no real evidence of escalating animosity on the parts of respective developers. Maybe among the respective fanboys. But that's what fanboys do. Does it really matter?

I'm a Linux advocate and I congratulate Sun on their excellent Dtrace tool. ZFS's admin tools look nice, too.

Reply Score: 4

RE: Palpable
by shapeshifter on Fri 3rd Aug 2007 23:33 UTC in reply to "Palpable"
shapeshifter Member since:
2006-09-19

Wow, the animosity between Sun advocates and Linux advocates continues to escalate. I think that's unfortunate for both parties, especially for people out there who use and enjoy both.

From Sun's perspective, I wonder how wise this is. The users that they want are almost all in the Linux camp. Picking fights tends to embolden, and not convert.

It's interesting how the roles are reversed now. Linux was the scrappy underdog, taking on a powerful but complacent and overconfident foe. Now, Sun is going guerrilla, the scrappy underdog taking on the Linux institution. If the Linux community does become too complacent, that would be Sun's opportunity.


Yeah, Sun developers are jealous.
Linux is getting all the attention and nobody talks about Solaris.
It's their fault (Sun's) though and they only have themselves to blame.
For so many years they've been playing in their own little sandbox ignoring the issues at large (mainly Microsoft's monopoly encroaching on every aspect of computing and wide adoption of x86 architecture hardware).
Linux has become successful by the hard work of unselfish people, dedicated community, and determined philosophy which manifests itself via the GPL that protects all the Linux code.
Linux now runs on off the shelf boxes, laptops, but also "big-iron" servers.
And Solaris is desperately catching up. But it's too late.
Sun releases a bit of code under a license purposefully incompatible with GPL and expects the world to throw its arms around them, jump onto the Solaris wagon, and forget about Linux.
So Sun and Solaris developers, come back with a little humility and some genuine willingness to cooperate and maybe then we'll talk.

Reply Score: 0

RE[2]: Palpable
by Valhalla on Sat 4th Aug 2007 00:11 UTC in reply to "RE: Palpable"
Valhalla Member since:
2006-01-24

shapeshifter wrote:
-"Yeah, Sun developers are jealous.
Linux is getting all the attention and nobody talks about Solaris.
It's their fault (Sun's) though and they only have themselves to blame.
For so many years they've been playing in their own little sandbox ignoring the issues at large (mainly Microsoft's monopoly encroaching on every aspect of computing and wide adoption of x86 architecture hardware).
Linux has become successful by the hard work of unselfish people, dedicated community, and determined philosophy which manifests itself via the GPL that protects all the Linux code.
Linux now runs on off the shelf boxes, laptops, but also "big-iron" servers.
And Solaris is desperately catching up. But it's too late.
Sun releases a bit of code under a license purposefully incompatible with GPL and expects the world to throw its arms around them, jump onto the Solaris wagon, and forget about Linux.
So Sun and Solaris developers, come back with a little humility and some genuine willingness to cooperate and maybe then we'll talk."


an online dictionary explaning the word 'bias' could do worse than link to this comment ;)

Reply Score: 4

RE[3]: Palpable
by shapeshifter on Sat 4th Aug 2007 07:34 UTC in reply to "RE[2]: Palpable"
shapeshifter Member since:
2006-09-19

an online dictionary explaning the word 'bias' could do worse than link to this comment ;)


Yeah, ofcourse I'm biased. So what's your point?

Reply Score: 0

RE[2]: Palpable
by Jondice on Sat 4th Aug 2007 00:16 UTC in reply to "RE: Palpable"
Jondice Member since:
2006-09-20

Stop trolling, we've all heard that one at least ten times before, and it didn't even sound good the first time.

A bit of code? Yeah, the most popular UNIX OS ever, the most popular VM ever, the second most popular Office Suite, and much more. I'd say that qualifies for a bit of code.

I think that (momentarily at least) making the code GPL incompatible may be a good thing for Solaris and Linux. As someone said above, maybe Linux will have to reinvent the wheel, and maybe it won't be a bad thing. As for Solaris, instead of just being zealous about this one OS, try to respect each of them. Obviously, Sun is a company and has a company's needs. Linux is also largely driven by corporate needs today. That's a good thing. That means people are getting paid to work on the OS. Sun isn't the only one that can make money from Solaris or Java, as should be easy to see. One interesting fact is that HP sells more Solaris servers than Sun.

Reply Score: 5

RE[3]: Palpable
by shapeshifter on Sat 4th Aug 2007 08:00 UTC in reply to "RE[2]: Palpable"
shapeshifter Member since:
2006-09-19

Stop trolling, we've all heard that one at least ten times before, and it didn't even sound good the first time.

A bit of code? Yeah, the most popular UNIX OS ever, the most popular VM ever, the second most popular Office Suite, and much more. I'd say that qualifies for a bit of code.


Lol, yeah, the most popular UNIX OS, but Linux is the most popular Unix-like OS.
Well, although I myself don't mind Java and regularly use java apps, I wouldn't say that Java is exactly loved by the end users. Some people even hate it with passion. And just about everyone prefers a native app instead of a Java app.
And for the longest time Java was utter crap and everyone hated using it. Only in recent times it got a bit better.

I think that (momentarily at least) making the code GPL incompatible may be a good thing for Solaris and Linux. As someone said above, maybe Linux will have to reinvent the wheel, and maybe it won't be a bad thing. As for Solaris, instead of just being zealous about this one OS, try to respect each of them. Obviously, Sun is a company and has a company's needs. Linux is also largely driven by corporate needs today. That's a good thing. That means people are getting paid to work on the OS. Sun isn't the only one that can make money from Solaris or Java, as should be easy to see. One interesting fact is that HP sells more Solaris servers than Sun.


Well, I don't agree that being driven by corporate needs is a good thing.
I want Linux to be by the people for the people and not by a corporation for a corporation.
I can't find one good thing that Sun has ever done for Linux and OSS.
Whatever code they released, they released under a fake oss license, incompatible with Linux.
If they released anything under GPL or license compatible with it, then it's only because alternatives existed or were shaping up nicely in Linux, so Sun had no longer anything to lose.
And as for contributing to oss projects, well, the jury is still out on that one.
At this point Openoffice.org is in pretty much unusable state.
Blurred menu fonts, jumping page thanks to the idiotic context sensitive toolbars, and no universal installer on Linux any more. And there is no option for end user to fix it themselves.
I've offered Oo.org to so many people and every single one of them rejected it. Pretty sad really.
So I am starting to suspect a conspiracy, Sun working for Microsoft, keeping adoption of Linux on the desktop slow thanks to no office suite available.
What better way to stifle Linux adoption than to offer an office suite and then keep it in a crappy state for years so nobody wants to use it.
There is no Linux desktop without an office suite, it's just not going to happen.
So, Sun, are you conspiring with Microsoft against Linux?
As if they'll tell the truth.

Reply Score: 0

(If You Love Somebody) Set Them Free
by pseudocode on Fri 3rd Aug 2007 21:11 UTC
pseudocode
Member since:
2007-05-30

Wow, the animosity between Sun advocates and Linux advocates continues to escalate.


It's not really between Sun and Linux, but between GPL and the-rest-of-the-world.

From the developers point of view: GPL is "fun" because it let you code, share your code and use someone else's code.

From the GPL-ists point of view: If the wheel is not under GPL, you have to reinvent it.

Edited 2007-08-03 21:12

Reply Score: 6

zztaz Member since:
2006-09-16

Or you could say that's between CDDL and the rest of the world.

Licenses matter. Putting code into the public domain has a poor track record in terms of building communities and long term support. Typical closed-source EULA's are even worse. A number of licenses, open and free in slightly different ways, have proven track records. The successful licenses all have limits or requirements of some sort.

The GPL is one of the most successful licenses available, so it must achieve a good balance between the conflicting goals of developers and users. Companies seem to like the fairness that comes with the tit-for-tat aspect; you can use my enhancements only if you make your own available, too. But the very features that make the GPL work make it incompatible with some other licenses.

The CDDL meets Sun's corporate needs. Again, the very features that Sun desired make it undesirable to others. That's the nature of licenses, and engineers should be used to the idea that optimizing for one parameter means trading off something else.

There is no one license that works for everyone. Some want the very feature that others object to. The licenses for the Linux kernel and DTrace don't mix, and neither will change. Perhaps that might have been avoided, but each project chose a license, and we must live with those choices. It's a done deal, drop it and move on.

Reinventing the wheel is not always a bad idea. Wheels were in use for thousands of years before someone thought of putting wheels on rails. Wheels had been in use for thousands of years before someone invented the pneumatic tire. You have to willing to reinvent the wheel before you can improve the wheel.

Linux developers may be behind in this sort of tool, but they now have an opportunity to observe DTrace in real-world use. Maybe they can reinvent the parts of DTrace that work the best, drop parts that are less useful, and improve on it here and there. There were license issues with BitKeeper, too, and that resulted in Git.

Reply Score: 7

renox Member since:
2005-07-06

I agree with your points except this one:
>>Putting code into the public domain has a poor track record in terms of building communities and long term support.<<

Uh, I wouldn't consider all the BSD OS, X Windows, Apache, a 'poor track record' sure Linux is even more successful but still that doesn't make those project unsuccessful..

Unless you consider 'gift' licenses (BSDv2, MIT, Apache) and 'public domain' different?
For all pratical purpose, there is no difference as those 'gift' licenses have basically no restriction.


About CDDL filling Sun business need: yes, and one of the big business need is to be incompatible with the Linux kernel.
I have no problem with this, I just wish that Sun would stop saying that Linux developers should port their CDDL code and acting surprised when Linux developers don't: that's also dishonest!

Reply Score: 4

zztaz Member since:
2006-09-16

Releasing software under ANY license differs from placing it in the public domain. You surrender ownership when you place your work in the public domain. With the BSD license, you keep the copyright but grant broad permissions to people who don't hold the copyright. I use the BSD license as an example, but the effects are similar across a range of licenses.

BSD might seem as permissive as public domain, but there is an essential difference: a copyright notice indicating who owns the rights. It's hard to build a community if you don't know who the founding members are.

BSD, X, Apache and other projects are successful because they impose at least some terms, however minimal, on distribution. That was the point I was trying to make. They are also successful because the projects have a formal organization with rules regarding membership, who can commit code changes, and mechanisms for resolving disputes. Projects often fail when they neglect to define rules.

DTrace and Linux follow different rules. I don't care how many developers tell me not to worry about how those rules may or may not conflict. It's not a technical issue. Show me a professional legal opinion that CDDL code and GPL code may be combined and distributed by third parties, and I'll stop worrying.

'Open' is not a synonym for 'anarchy'. 'Open' does not mean 'no rules', it means that the rules are published and apply equally to all. I respect the GPL and CDDL, among other licenses. I don't speak for Linux or Sun, nor am I a lawyer, so my identity doesn't matter. But I have a question for the 'just do it' crowd:

Which license do you want me to violate? Either, or both?

Reply Score: 1

John Nilsson Member since:
2005-07-06

So you are saying that without the copyright regime there would no way of knowing who created what software, and as a result there would be no way to build communities around software projectes?

Do yuo have anything to back that up?

Reply Score: 2

We'll help
by bgregg on Sat 4th Aug 2007 04:15 UTC
bgregg
Member since:
2007-08-04

G'Day - I'm the "Brendan Gregg" who Adam referred to in his blog entry. I've used DTrace heavily since early 2004, including authoring most of the scripts in the DTraceToolkit. Several months ago I joined Sun.

I do feel sorry for the Linux end users. Before I joined Sun I did use Linux (and Solaris), and frequently worked with Linux professionals. I know how they feel (and I felt).

DTrace was publically available and CDDL'd over two years ago. There was already a precedent for non-GPL code in the Linux kernel (nvidia and others). But instead of porting DTrace to Linux, we've had two years of license battles and promises. How does that help me, the end user? ... Hello?

I'd still love to see DTrace on Linux - one reason is so that I can port the DTraceToolkit to Linux. I'm looking forward to it as I really love writing DTrace scripts, and as a performance and troubleshooting guy the capabilities of DTrace are mindblowing. DTrace will be a big part of my life until the day I die; and I'm hoping that Linux can be part of that too.

There are others like me in the DTrace community who will no doubt help to make Linux DTrace a success. Just like we'll strive to help make MacOS DTrace a success, and wherever else DTrace is ported to.

If (or when) the DTrace Linux port does finally happen, I'll not only help out with scripts, but I'll help out with how-to documentation and presentations too. No need to borrow without asking. ;-)

Reply Score: 7

RE: We'll help
by smitty on Sat 4th Aug 2007 05:19 UTC in reply to "We'll help"
smitty Member since:
2005-10-13

If (or when) the DTrace Linux port does finally happen

It is NEVER going to happen, and neither will ZFS at least in the kernel. As you said, there is precedent for putting non GPL code in the kernel, like the nvidia drivers. But it is important to note that the kernel devs do not put that in themselves and it isn't in the vanilla official kernel. In fact, one of the LiveCD distros got in legal trouble just for distributing a kernel with the 3d drivers in it. An analogous situation would be Sun providing ZFS and DTrace ports that individual users could download from their website and load into their own kernels manually. Sun has no interest in doing this. Perhaps some other 3rd party will take up the task?

Sun could end this whole mess right now by dual licensing their code under the GPL, but for many reasons they don't want to. Which is their right. They can put the code under whatever license they want to, but to put it under an incompatible license and then complain it isn't being used in Linux is a little ridiculous. I think they know this and just like the attention it brings them and being able to point to a very concrete and useful advantage Solaris has over Linux right now.

The real losers right now are the linux users who have a subpar implementation, but hopefully that will change in the future.

Reply Score: 3

RE[2]: We'll help
by bgregg on Sat 4th Aug 2007 21:16 UTC in reply to "RE: We'll help"
bgregg Member since:
2007-08-04

It is surprising how many people have posted to say that CDDL DTrace Linux will never happen, that this was Sun's plan all along, that Sun isn't genuinely interested in co-operating, etc.

I'm sorry, but who are these people?

smitty: Matt
shapeshifter: Shapeshifter
zztaz: zztaz
pseudocode: xphilipp

Those are the realnames provided to their osnews profiles. You might want to update them, if you expect any more credibility than an anonymous coward.

Both Sun's CEO and members of team DTrace have said that they want to see DTrace on Linux. This isn't some conspiracy. I work with team DTrace, we've discussed this at length, and it is genuinely what we want.

So if you'd like to post about what Sun thinks or is planning - then you'll need to explain how you know better than both Sun's CEO and team DTrace.

And if you'd like to post about what Linux would never do, then please explain how you know this. Are you a respected spokesperson for Linux? Are you a major contributor to the Linux kernel?

Reply Score: 5

RE[3]: We'll help
by smitty on Sat 4th Aug 2007 23:17 UTC in reply to "RE[2]: We'll help"
smitty Member since:
2005-10-13

Those are the realnames provided to their osnews profiles. You might want to update them, if you expect any more credibility than an anonymous coward.

Actually, my name really is Matt. You might be able to guess my last name as well. :> If you're looking for my address and home phone #, sorry. But I'm not putting those out on the internet, even if you do call me an anonymous coward.

This isn't some conspiracy.

I never said it was. Just that Sun has no intention of putting DTrace of ZFS under the GPL license and that unless that happens, it is unlikely to get into the kernel.

So if you'd like to post about what Sun thinks or is planning - then you'll need to explain how you know better than both Sun's CEO and team DTrace.

If you're saying that Sun is going to put it under the GPL, or code up the support in Linux themselves and provide it as a 3rd party, well that's great news for linux users. Somehow, I don't see them doing this though. I'd be happy to be proven wrong.

And if you'd like to post about what Linux would never do, then please explain how you know this. Are you a respected spokesperson for Linux? Are you a major contributor to the Linux kernel?

No, I have no connection to Linux at all (except as a user) but I do read the LKML sometimes. And every time something about CDDL code comes up (like getting ZFS support in the kernel) dozens of real kernel developers post there that it isn't going to happen because the CDDL is incompatible with GPL. If you don't believe me, then I'm not sure how I can convince you - it's right there on the internet for everyone to see.

Do you have some lawyer telling you that it is legal?

I work with team DTrace

Congratulations, and thanks for the great tool you helped create.

Edited 2007-08-04 23:32

Reply Score: 3

RE[4]: We'll help
by bgregg on Sun 5th Aug 2007 03:50 UTC in reply to "RE[3]: We'll help"
bgregg Member since:
2007-08-04

Actually, my name really is Matt. You might be able to guess my last name as well. :> If you're looking for my address and home phone #, sorry.

No, I'm not after addresses and phone numbers, I just want to know who I'm talking to - full name and company name. If company name isn't appropriate, then some other way to identify you from others with the same name (city/country, authored product, etc).

I do read the LKML sometimes. And every time something about CDDL code comes up (like getting ZFS support in the kernel) dozens of real kernel developers post there

Thanks - that is useful to know. I'm now catching up on LKML mail.

If you're saying that Sun is going to put it under the GPL, or code up the support in Linux themselves and provide it as a 3rd party, well that's great news for linux users.

Why must it be Sun that ports and provides it as a 3rd party? Sounds like an opportunity for another company, backed with legal advise. And it shouldn't take much time to do a port; perhaps several weeks. Sun's DTrace engineers will help too, as time permits, just like with the MacOS port.

Do you have some lawyer telling you that it is legal?

I can't say whether it is or isn't, but some good points were raised in the comments on Adam's blog, which may mean that this is a non-issue (just like it is in user-land).

Reply Score: 3

RE[5]: We'll help
by smitty on Sun 5th Aug 2007 05:35 UTC in reply to "RE[4]: We'll help"
smitty Member since:
2005-10-13

I just want to know who I'm talking to - full name and company name

I suppose I could give that, but really it doesn't have any bearing on the subject. We're a MS shop and don't use any Solaris/Linux at all, except for one old svn server. I never meant to give any other impression, but I still think my thoughts on the subject are valid even if I'm not an insider like you.

Thanks - that is useful to know. I'm now catching up on LKML mail.

I had rather assumed everyone at Sun was already aware of the discussions on the LKML, and that was why I earlier assumed that there had to be some other reason for bringing up the subject again. After all, if I had an advantage over my competitor I'd be talking about it too.

Why must it be Sun that ports and provides it as a 3rd party? Sounds like an opportunity for another company, backed with legal advise. And it shouldn't take much time to do a port; perhaps several weeks.

It doesn't, but you did say Sun wanted to see it on Linux. Just not that badly? Anyway, I don't think anyone really expects Sun to do so. In my mind they've done plenty just creating the system and releasing it as OSS. I just get tired of people always blaming the GPL for this situation and then usually commenting about how superior the BSDs are. I guess I get a little cranky when the subject comes up. ;) I had assumed that it would take a bit longer than a couple weeks to do the port - if it could be done that easily then perhaps another 3rd party will step up. I still think the legal situation would make that tricky, though, and getting it into the main branch is just not going to happen. It would be nice if they could come up with a compatible system, though, so that users could take their DTrace knowledge from one system and use it on the other.

Edited 2007-08-05 05:42

Reply Score: 3

RE[3]: We'll help
by renox on Sat 4th Aug 2007 23:22 UTC in reply to "RE[2]: We'll help"
renox Member since:
2005-07-06

>>Both Sun's CEO and members of team DTrace have said that they want to see DTrace on Linux. This isn't some conspiracy. I work with team DTrace, we've discussed this at length, and it is genuinely what we want.<<

But they also chose an incompatible license: acts are more reliable than speech.
Somehow, it's hard to believe the 'genuine' part, when the excuse for not having either a different compatible license or a dual license are just that: excuse.

As for deducing what Linux kernel developers will do or won't do: well you know, many of us have been reading linux kernel mailing list for years now, so 'never' is too strong, but according to their past behaviour, it's very unlikely that CDDL code would be accepted in the main kernel(s)..

Too bad though: it would really be nice for developers if there was only one unified API to instrument user space.
Same thing for D script: it would be nice for system administrators to use the same scripts on MacOS, *BSD, OpenSolaris and Linux!

Reply Score: 2

RE[4]: We'll help
by drdoug on Sun 5th Aug 2007 02:17 UTC in reply to "RE[3]: We'll help"
drdoug Member since:
2006-04-30

> But they also chose an incompatible license: acts are
> more reliable than speech.
> Somehow, it's hard to believe the 'genuine' part, when > the excuse for not having either a different compatible > license or a dual license are just that: excuse.

The choice for a 'incompatible license' was purely the fact that OpenSolaris could not be released under GPL at the time as Sun did not have the rights to change the license of some of the code. Sorry, 'no excuse' is just the way the cookie crumbles.

Sun have demonstrated that when they can they do release code and hardware under GPL.

Reply Score: 3

RE[5]: We'll help
by mjg59 on Sun 5th Aug 2007 04:16 UTC in reply to "RE[4]: We'll help"
mjg59 Member since:
2005-10-17

at the time as Sun did not have the rights to change the license of some of the code

Then how did they have the rights to release it under the CDDL? The CDDL didn't exist when the vast majority of Solaris was written.

Making the "Sun released some code under the GPL, therefore they'd have released all of it under the GPL if they could" argument assumes that Sun is a single homogeneous body. It's not. Different sections of the company behave differently, have different traditions and have different beliefs. The way that Sun has developed over the years means that there's plenty of internal power conflicts, and the somewhat odd approach to releasing different chunks of code under different licenses is a reflection of this.

Reply Score: 1

RE[6]: We'll help
by drdoug on Mon 6th Aug 2007 08:42 UTC in reply to "RE[5]: We'll help"
drdoug Member since:
2006-04-30

>> at the time as Sun did not have the rights to change > the license of some of the code
>
> Then how did they have the rights to release it under
> the CDDL? The CDDL didn't exist when the vast majority
> of Solaris was written.

They didn't release the source code of the parts which they did not have the permission to release. The closed source parts of Solaris are released in binary form. This is allowed under CDDL.

Reply Score: 2

RE[3]: We'll help
by mjg59 on Sun 5th Aug 2007 04:32 UTC in reply to "RE[2]: We'll help"
mjg59 Member since:
2005-10-17

Are you a respected spokesperson for Linux?

Probably not, but my opinion is consistent with that expressed by pretty much every Linux kernel developer I've had this discussion with. However, as one of the four members of the technical board that determines the technical direction of Ubuntu, I'm probably a respected spokesperson for one of the largest distributions.

Are you a major contributor to the Linux kernel?

Three different Linux distributions have expressed interest in hiring me over the past month, along with a large company in the Linux market with a market cap something like 10x that of Sun. Linus occasionally emails me with questions. I wouldn't necessarily say "Major", but I'd probably go with "Significant".

I spent a while discussing this issue on Adam's blog. My opinion (and, I believe, the opinion of the majority of core Linux contributors) is that there's no realistic chance of a CDDLed body of code being incorporated into the core Linux kernel. Nor is there any real chance of a GPLed shim layer being incorporated in order to implement enough of the Solaris kernel ABI to allow a set of CDDLed modules to take be loaded.

I'm not going to say that the release of DTrace under the CDDL is any sort of deliberate snub to the Linux community - the fact that Sun have released their code under any sort of open license at all is a massively generous donation to the community, and I sincerely hope that it benefits Sun in the long run. However, one of the consequences of the license choice is that it's not obviously possible to incorporate chunks of that code into Linux. Linux misses out on the functionality and Solaris misses out on the developer base. I think finding some way to make this work would benefit both sides, but I strongly suspect that the only solution would be to relicense the code in question.

Reply Score: 2

RE[3]: We'll help
by shapeshifter on Sun 5th Aug 2007 23:48 UTC in reply to "RE[2]: We'll help"
shapeshifter Member since:
2006-09-19

It is surprising how many people have posted to say that CDDL DTrace Linux will never happen, that this was Sun's plan all along, that Sun isn't genuinely interested in co-operating, etc.

I'm sorry, but who are these people?

smitty: Matt
shapeshifter: Shapeshifter
zztaz: zztaz
pseudocode: xphilipp

Those are the realnames provided to their osnews profiles. You might want to update them, if you expect any more credibility than an anonymous coward.


Umm, dude, sorry this the Internet and a public forum, in case you haven't noticed. We're not giving testimony in front of the Congress.
Some prefere to be anonymous and credibility is only "skin deep".

Both Sun's CEO and members of team DTrace have said that they want to see DTrace on Linux. This isn't some conspiracy. I work with team DTrace, we've discussed this at length, and it is genuinely what we want.
So if you'd like to post about what Sun thinks or is planning - then you'll need to explain how you know better than both Sun's CEO and team DTrace.


Umm, Microsoft has been saying the same thing sometime.
They "want" open standards, cooperation, and interoperability with Linux.
Words say one thing but their actions prove otherwise.
So, sorry if I don't just take your CEO's and yours word for it. Untill we see some code licensed compatible with Linux, it's all just empty words.
Only some actions will prove otherwise.


And if you'd like to post about what Linux would never do, then please explain how you know this. Are you a respected spokesperson for Linux? Are you a major contributor to the Linux kernel?


No, I'm neither. Just a bystander voiceing his opinion.
And just responding to the Adam Leventhal's weblog post which is nothing but talking trash about Linux and its developers.

The whole post can be summed up like this.

"Bhaahahahha, the Linux guys tried to steal Dtrace, went and copied, and came up with SystemTap.
Except they did such a lousy job, SystemTap sucks compared to Dtrace. What a bunch of losers.
Better come to us and if you're nice little Linux boys and beg us, we'll help you port our great, flawless work of art Dtrace to Linux. "

Talking credibility, it's Adam that is on shaky ground here when trashing the work of Linux developers.

Reply Score: 0

CEO
by Luminair on Sat 4th Aug 2007 05:07 UTC
Luminair
Member since:
2007-03-30

The head of Sun responded to that blog post, by the way:

And for the record... as Adam states so eloquently, we're thrilled to see ports of DTrace running everywhere. ZFS and DTrace's availability beyond OpenSolaris/Solaris, on Apple's Mac OS (and BSD) prove the point - joining communities is a good thing.

Posted by Jonathan Schwartz on August 02, 2007 at 02:39 PM PDT #


Edited 2007-08-04 05:08 UTC

Reply Score: 3

Honk! Honk!
by Weeman on Sun 5th Aug 2007 11:16 UTC
Weeman
Member since:
2006-03-20

Just that everyone knows, it was considered to put OS/Net out there GPLv3 licensed, but was rejected by the board of advisors of the OpenSolaris project.

And I'm think they're right to have done this. Because don't let kid ourselves. If it's out as GPL, the Linux developer community will fall over it like locusts.

Sun's idea for Solaris is for it to gain popularity again, that includes the Project Indiana. There won't be GPLv3 Solaris until the desired popularity has been gained. And even then.

Reply Score: 1

@shapeshifter
by aleventhal on Mon 6th Aug 2007 07:03 UTC
aleventhal
Member since:
2005-12-14

yeah, that's pretty much what I was trying to say.

Browser: Mozilla/5.0 (iPhone; U; CPU like Mac OS X; en) AppleWebKit/420+ (KHTML, like Gecko) Version/3.0 Mobile/1C25 Safari/419.3

Reply Score: 2

Licensing issue for FreeBSD too
by renox on Mon 6th Aug 2007 11:52 UTC
renox
Member since:
2005-07-06

Apparently FreeBSD dev won't include DTrace in their v7 release because there are licensing issue for headers:
http://marc.info/?l=freebsd-current&m=117469583227359&w=2

I was wondering how FreeBSD dev could include DTrace in their kernel and keeping the kernel under BSD, the answer is that they will use DTrace as a kind of module not provided inside the kernel (it must be loaded) but to compile the kernel, they need DTrace header and that why they're wondering if it's legal..

Reply Score: 2

Kebabbert
Member since:
2007-07-27

What Sun has "contributed a bit of code"?? Sun is the company that has contributed the most to Open Source, of all companys. There are many studies and articles of this, just google a bit.

Why are there no complaints on IBM aix, or HP ux? Or MICROS~1? Sun has opened up all it's crown jewels (java, Solaris, ZFS, DTrace, layout and specifications of OpenSparc CPU, etc, etc), and its there free for anyone to see and look into. Which other proprietary company has done that? Can you imagine the debate within Sun when talking about giving away all it's crown jewels? No other company has tried what Sun is doing. Not IBM, not HP, not MS, etc. Maybe this will doom Sun? I hope not.

Why all the complaints? We all win on Sun's acting, dont we? Sun has almost nothing left to open up. And still the complaints. Dont you want IBM to act like this? Should Sun go back to closed source again?

Reply Score: 1