
Probably just Russian Meddling
Big Red, eh? Driving us crazy, is that the plan?
Apart from that, this sounds excellent, and I will gladly be driven crazy, even if Vladimir Ellison wins!
DTrace
under the GPL Oracle appears to have open-sourced DTrace, the system instrumentation tool that Sun Microsystems created in the early 2000s and which has been beloved of many-a-sysadmin ever since. As noted by developer Mark J. Wielaard, this commit by an Oracle developer shows that something is afoot. Big Red recently listed DTrace as one …
... that the "ancient thread" is just that ... Absolutely nothing that they were "debating" then has any bearing whatsoever in today's world. And the protagonists are 22 years older and wiser. The world has moved on.
That said, DTrace is a useful tool ... If you're a sysadmin and have never used it, there is no time like the present :-)
Some of what they were debating in that thread 2 decades ago resulted in the Meltdown/Spectre mess we have now.
There are also some other interesting things in that discussion. The Linux TCP stack was faster than the SunOS one at the time because Sun used a modular network stack based on isolated layers of the Streams concept while Linux allowed for anything that worked fast to be in the Kernel. Oddly enough the instance of preserving the modular layered nature of the Linux file system layers is why the Linux ZFS port is behind other operating systems and still would be even if Oracle reopened the source for that.
That "ancient thread" has a large number of familiar names that are still active.
"Some of what they were debating in that thread 2 decades ago resulted in the Meltdown/Spectre mess we have now."
Absolute bullshit. NOTHING that they were talking about had anything at all to do with the decisions about CPU design changes that were already in the works at Intel, AMD etc.
"NOTHING that they were talking about had anything at all to do with the decisions about CPU design changes that were already in the works at Intel, AMD etc."
About that time CPUs were getting fast enough that the average program was doing enough system calls that the multitasking interrupt was no longer the major cause for the kernel to shift things around. Sun had already announced their newer processor and it included lots of performance improvements for doing things like flushing page tables and switching between system mode and user mode and between cores. Intel at the time was still trying to catch up in the server market and introduced their solutions to those problem about 3 years latter. At that time Intel was trying to break into the server market and at least one of the people in the thread did architectural work at Intel.
Nothing in that thread had anything to do with either Spectre or Meltdown happening.
I have to agree. Speculative-execution algorithms and architectural designs date from the late 1960s, and actual on-chip implementations from the late 1980s. The use of hardware side channels to break security on general-purpose computers was public at least since the mid 1990s. That Usenet thread is connected to Spectre attacks (of which Meltdown is a subclass) only in the most tenuous fashion.
You do realize they did the relicensing back in August, right?
You'd figure people would pay attention to changes like this with a major impact on the way sysadmins do their job.
That being said, those of us using FreeBSD, macOS, or old Slowlaris have had DTRACE for what seems like eons now, since like FreeBSD 7.1, Leopard and Solaris 10's initial release. Yet the Penguin bothers will probably act like this is something new and revolutionary, like they usually do when they finally catch up, a decade (or more) later.
Yep. FreeBSD has had DTrace for years.
OSX has had a hobbled form for years too.
However, Ive a lot of Linux platforms to maintain.
Having dtrace on Linux too means I have a single tool that does the job of a ragbag collection of whatnots.
Linux is the most important platform in the cloud. Its good to have a decent analysis tool for when things get weird.
I'm suprised that DTrace hasn't been ported to Linux a long time ago.
Maybe I'm wrong, but I thought that DTrace is one of many tools which Sun Micrososystems open sourced before they were acquired by Oracle - ZFS and Zones were both definitely open sourced.
Joyent, one of the main vendors of OpenSolaris, are promoting DTrace: https://www.joyent.com/dtrace
There is a great video on Yoochoob somewhere by a one of the many ex-Sun egngineers who went to Joyent, who reveals the history of Solaris and how it became the OpenSolaris that we know today.
Its been in Oracle Linux for years.
The problem is the license and Oracle's lawyers.
Strictly speaking it's not Oracle's lawyers that are the problem. It's GPL2 that has the problem accepting other licenses. Oracle (Sun) can license their code in whichever way they want to; it's up to the rest of us to respect that.
The problem with GPL is it is not supported because it is a good license, it is supported because it is the license of choice behind everyones favourite OS. GPL is a problem as it is not an open source license. It is a copyleft license and doesn't grant full free rights to the code.
I support open and free source code, so all the code I write and publish is released under an open source MIT license, rather than a copyleft license.
This means you can do what you want with it, you don't have to restrict yourself to using only certain other 'correctly' licensed code, it is true free and open source. The only restriction with MIT is that somewhere you have to include a copy of the MIT license, that's it. You can use it commercially without problem, or you can just repackage the code and sell it directly, there are no restrictions.
Try doing any of that with GPL code, and survive the uproar that would go around it.
Interestingly, Microsoft released all it's open source code (.Net Core etc) under the MIT license, rather than the much more restrictive GPL.
yeah, that "license incompatibility" thing and "purity" for distribution _WITH_ the kernel, like some kind of religious requirement or something.
This, of course, never means that someone couldn't create a package that you could install on a particular Linux distro, or even include it with a distro as a "non-Free" package [like Debian derivatives might be inclined to do, although it appears they don't].
Nor does it stop anyone from doing a 'build from source' install.
This sounds an awful lot like the "problems" associated with ZFS on Linux. Has _THAT_ even been fully resolved?
Linux' licensing "requirements" have bothered me for a while. It's one of many reasons I use FreeBSD for my primary desktop and server OS, rather than Linux.
Still, a welcome change from Oracle with the GPL license. As for me, I like "dual licensing" things instead. That way you can ship your derived version as GPL, _OR_ "some other license" (say MIT or BSD license), your choice. This avoids having to 'cave' to the GPL way of thinking, while simultaneously allowing it. Yeah, options. Freedom IS a good thing!
Less religion/purity.
More a case of having a simple 'Everything in the kernel is GPL' thats its.
It makes auditing and any legal issues a no brainer.
In my time, Ive had to deal with products containing many different licenses. They make life very hard.
The only hing worse than dealing with multiple license is dealing with 3rd party code that has been submited without any form of certification.
"Still, a welcome change from Oracle with the GPL license."
^ I fully agree with this and the question then becomes: Is this just a one off or is there more to follow? Open source licensing means that Oracle could itself benefit from the improvement work by external developers.
Come on Oracle, please, relicense to GPL, you know you want to.
Please Oracle, DO NOT license under the GPL - the GPL is a legal nightmare. I suppose dual-licensing would work, but why should they appease the GPL which is the thing that is creating the barriers to usr in the first place?
Maybe that's because Oracle doesn't like someone to take unfair advantage of their code ? Microsoft and Apple have built empires using pieces of BSD licensed code. Did they ever send back a thank you postcard to those developers who created the software ?
Contrarily to your statement, GPL does not create barriers for users. Only for greedy developers trying to make money by cheating other developers.
Even a brain dead monkey knows by now that you can use GPL code as you please. You may even give it away under the same conditions in place when you received it. If you care for your work and want to be paid, write your code from scratch (i.e. don't rip off) and license it under a proprietary license (or BSD if you feel generous).
On the one you hand you claim it is free open source software, on the other, you then tell me precisely what I can and can't do with the code. Not exactly free and open source is it?
If you were really a fan of open source you would use the MIT license, after all, the point of open source software development is not to get paid for it, but to freely release software for others to use.
What anyone else does with free software should be none of the developers concern, after all, they released it as free open source software, except when they don't, and license it under a restrictive copyleft license like GPL.
I always use MIT for any published source code, never GPL, but then I believe if a piece of code is claimed to be free and open source, it really should be.
"the knowledge that Linux is truly the One True OS"
Yeah, yeah, and the Book of Linux was translated from assembly language written on golden disk platters using a pair of GNU tools inside Linus' favorite hat. And, because of the burning in our monitors, we must _BELIEVE_ !!! [and when I look into that burning image I can see the image of J.R. Dobbs]
(I couldn't pass up THIS opportunity!)
Well, It is turning out to be the case...
You get Linux on virtually anything these days, and it works great, and keeps getting better and better, it has its issues, and like everything in life it is not perfect, but surely it is already got to the point where as an engineering tool is an absolute killer app...
That one liner from Bryan Cantrill was a belter.
He quoted 251 lines of technical content from David S. Miller and added one line:
Have you ever kissed a girl?
To which Miller responded:
No, but I can kiss the sky:$Id: bw_tcp.c,v 1.3 1995/06/21 21:02:49 lm Exp $
Socket bandwidth using xenophanes: 10.49 MB/sec
And I know how to make it even faster.
Just remember, if ZFS and Dtrace were just so totally awesome and amazing, Sun Microsystems and/or an Oracle run Sun set of stuff would dominate today. It's nice and all, but maybe not as nice as some want us to believe.
I remember working for a company with a very mature software product (existed years before Linux), but its performance on Solaris sucked vs. Linux. So Sun sent their best expert to our site for month armed with the mighty Dtrace to find out where "our" flaw was. Needless to say, after a month he left with head down sobbing.
It was fun and sad to watch.
Still, interesting to see Oracle GPL something... but usually means they're abandoning ship... I know Larry isn't a nice guy and maybe he isn't in full control, but still....
Just remember, if ZFS and Dtrace were just so totally awesome and amazing, Sun Microsystems and/or an Oracle run Sun set of stuff would dominate today. It's nice and all, but maybe not as nice as some want us to believe.
Yes yes, and how did you Linux guys react when Windows people said the same thing about Linux?
It's a totally stupid claim in either case.
Yeah but ....
Solaris does run 10-20% slower than Linux.
There is a reason for that.
Solaris has to operate in environment where if the kernel crashes customer want to know why.
There's a lot of checkpointing going on as a system.
And Solaris has STREAMs networking thats a bad idea.
Both dtrace and ZFS sit apart from Solaris. Well, zfs sits on top of the BSD disk block.
As a large, enterprise file system ZFS is the dogs bollocks. It really is. The pin, the gnashing of teeth all goes away. Sure, you have to throw RAM at it but DRAM is cheaper than my time.
Dtrace is great. Its a proper find anything tools. Nothgn comes close.
The lack of uptake is due to zfs and trace coming out just as Sun fill to bits.
This is just silly. Anyone that wanted to use dtrace or zfs was building the packages as modules and adding it themselves.
I'm glad this makes the communists feel better. However, me thinks Oracle has an ulterior motive. Ten bucks says they're going to kill off the Solaris project all together and replace it with Oracle Linux and needed this license change to avoid sticky-ness with the GNU project.
Oracle appears to have open-sourced DTrace, the system instrumentation tool that Sun Microsystems created in the early 2000s and which has been beloved of many-a-sysadmin ever since.
It's been open-source for ages, under the CDDL.
CDDL is an open-source license.
Open-Source != GPL
Ahhh it would be so good to have SMF instead crappy systemd under Linux ..
Fireworks, FMA, ILB, ipfilter, projects, trusted boot with signed ELF binaries, zones, kernel zones and many other things ..
I want IPS instead crappy rpm or dpkg. I want BEs instead crappy os-tree, I want .. this list is really long :-/
But moment all this already is available for free in OpenIndiana!!! so why do we want to have this under Linux if core developers ignoring wishes of normal users don't want this?
DTrace source code is available more than the decade. In meantime on Linux failed LTT, LTTng, sytemtap and few less known wannabe-DTrace Linux projects. Now it is even more such projects and looks like every CS student is trying to write MyOwnDtrace(tm).
Why does anyone need to care about Linux offering technologies like DTrace or ZFS if Linux developers community don't care?
FreeBSD, OpenIndiana are still growing. It took few years kind of critical mass of the developers to fully understand OpenSolaris code to be able to conserve it and start introducing new features.
Sun then Oracle did not understand this delay and now Oracle is losing momentum to have the active support of independent developers from outside Oracle.
It is still not too late for Oracle ..
Oracle instead trying help crappy Linux should focus on building Open Source community around own products. They started publishing own code in git repos but those repos have disabled submission of pull requests =:-o
Most of the System Engineers knowing the real value of some technologies are still between hammer and anvil. Linux hammer of the core developers who don't want o understand things which they do not write, and Oracle anvil which still does not understand long-term real potential of Open Source community.
At the moment still, amongst those two, I don't see the potential winner. I see only customers and normal users losers ..