Oracle's port of DTrace to Linux is not DTrace(dtrace.org)
dtrace.org
Oracle's port of DTrace to Linux is not DTrace
http://dtrace.org/blogs/ahl/2011/10/10/oel-this-is-not-dtrace/
1 comments
When a company the size of Oracle announces something at their premiere conference, it's reliably not something they thought of for the first time that morning. If Ed Screven was a developer, hacking up DTrace for OEL, and writing on his blog and not the Chief Corporate Architect, reporting to Oracle's CEO, Larry Ellison, speaking at a keynote then I would totally agree with you.
Apple's WWDC in 2006 they announced DTrace. They handed it out; it worked. It had rough edges, but it was obviously worthy of release.
Paul Fox started hacking up DTrace for Linux in 2008. It's not perfect, and there's work to be done, but he's making progress and representing what he has appropriately.
Here's what Oracle's Linux blog had to say earlier today: http://blogs.oracle.com/linux/entry/looking_back_at_oracle_o... Another interesting feature we announced and is already available to Oracle Linux customers is "DTrace." We released a preview of DTrace which is made available to Oracle Linux support subscribers. Wim is also on a roll as you can see from his detailed blog entry on DTrace!
Apple's WWDC in 2006 they announced DTrace. They handed it out; it worked. It had rough edges, but it was obviously worthy of release.
Paul Fox started hacking up DTrace for Linux in 2008. It's not perfect, and there's work to be done, but he's making progress and representing what he has appropriately.
Here's what Oracle's Linux blog had to say earlier today: http://blogs.oracle.com/linux/entry/looking_back_at_oracle_o... Another interesting feature we announced and is already available to Oracle Linux customers is "DTrace." We released a preview of DTrace which is made available to Oracle Linux support subscribers. Wim is also on a roll as you can see from his detailed blog entry on DTrace!
You're missing an essential bit here: the use of DTrace damaged the system. This violates the Hippocratic promise of DTrace: to do no harm.[1] This transgression should not be written off as a youthful indiscretion of the port, for it reflects that those undertaking it do not understand this most essential of constraints. And to be honest, I'm personally astounded that a port with such modest ambition could be so busted: if you can't get the syscall provider right, FBT, SDT, USDT and so on will remain firmly out of reach.
[1] http://dtrace.org/blogs/bmc/2005/07/19/dtrace-safety/
[1] http://dtrace.org/blogs/bmc/2005/07/19/dtrace-safety/
There's a difference between an implementation which clearly violates the intended design and a tricky bug that is breaking something. Without knowing exactly why it is broken, it just seems to be a missed first opportunity for them to wow everyone. I'd be more concerned if it stayed broken more than a short period of time now that it has been publicly pointed out to them. It would be helpful if the Oracle guys posted the most efficient way for users to funnel them bug reports.
... or if they posted the source code...
Yesterday (October 4, 2011) Oracle made the surprising announcement that they would be porting some key Solaris features, DTrace and Zones, to Oracle Enterprise Linux.
Key word: would be.
It's been a week since they announced they have decided to port it, its' probably been in development for 2 weeks, and the author is upset it's not mature. Hmm...