JDK 21 released

Posted Sep 20, 2023 12:45 UTC (Wed) by paulj (subscriber, #341)
In reply to: JDK 21 released by vasvir
Parent article: JDK 21 released

It is funny indeed. I'm /very/ sure there were many debates inside Sun Microsystems even on the best threading model for Solaris, motivated in great part by the JVM. I think Solaris even did have an M:N thread model for a while, with M threads in libc mapped onto N OS threads. It was complex and problematic, and eventually removed, cause it was just simpler and more efficient (and less buggy!) to just have /1/ thread management system (scheduler, etc.) in the OS.

JDK 21 released

Posted Sep 20, 2023 14:13 UTC (Wed) by nim-nim (subscriber, #34454) [Link] (1 responses)

> I'm /very/ sure there were many debates inside Sun Microsystems even
> on the best threading model for Solaris, motivated in great part by the JVM.

I’m quite sure there were many such debates, and I’m quite sure they were not motivated by the JVM.

There was a lot of bad blood and turf wars between Solaris engineers and Java engineers at SUN, that explain why “write once run everywhere” quickly degenerated into “pathologically unable to integrate with any system service, especially Unix-side”.

There is no way in hell Solaris engineers would have let Java engineers define the way the Solaris threading model should look like, and there is no way in hell Java engineers would have felt constrained by the choices Solaris made.

From the Java side of the equation Solaris was a more and more marginalised OS and they had no intention of sinking with it, placing their bets on Windows and then (with a little IBM/BSD prodding) on Linux. Plus there was the great hope of eventually rewriting everything, OS included, the right way (in Java).

From the Solaris side of the equation Java was this thing that made so many wrong choices (from an OS and performance point of view) and dared dictate how the OS should behave. Plus they were traitors that helped move software to competitor OSes.

And SUN management failed to articulate a vision that would have made those bright minds achieve success toguether.

JDK 21 released

Posted Sep 20, 2023 14:21 UTC (Wed) by paulj (subscriber, #341) [Link]

Solaris engineers I'm sure wouldn't have let JVM decide the Solaris threading model. That's not at all at odds with Solaris engineers taking into account how /they/ thought users like JVM could be best served by the threading support in Solaris.

Looking it up, the M:N threading was in the original thread library (Solaris 2.5 or 2.6?) Which was deprecated by Solaris 9 odd by the T2 1:1 library, and removed for Solaris 10.