On 4/18/09 3:47 PM, "Kirk Wolf" <[email protected]> wrote: > I will defer to your judgment that having any kind of Sun JRE is > better than an IBM JRE, from a "selling" point of view.
Well, not just mine. Part of IBM's investigation into participating in the OpenSolaris project was whether the IBM JDK was sufficient, or whether people really wanted the Sun JDK. They did a CIO survey. The response was overwhelmingly that they wanted the Sun JDK, both on Linux and (especially) on OpenSolaris. > - Port/write a JIT for the Sun JRE for z architecture. A good JIT, > especially for z, is really a very sophisticated real-time optimizing > compiler; a rather large effort. Not as hard as it used to be. The Zero-assembler project has eliminated a lot of really icky stuff in the JDK implementation. But, yes. Non-trivial effort. Fortunately, I have really smart people working on it. 8-) > Most of > the portability issues have to do with using non-API classes, which > might not be in OpenJDK either. But if this were done, it would be > easy to port it to Linux, which would be great. If you look closely at the sample output I posted, that's actually on Linux. One of the reasons we kept the Linux ABI in OpenSolaris was to make things more easily portable. 8-) > If you could pull > this off so that the performance were comparable to the IBM JVM, there > is a team of IBM JIT developers that would certainly get excited. > This kind of competition would be great. Actually, it'll just add Hursley to the list of death threat origins from various divisions of IBM...8-) I don't want to compete with them, I want the whole issue to GO AWAY so that this whole Java thing can settle down to being something other than a gigantic PITA. If they (IBM and/or Sun) would share their work, I'd assemble it into the Grand Unified Java Package and then we wouldn't have to argue about it any more. Unfortunately, some parts of Sun and IBM don't seem to want that outcome to happen. They have their reasons. > - (Assuming that IBM ported their JVM from Linux z to Solaris z) Why > not explore the possibility of running the Sun OpenJDK Java libraries > with the IBM JVM? I'm not sure how difficult this would be, but it > seems possible in theory. Alternatively, IBM somehow could add > non-public Sun classes to their JDK so that compatibility issues are > relieved. If. If is good. I'd put that question directly to IBM and tell them how badly you want it to happen. You might get an answer. Or not. > Again, customers whose applications only support the Sun JRE are > saying that they don't conform to the public API. They will probably > get whacked when they try to migrate to a new version of Java. These > are likely the same goof-balls that complain how non-portable Java is, > and how slow it is since it is "interpreted". Do these people really > think that porting their Solaris workloads from Sparc or Intel to z > will be easy, so long as they don't have to support a different JRE? Again, not my battle. You have the checkbox decision makers that do evaluations based on presence or absence of specific things. Check, you win. No check, you fail. This is one of those checkbox items. Ask the rest of the list how often they get the whine that "we need the Sun JDK to run application PDQ?" I agree that apps that require a specific JDK are probably going to lose at some point, but that's not my problem at the moment. If IBM or Sun wants to engage with us, they are welcome to do so. There are at least a dozen large System z deals that I'm aware of that IBM will lose if this does not happen in a timely manner. But, again, just like with Linux on Z itself, they can be part of the problem or part of the solution. It's up to them. -- db ---------------------------------------------------------------------- For LINUX-390 subscribe / signoff / archive access instructions, send email to [email protected] with the message: INFO LINUX-390 or visit http://www.marist.edu/htbin/wlvindex?LINUX-390
