I really don't think it is, as I think scalability depends on what you are trying to do. For example, I was dealing with both distributed systems and vector architectures in the 80s and 90s - neither was generally 'scalable', unless you stated what the specific problem was (I came across many examples of developers trying to throw all sorts of problems at one or other of these architectures and failing badly in many cases). The same applies, I think to systems which are highly non-uniform in memory architecture. They can be a waste unless you code for them correctly. Something like a well-written Fortran program may be able to great performance, but it may be pointless to try and run a high-load web server.
Their smaller Altix systems (still larger than the largest Sun system) are also used very much for IO, networking, etc. and not just userspace computation
Are they? From past experience, and looking at websites now, SGI has always been a provider of primarily technical and numeric workstations and machines - things like video and imaging computation and chemical/biochemical modelling. These tend to be things that NUMA would be appropriate for, as the problems can be divided up into localised spaces. On the contrary, Sun has tended to be used for commerce, with less predictable multi-threaded work, where NUMA is likely to be a lot less effective.
That has absolutely no bearing on anything. DOS is more mature than Linux ferchistsake
What I was implying was that these systems were more mature in terms of tuning for multiprocessing, and I think I put that reasonably clearly.
There is no way Solaris could keep up with Linux on a fancy, modern interconnect like SGI's NUMAflex.
I don't see why not, as Solaris is just an operating system - your point is about hardware, and Solaris can run on Intel and AMD, and presumably make use of NUMAflex.
Well are you skeptical about Linux, or just skeptical about an SSI that large?
I'm skeptical about any system trying to do this. I think its only going to work with very careful coding. My feeling is that this isn't really 'scalability', its just giving specialised developers in highly specific situation the opportunity to use more CPU power. (As someone interested in numeric work, I think this is a good thing).
I mean, is there any reason to think Linux will perform significantly worse than IRIX in this regard? (I say "significantly" because of course IRIX should perform a bit better just by virtue of being more mature,
I would argue that IRIX (and Solaris) should perform and scale better on multiple CPUs because they are very much mature in this respect than Linux. Linux will catch up in time, but in spite of all the good work that has been put into the kernel, I just can't believe it can yet match the multiprocessing efficiency of some proprietary kernels that have been highly tuned for this for many years.
Thanks for all the info - I had forgotten the SGI/Cray connection. However, I'm still not sure I see how it could be said that Sun purchased a NUMA-type architecture off SGI - it looks like they purchased the SMP technology?
The SGI is definitely NUMA, the Sun is either SMP or "slightly NUMA"
I think this is my point, which is that the Sun (or any) SMP system is more symmetric, and so probably provides more 'scalable' power for general purpose use?
Anyone got any suggestions how a positronic brain works exactly? The Star Trek TNG writers gave Data one too,
The Star Trek writers would probably give the same response as when they were asked how the Heisenberg Compensators (used to prevent quantum effects distrupting transportation) worked:
Assuming the positronic brain has a few billion of these,
Surely its likely to be a lot more. I mean, our brains have at least 100 billion cells, each of which can have 10,000 synapses. If positrons take the place of electrons, you may have 10s of them for each synapse (at least!).
So, its 100 x 10,000 x 10 = 10 million times worse than you calculate....
My opinion is that Linux on a 1024-way is a spectacularly stupid idea, introduced more for the sexiness of having a 1024-way machine than for any practical benefits.
I think there are practical benefits, but only for very specialised applications. As far as I understand it, Linux only really works highly multiprocessor on architectures that aren't really symmetrical (like this SGI machine). Different CPUs don't have the same speed of access to different memory areas as in true SMP. So, it's not comparable to true SMP as in systems from Sun and IBM.
but I was always impressed with the NUMAFLEX archetecture....
I was under the impression that NUMA-type architectures were actually cheap compromises, providing massive multiprocessing for systems that could not really cope with fully symmetric CPU access to memory. Have things changed, or am I wrong? I'd be interested to know, as I have not done any development on SMP-type machines for years.
Well, I hope it's not getting us closer to 'I, Robot'. Remember, the Azimov robots used positronic brains. Even a simple headache could result in the destruction of a city block.
The UNIX made by SGI (the company making the machine referenced in the article) is more scalable than Solaris. Remember, IRIX was the first OS to scale a single Unix OS image across 512 CPUs. And now they've eclipsed that, with Linux.
Scalability is a complex issue. SGI has put a whole lot of processors together and put a single Linux image on it (so that a single program can use all memory), but this says nothing about how that setup will actually perform for general purpose use. Just because the hardware allows threads on hundreds of processors to make calls into a single Linux kernel, does not mean that there will not be major performance issues if this actually happens.
There are performance issues with memory even on single processor systems with nominally a single large address space, and a developer may need to put a lot of work into ensuring that data is arranged to make best use of the various levels of cache.
Many of the multi-processor architectures require even greater care to ensure that the processors are actually used effectively.
The fact that a single Linux image has been attached to hundreds of processors is no indication of scalability. A certain program may scale well, or not.
Gov, people dont want to have their documents open correctly 95% of the time. they want 100%.
This is why I recommend users install Star/Open Office in addition to MS Office. Several times the following has happened: User tries to open MS Office document in MS Office. MS Office application refuses to open document or crashes. User opens same document in Star/Open Office. Document opens fine. User saves document in MS Office format. User can then open document in MS Office.
It is rumored that these cars will be able to run on both diesel and unleaded fuels, but will not be able to go over ~30MPH.
A strange thing about some geeks/nerds. They are are supposed to be a class of people who are interested in the latest developments and have an in-depth understanding of technology, yet persist in spreading FUD that has not been true for 5 years.
Java is not slow. Well, some Java is. Some C is slow. Some java GUI stuff is slow. Some C GUI stuff is slow. But, on the whole, Java is not slow. Get up to date!
Unfortunately for Sun, they're not innovators and there are no current trends directly in their area for them to latch on to.
Sun have always been innovators. They were the main drivers for Unix Workstations in the 1980s, they pioneered GUIs for Unix. They helped push 'open systems' in which different OS providers wrote to common APIs. They pioneered donating APIs to the community to assist with market growth (NFS is a good example). They were one of the first users of RISC. They helped make binary portability a practical reality with Java (it was not the first such system, but the first designed for the mass market). Sun is helping to establish new markets and technologies. Turn on any TV channel and you will soon hear an advert to download Java games and programs to your phone or PDA. This market is huge, and Sun gets paid for consultancy services by the companies that develop the phones. This year major car builders are working with Sun to enable the use of Java as a real-time embedded control system. Sun get paid for this work.
Java is not going to make it any money as a product
Java is making Sun millions, through licencing, and as a mechanism for enabling Sun's recent sigificant growth as a software consultancy company.
Sun may have many flaws, but lack of innovation has never been one of them.
It's just circling in the middle of no where
Not at all. The company is changing, with a major and increasing part of its revenue and direction being nothing to do with hardware.
Oh, I think PHP has relegated Perl to the scrap heap. In absolute numbers, Perl usage may still have grown
Sorry, I don't understand as these two statements are mutually contradictory.
Yup, but they are targeting exactly that subset that people need to move their Windows-based ASP.NET applications to Linux.
Its the ASP.NET/Windows Forms/Web Services part of Mono is that is incomplete.
Well, if you actually believe that Solaris has much of a future, then that may be a reasonable argument to make...
No, this was not the point I was making. Read the post - it was about relevant comparisons.
Well, that isn't all that surprising, since other web development platforms often don't have a need for complex "clustered application servers"--they scale without it.
No they don't. If you understood the technology you would realise that the clustering is usually nothing to do with the web, its to do with the databases and other services. To handle these the application servers need to be clustered and services need to be able to dynamically located using facilities such as Naming, and interaction with the services (and not simply the database) has to be transactional and scalable. To say that other systems 'scale without it' is just silly - these clustering systems aren't set up just for fun - they are an important part of commercial high-volume systems.
Yeah, sure. Is Java running the real-time systems in my car?
Yes. This year there has been a big investment in real-time automobile systems using Java. The largest car makers are working on a common standard for real-time Java for this purpose,
Is it in my router?
Yes. For an example see the recently released Possio PX30.
My webcam?
Yes. Cat Eye Technogies is an example.
My digital camera?
Yes. Java VMs such as JLite and Chai (from HP) are used widely for this.
It isn't.
Sorry - you are wrong - see above.
Sun keeps grinding out one set of APIs after another, and people may try them out for a year or two and say that they are "using Java", but that is not the same as actually spreading into a market.
Well, it is, if by "using Java" these "people" (mostly "large companies") are really using Java, and they are, in their millions.
Last year there were an estimated 3 million Java developers. This year its over 4 million.
Look : you may not actually like Java, for whatever reason, but not facing up to the facts (facts that are easily available), won't actually make Java disappear.
Please tell me: where can I find that "free, certified complete and compatible, open source implementation of J2EE"?
Am I missing something, but why would anyone be reguralrly doing such long compiles? I remember C++ development in the mid-80s when long compiles and links were common, but that was in the days of 20 MHz 386s. Even then, we soon moved on to the use of incremental compilers - only the relevant (and small) parts of even a large project would be compiled, and then linked back into the pre-compiled code. I can understand why the first compilation of a large project could take some time, but from then on, shouldn't compilation and linking be a minor matter?
Its certainly true that they can be used for GUI development, however Java hardly is ever used for this (its one of Java's failures). Java is used primarily for server-side middleware - interfacing with databases. C# and Python have not impacted this market at all (are there any C# or Python CORBA server components?
Perl has relegated awk to the scrap heap. And Python has pretty much killed Perl's aspiration in several areas (GUI development, Matlab replacement, etc.). PHP is probably far more common than Perl for server-side development now. And all of them have taken away a lot of "market share" from C.
This are are interesting points, but these are not the answers to the questions I asked.
Has Python 'relegated Perl to the scrap heap'? No. Has PHP done this? No. Have they 'relegated C to the scrap heap?' No (show me a kernel driver written in PHP). Will they relegate Java 'to the scrap heap'? No.
ASP.NET is the biggest threat to Java: that's where server-side development is moving on Windows
I might just as well say that ASP.NET is doomed because server side development on Solaris is moving to Java! We are talking about different platforms... Microsoft is, and have always been, a minor player in the application server market. Microsoft servers tend to be used for specialist single-function situations - file servers, print servers or Exchange servers. They really aren't used that much as large-scale application servers. There is a HUGE difference between a server running a few (or even many) ASP pages and clustered application servers, where Java dominates. But, even where Windows is the system used for servers, guess what one of the main languages used for development is? Java! Why - partly because its portable: managers know they have the option to migrate those server to something else (such as Linux) with no recoding.
Mono's.NET implementation then gives those people the option to deploy on Linux when they come to their senses.
Mono's.Net is not.Net - its an incomplete subset, and on Mono's website they say that it always is likely to remain an incomplete subset. By contrast, there are quality open-source implementations of the totality of enterprise Java (JBoss), and unlike.Net, the implementations are actually supported by company that designed the language! Which do you think developers are going to go for - a subset of a Microsoft product with no guarantees of current or future compatibility, or free, certified complete and compatible, open source implementation of a system like J2EE that is supported by dozens of companies? Well, just look at the IT press and the job market and the answer is clear.
Languages don't usually die, but they can become less and less relevant. And that can even happen pretty quickly.
Absolutely. I guess when Java stops its phenomenal year-on-year growth, and its constant spread in to new markets and technology (this year, embedded and real-time applications), then this could happen.
I am glad, you are so blind people. Every few months, there is a news about "potential life in the Universe is more abundant... blah, blah, blah". Every such news just fades away.
Unlike the water, which you said did not exist, and has been there for millions of years.
Jesus. Just give me a version number so I can track what it's compatible with, and what features it has. If you're bumping up your version number for a product, bump them for all related ones as well, in the same increment. Don't make me try to figure out what version number of the language is supported by which version number of the developer's kit for god's sake. Is it so damn hard?
Obviously it isn't, because that is exactly what they have done. The Java version number and SDK number now match. Its Java 5.0 throughout: Java 5.0, SDK 5.0.
Slashdot Mirror
at least for their applications, they get good scalability.
But that is exactly my point - its for their applications. But would you get scalable performance for a Web server? C compiling? Doom III?
definitely not SMP by definition.
You may be right, but I'm not sure where that gets us as almost everyone else calls Sun's architecture SMP.
This is the very definition of scalability.
I really don't think it is, as I think scalability depends on what you are trying to do. For example, I was dealing with both distributed systems and vector architectures in the 80s and 90s - neither was generally 'scalable', unless you stated what the specific problem was (I came across many examples of developers trying to throw all sorts of problems at one or other of these architectures and failing badly in many cases). The same applies, I think to systems which are highly non-uniform in memory architecture. They can be a waste unless you code for them correctly. Something like a well-written Fortran program may be able to great performance, but it may be pointless to try and run a high-load web server.
Their smaller Altix systems (still larger than the largest Sun system) are also used very much for IO, networking, etc. and not just userspace computation
Are they? From past experience, and looking at websites now, SGI has always been a provider of primarily technical and numeric workstations and machines - things like video and imaging computation and chemical/biochemical modelling. These tend to be things that NUMA would be appropriate for, as the problems can be divided up into localised spaces. On the contrary, Sun has tended to be used for commerce, with less predictable multi-threaded work, where NUMA is likely to be a lot less effective.
That has absolutely no bearing on anything. DOS is more mature than Linux ferchistsake
What I was implying was that these systems were more mature in terms of tuning for multiprocessing, and I think I put that reasonably clearly.
There is no way Solaris could keep up with Linux on a fancy, modern interconnect like SGI's NUMAflex.
I don't see why not, as Solaris is just an operating system - your point is about hardware, and Solaris can run on Intel and AMD, and presumably make use of NUMAflex.
Well are you skeptical about Linux, or just skeptical about an SSI that large?
I'm skeptical about any system trying to do this. I think its only going to work with very careful coding. My feeling is that this isn't really 'scalability', its just giving specialised developers in highly specific situation the opportunity to use more CPU power. (As someone interested in numeric work, I think this is a good thing).
I mean, is there any reason to think Linux will perform significantly worse than IRIX in this regard? (I say "significantly" because of course IRIX should perform a bit better just by virtue of being more mature,
I would argue that IRIX (and Solaris) should perform and scale better on multiple CPUs because they are very much mature in this respect than Linux. Linux will catch up in time, but in spite of all the good work that has been put into the kernel, I just can't believe it can yet match the multiprocessing efficiency of some proprietary kernels that have been highly tuned for this for many years.
Thanks for all the info - I had forgotten the SGI/Cray connection. However, I'm still not sure I see how it could be said that Sun purchased a NUMA-type architecture off SGI - it looks like they purchased the SMP technology?
The SGI is definitely NUMA, the Sun is either SMP or "slightly NUMA"
I think this is my point, which is that the Sun (or any) SMP system is more symmetric, and so probably provides more 'scalable' power for general purpose use?
Anyone got any suggestions how a positronic brain works exactly? The Star Trek TNG writers gave Data one too,
The Star Trek writers would probably give the same response as when they were asked how the Heisenberg Compensators (used to prevent quantum effects distrupting transportation) worked:
'Very well thank you!'
Assuming the positronic brain has a few billion of these,
Surely its likely to be a lot more. I mean, our brains have at least 100 billion cells, each of which can have 10,000 synapses. If positrons take the place of electrons, you may have 10s of them for each synapse (at least!).
So, its 100 x 10,000 x 10 = 10 million times worse than you calculate....
My opinion is that Linux on a 1024-way is a spectacularly stupid idea, introduced more for the sexiness of having a 1024-way machine than for any practical benefits.
I think there are practical benefits, but only for very specialised applications. As far as I understand it, Linux only really works highly multiprocessor on architectures that aren't really symmetrical (like this SGI machine). Different CPUs don't have the same speed of access to different memory areas as in true SMP.
So, it's not comparable to true SMP as in systems from Sun and IBM.
but I was always impressed with the NUMAFLEX archetecture....
I was under the impression that NUMA-type architectures were actually cheap compromises, providing massive multiprocessing for systems that could not really cope with fully symmetric CPU access to memory. Have things changed, or am I wrong? I'd be interested to know, as I have not done any development on SMP-type machines for years.
SUN BOUGHT IT'S NUMA TECH FROM SGI IN THE FIRST PLACE.
Interesting. I thought Sun always used pure SMP, not NUMA. Do you have a URL to back up this statement?
Well, I hope it's not getting us closer to 'I, Robot'. Remember, the Azimov robots used positronic brains. Even a simple headache could result in the destruction of a city block.
The UNIX made by SGI (the company making the machine referenced in the article) is more scalable than Solaris. Remember, IRIX was the first OS to scale a single Unix OS image across 512 CPUs. And now they've eclipsed that, with Linux.
Scalability is a complex issue. SGI has put a whole lot of processors together and put a single Linux image on it (so that a single program can use all memory), but this says nothing about how that setup will actually perform for general purpose use. Just because the hardware allows threads on hundreds of processors to make calls into a single Linux kernel, does not mean that there will not be major performance issues if this actually happens.
There are performance issues with memory even on single processor systems with nominally a single large address space, and a developer may need to put a lot of work into ensuring that data is arranged to make best use of the various levels of cache.
Many of the multi-processor architectures require even greater care to ensure that the processors are actually used effectively.
The fact that a single Linux image has been attached to hundreds of processors is no indication of scalability. A certain program may scale well, or not.
Gov, people dont want to have their documents open correctly 95% of the time. they want 100%.
This is why I recommend users install Star/Open Office in addition to MS Office. Several times the following has happened: User tries to open MS Office document in MS Office. MS Office application refuses to open document or crashes. User opens same document in Star/Open Office. Document opens fine. User saves document in MS Office format. User can then open document in MS Office.
It is rumored that these cars will be able to run on both diesel and unleaded fuels, but will not be able to go over ~30MPH.
A strange thing about some geeks/nerds. They are are supposed to be a class of people who are interested in the latest developments and have an in-depth understanding of technology, yet persist in spreading FUD that has not been true for 5 years.
Java is not slow. Well, some Java is. Some C is slow. Some java GUI stuff is slow. Some C GUI stuff is slow. But, on the whole, Java is not slow. Get up to date!
Unfortunately for Sun, they're not innovators and there are no current trends directly in their area for them to latch on to.
Sun have always been innovators. They were the main drivers for Unix Workstations in the 1980s, they pioneered GUIs for Unix. They helped push 'open systems' in which different OS providers wrote to common APIs. They pioneered donating APIs to the community to assist with market growth (NFS is a good example). They were one of the first users of RISC. They helped make binary portability a practical reality with Java (it was not the first such system, but the first designed for the mass market). Sun is helping to establish new markets and technologies. Turn on any TV channel and you will soon hear an advert to download Java games and programs to your phone or PDA. This market is huge, and Sun gets paid for consultancy services by the companies that develop the phones. This year major car builders are working with Sun to enable the use of Java
as a real-time embedded control system. Sun get paid for this work.
Java is not going to make it any money as a product
Java is making Sun millions, through licencing, and as a mechanism for enabling Sun's recent sigificant growth as a software consultancy company.
Sun may have many flaws, but lack of innovation has never been one of them.
It's just circling in the middle of no where
Not at all. The company is changing, with a major and increasing part of its revenue and direction being nothing to do with hardware.
Oh, I think PHP has relegated Perl to the scrap heap. In absolute numbers, Perl usage may still have grown
Sorry, I don't understand as these two statements are mutually contradictory.
Yup, but they are targeting exactly that subset that people need to move their Windows-based ASP.NET applications to Linux.
Its the ASP.NET/Windows Forms/Web Services part of Mono is that is incomplete.
Well, if you actually believe that Solaris has much of a future, then that may be a reasonable argument to make...
No, this was not the point I was making. Read the post - it was about relevant comparisons.
Well, that isn't all that surprising, since other web development platforms often don't have a need for complex "clustered application servers"--they scale without it.
No they don't. If you understood the technology you would realise that the clustering is usually nothing to do with the web, its to do with the databases and other services. To handle these the application servers need to be clustered and services need to be able to dynamically located using facilities such as Naming, and interaction with the services (and not simply the database) has to be transactional and scalable. To say that other systems 'scale without it' is just silly - these clustering systems aren't set up just for fun - they are an important part of commercial high-volume systems.
Yeah, sure. Is Java running the real-time systems in my car?
Yes. This year there has been a big investment in real-time automobile systems using Java. The largest car makers are working on a common standard for real-time Java for this purpose,
Is it in my router?
Yes. For an example see the recently released Possio PX30.
My webcam?
Yes. Cat Eye Technogies is an example.
My digital camera?
Yes. Java VMs such as JLite and Chai (from HP) are used widely for this.
It isn't.
Sorry - you are wrong - see above.
Sun keeps grinding out one set of APIs after another, and people may try them out for a year or two and say that they are "using Java", but that is not the same as actually spreading into a market.
Well, it is, if by "using Java" these "people" (mostly "large companies") are really using Java, and they are, in their millions.
Last year there were an estimated 3 million Java developers. This year its over 4 million.
Look : you may not actually like Java, for whatever reason, but not facing up to the facts (facts that are easily available), won't actually make Java disappear.
Please tell me: where can I find that "free, certified complete and compatible, open source implementation of J2EE"?
JBoss and Tomcat.
Today, Calc's internal computations are done with infinite precision for basic operations (addition, subtraction, multiplication, division)
....
Neat. So, Pi + 1 =
Well distcc can greatly reduce the time.
I see what you mean, but my point remain valid, I think: this is a one-off compile. After that.. what is the demand?
Am I missing something, but why would anyone be reguralrly doing such long compiles? I remember C++ development in the mid-80s when long compiles and links were common, but that was in the days of 20 MHz 386s. Even then, we soon moved on to the use of incremental compilers - only the relevant (and small) parts of even a large project would be compiled, and then linked back into the pre-compiled code. I can understand why the first compilation of a large project could take some time, but from then on, shouldn't compilation and linking be a minor matter?
Nope, It's verion 5.0 of the "Java 2 Standard Edition.
You are right. I was wrong. I thought I understood what was going on, but now I'm getting a headache just thinking about it.
GUI development: C#/.NET (Windows), C#/Gtk# (Linux), Python/Gtk (Linux), Python/wxWindows (cross-platform)
.NET implementation then gives those people the option to deploy on Linux when they come to their senses.
.Net is not .Net - its an incomplete subset, and on Mono's website they say that it always is likely to remain an incomplete subset. By contrast, there are quality open-source implementations of the totality of enterprise Java (JBoss), and unlike .Net, the implementations are actually supported by company that designed the language! Which do you think developers are going to go for - a subset of a Microsoft product with no guarantees of current or future compatibility, or free, certified complete and compatible, open source implementation of a system like J2EE that is supported by dozens of companies? Well, just look at the IT press and the job market and the answer is clear.
Its certainly true that they can be used for GUI development, however Java hardly is ever used for this (its one of Java's failures). Java is used primarily for server-side middleware - interfacing with databases. C# and Python have not impacted this market at all (are there any C# or Python CORBA server components?
Perl has relegated awk to the scrap heap. And Python has pretty much killed Perl's aspiration in several areas (GUI development, Matlab replacement, etc.). PHP is probably far more common than Perl for server-side development now. And all of them have taken away a lot of "market share" from C.
This are are interesting points, but these are not the answers to the questions I asked.
Has Python 'relegated Perl to the scrap heap'? No. Has PHP done this? No. Have they 'relegated C to the scrap heap?' No (show me a kernel driver written in PHP). Will they relegate Java 'to the scrap heap'? No.
ASP.NET is the biggest threat to Java: that's where server-side development is moving on Windows
I might just as well say that ASP.NET is doomed because server side development on Solaris is moving to Java! We are talking about different platforms... Microsoft is, and have always been, a minor player in the application server market. Microsoft servers tend to be used for specialist single-function situations - file servers, print servers or Exchange servers. They really aren't used that much as large-scale application servers. There is a HUGE difference between a server running a few (or even many) ASP pages and clustered application servers, where Java dominates. But, even where Windows is the system used for servers, guess what one of the main languages used for development is? Java! Why - partly because its portable: managers know they have the option to migrate those server to something else (such as Linux) with no recoding.
Mono's
Mono's
Languages don't usually die, but they can become less and less relevant. And that can even happen pretty quickly.
Absolutely. I guess when Java stops its phenomenal year-on-year growth, and its constant spread in to new markets and technology (this year, embedded and real-time applications), then this could happen.
I am glad, you are so blind people. Every few months, there is a news about "potential life in the Universe is more abundant ... blah, blah, blah". Every such news just fades away.
Unlike the water, which you said did not exist, and has been there for millions of years.
You are right, I was wrong. I was basing what I said on what I had been told at the JavaOne conference. I should have RTFA.
What's the 2 doing in there if it's just a simple versioning scheme?
Proving me wrong and getting me confused!
Jesus. Just give me a version number so I can track what it's compatible with, and what features it has. If you're bumping up your version number for a product, bump them for all related ones as well, in the same increment. Don't make me try to figure out what version number of the language is supported by which version number of the developer's kit for god's sake. Is it so damn hard?
Obviously it isn't, because that is exactly what they have done. The Java version number and SDK number now match. Its Java 5.0 throughout: Java 5.0, SDK 5.0.