Slashdot Mirror

You can still construct a program with a memory leak in Java.

Not in the way you propose.

Java currently uses a Concurrent Mark-and-Sweep garbage collector, whereby it can (conceptually) traverse the object tree from one or more root objects (in your example, Main), with any objects that are reachable being marked for preservation. In your example, both A and B would be left unmarked, as you can't get to them from Main any longer.

Once the marking phase is completed, the set of unmarked objects is garbage collected. In your example, both A and B would be collected.

This is a serious over-simplification of how the entire system works; for further information I'd suggest reading Memory Management in the Java HotSpot Virtual Machine.

There are ways to leak memory in Java, but what you described isn't one of them.

Yaz.

I think AC meant something like http://www.sun.com/storage/flash/module.jsp.

I do not know how flash is done in laptops now, the best ones cannot be far from the CPU.

Yes http://www.sun.com/storage/flash/module.jsp.
It does not use SATA, but something JEDEC will(?) standardise.

I think it is very interesting idea, whether it will take off, I do not know.

That would be something I'd like to see.

I bash Java, but when I need to knock up things quickly, it is the first language I turn to. Usual old "right tool for the job" argument. A direct boot JVM would enable quite a few more classes of jobs.

Actually, now I think about it, wasn't there this?. Have you ever tried it? I haven't. Although I don't think that was the one I was thinking of.

Just firing random thoughts.

All Java code runs under a security manager.
Wrong: see http://java.sun.com/javase/6/docs/technotes/guides/security/overview/jsoverview.html

"Java applets and Javaâ Web Start applications are automatically run with a SecurityManager installed. However, local applications executed via the java command are by default not run with a SecurityManager installed. In order to run local applications with a SecurityManager, either the application itself must programmatically set one via the setSecurityManager method (in the java.lang.System class), or java must be invoked with a -Djava.security.manager argument on the commandline."

Default location of newer JRE releases is
C:\Program Files\Java\jre6

As for native look and feel, Swing does have several Windows look and feels in which widgets have native looks:
http://java.sun.com/docs/books/tutorial/uiswing/lookandfeel/plaf.html#available

Otherwise, I agree with you that Java desktop app are hard to distribute and necessary hurdle and that Sun is directly responsible for. It seems that Java desktop app platform was never given full attention.

... won't businesses wait for Windows 7 SP1 anyway?

That said, every geek worth his salt (let alone any actual IT professional) should take advantage of the fact that MS will let you download and run the Release Candidate Customer Preview of Vista 7 Ultimate for free for a year. Works just fine in VirtualBox (also free, for Win, Lin, and OS X) as described here. Even if you hate MS for whatever reason, it's still worth knowing what they're doing, especially if you can do so for free on whatever platform you're (probably) currently using.

But how much of those libraries exist to achieve Java's religious beliefs on abstraction?

Wow, how did this get modded insightful? For one, calling the design of a programming language a "religious belief", then asking a vague question about it without providing even a basis of an answer is just inflammatory.

But the answer that anybody who knows what they're talking about will tell you is, none of them. Java's abstraction mechanisms are built into the language. None of the standard libraries are necessary to support it. They take advantage of it, of course, and you'd be crazy to not take advantage of one of the language's features. Try taking a look at a tree representation of all of the classes in the standard library. The vast majority of classes are not more than one or two levels down from the top-level Object. The things that are deeper are typically things that are complex in any language -- CORBA, GUI toolkits, etc. It certainly looks much cleaner than many graphs I've seen of C++ libraries that abused multiple inheritence.

Waiting for similar hardware to become available for other languages.

I think Tim Bray is on the right track with his widefinder idea.

See Widefinder 1 and Widefinder 2 for details.

But would you buy from them again, that's the thing?

Obviously not, since they stopped making graphics workstations a long time ago. Like Sun, they've long since conceded that market to commodity PCs and specialized in servers and HPC clusters.

Correction: Sun does still have a couple of workstations left. But sadly, they're Intel-based.

Just get one of these.

Maybe not directly relevant to the OP's question but since I see a bunch of folks mentioning using RAID, i thought i'd chime in about RAID5 survivability.

RAID5 protects you against one failure in a stripe. if you lose a drive, that's a failure. If you have a read error on a particular sector, that is another failure, and your data is gone.

the probability of a read error *somewhere* on a 1TB drive is actually quite high.

So, you lose a drive, you go to rebuild, you find you have a read error and can't get your data.

This can mean a few things.

1) lose a particular bit of data. Maybe you don't care, if you're archiving DVDs you'll probably cope just fine. If it's important data you'll be sad.

2) can't rebuild your RAID. Some RAID controllers will just give up if they get a read error during a rebuild, so then you have to back up the recoverable portion of your data (probably the vast majority), rebuild the RAID, etc.

I don't know how the various software RAIDs cope with this. I had this happen with a dell/lsilogic hardware raid card.

In my case, the read error was not something i noticed when i backed up and restored the data onto a new raid, but the parity didn't match so it wouldn't rebuild. It very well may have been on an unused portion of the filesystem.

solutions/mitigations:

1) scrub your RAID5's regularly. this process checks everything over and fixes any errors while you still have a full RAID5 set. This will reduce your chance of failure greatly.

2) use RAID6. it adds an extra drive's worth of redundancy.

3) use http://en.wikipedia.org/wiki/Parchive or some similar additional layer of redundancy.

You should also consider running an OS that supports ZFS (FreeBSD, OpenSolaris, Nexenta). It has additional data checksumming that can help. ZFS has software raid built in.

Several good blog posts on this subject here:
http://blogs.sun.com/relling/tags/mttdl

He talks specifically about Solaris & ZFS, but the reliability stuff is generally applicable. RAID-Z is basically equivalent to RAID5; RAID-Z2 is basically RAID6.

Come to think of it, I might be confusing it with the TYAN SMDC card or possibly with something I read about the N1 provisioning system. It's been a while since I looked into it.

Looking at the ILOM configuration docs the only reference I found about "same subnet" is that you need to be on a computer that has "access to the same subnet as the ILOM" which isn't the same as having to be on the same subnet.

My mistake. Somewhere along the lines I read a forum post that some of Sun's motherboards were manufactured by TYAN and used the same IPMI card. Just goes to show you can't believe everything you read online.

Intel and AMD aren't doing 64+ execution threads on a chip yet.

(Most people's knowledge of SPARC - and MySQL, coincidentally - seems not to have been updated since 1998...)

With all the talk of container and "lego" data centers, Oracle wants to become fully vertically integrated so that you can go to Oracle and say: "I've got $10 million -- sell my data center blocks".

Sun's already been developing their own data-center-in-a-shipping-container, and Oracle now has all the bits and pieces:

• Hardware that runs Oracle really well -- Sun SPARC
• The operating system for big data centers -- Solaris
• The Java application server -- BEA's WebLogic
• The Database -- well duh!

Also, having a horde of hardware engineers is Ellison's wet dream. As I said before, Larry Ellison wakes up every morning and asks himself, "How can I [fsck] Microsoft today?" Larry has stated in the past he wouldn't mind moving beyond databases, and with Sun's hardware and Java, he's poised to do pretty much anything he wants. So he might entertain delusions of mobile, return of the net appliances, home multimedia, etc. In the short term, though, I think he's hoping he can create custom hardware to make Oracle and Java run much faster. Will he succeed? Dunno, but Larry Ellison has a ferocious desire to succeed, and often, that's all you need.

I think Niagara may fit the bill.

Interesting OO blog entry describing what the bottlenecks are and how they're improving OO startup time on Windows:

    http://blogs.sun.com/GullFOSS/entry/achievements_for_a_better_start

In this case, Sun said "about 500.000 lines of code were changed/rewritten" just for antialiasing, a task that "took 5 years to complete".

And they didn't even simply abstract out the drawing and use Cairo -- they wrote their own antialiasing. So it looks a little different from all my other apps, and now I'll have yet another implementation of AA in memory now. (If that's not bloat, what is?)

The screenshots look prettier, but I'm not sure if this is progress or not. If they want to add another graphics feature, will they have to touch 500000 lines of code again, and take another 5 years? Or do they have their own ad-hoc Cairo-like abstraction internally so they can just add it in one area of the code? They don't say.

Slashdot:
http://developers.slashdot.org/article.pl?sid=09/04/20/128246

Sun:
http://www.sun.com/third-party/global/oracle/

Oracle:
http://www.oracle.com/sun/index.html

Chicago Tribune:
http://www.chicagotribune.com/business/yourmoney/sns-yourmoney-0503leckeyfile,0,5105395.story

http://www.sun.com/products/sunmd/s20/

In terms of hardware multi-cores are delivering quite fine, its only the software that is lacking behind a lot, as todays programming language are pretty ill suited for parallel programming, but then the brain itself is a parallel processor, so that might not be that big of a problem for AI.

And anyway, just imagine a ship full of these these. Just because we might no longer be able to cram everything on a single chip, doesn't mean that we won't have a shitload of computing power in the future. The only real problem I see in slowing down Kurzweil's vision is the lack of a market. Its like the flying car thing, of course we can build them, but there wouldn't be many people with enough money to actually buy them, so they never got created in the first place.

Is this really a surprise?
Poor Dead Horse (MS) ... twack!

Sun ODF Plugin for Microsoft Office:

http://www.sun.com/software/star/odf_plugin/index.jsp

http://www.sun.com/software/star/odf_plugin/get.jsp

Is this really a surprise?
Poor Dead Horse (MS) ... twack!

Sun ODF Plugin for Microsoft Office:

http://www.sun.com/software/star/odf_plugin/index.jsp

http://www.sun.com/software/star/odf_plugin/get.jsp

Have they actually done so? The only ODF plug-in I see for Office is from Sun. http://www.sun.com/software/star/odf_plugin/

That link goes to the 1.2 version of the Sun plugin. The current one, which is also reviewed in TFA, is version 3.0.

Sun's ODF plugin for Office.

I've used Sparc desktops in the past. I even used one as my main home machine for a while. You could even get Sparc laptops.

In their time they beat the Intel option imo and they are still in use in some places.

This 2-axis rate gyro part came out in 2006. Analog Devices and Motorola have had comparable parts for years, but at a higher price.

It's only 2-axis. If they could do all 3 axes on a flat chip, that would be something. Usually, you need a second chip mounted vertically to the first one to get all three axes.

An elegant design is to use four MEMS gyros oriented along the axes of a tetrahedron. With that redundancy, you can detect faults. The Segway does that, for safety reasons.

Title says it all: http://www.sun.com/software/solaris/get.jsp

Actually, John McCreesh, head of marketing at OpenOffice.org, says that OpenOffice is NOT part of the same patent deal as StarOffice. You can see the link at Sun's Web site here

Yep, there was definitely a jump in postgresql around that time too, but the slope for MySQL appears to have gone up. There wasn't any sort of mass exodus from MySQL to PostgreSQL.

Maybe some of that was people afraid of what would happen to MySQL as a result of Sun's acquisition but it could also be a result of Sun providing support for PostgreSQL and including it in Solaris 10. Around June 2007 there was a big spike and that was when Sun came out with the first industry standard benchmark Sun put out with PostgreSQL, Sun Java App Server on a T2000 UltraSPARC T1 server.

I could really care less what happens to MySQL, I'm more concerned what happens with Sun's future contributions to PostgreSQL.

Here is the comparison. The indemnity is listed. Sun's Comparison

Walmart doesn't carry it, but there is a retail box version.

http://www.sun.com/software/staroffice/

"To be fair, every OS that I've ever dealt with has had issues with major upgrades. Whether it's glibc problems with older Linux binaries, or compatibility and driver problems moving to newer versions of Windows, there's always pain."

I just fired-up a bunch of programs I compiled while in college on a number of Solaris boxes ranging from 2.5, 2.5.1, and 2.6 and they all ran perfectly fine just now on Solaris 10. That is a span of something like 15 years of compatibility. So if your OS decides to do things like change the way thread local storage is done, add international support in a non backwards compatible way, move from COFF to ELF, then in ELF change and extend the format of the relocation entries, not to mention use a C++ tool chain that keeps modifying the ABI, then yeah you really have no recourse. Oh and Solaris has not really changed the device driver model much from SunOS 4.x, pretty much any driver (well it has to be 64-bit for a 64-bit kernel) from 2.6 is going to work in Solaris 10, unless it used something esoteric, going back before then will need some minor tweaks to some structures. Check out the man pages in section 9E or the funtions that start with 'ddi' sometime (alternately http://docs.sun.com/app/docs/doc/819-3196/ )

Brazil seem to have an amazing electronic healthcare system using Java. Maybe that pushed oracle to buy sun. http://java.sun.com/developer/technicalArticles /xml/brazil/index.html

I'm no expert on compilers. But my source certainly was: David Griswold, the original lead for HotSpot project. Now, this conversation happened more than a decade ago, so it's possible I misremember it. And of course it's always possible I misunderstood it. But I don't think I did.

They can't just drop support.

Sure they can, the only question is how long it will take before they can do it cost-effectively. Sun's normal sales terms talk about how they can pull out of their side of a support agreement in section 5.5. Typically Sun's contracts are executed for some number of years at a time, with even their Perpetual Entitlement Contracts having a renewal date in them. So Oracle could announce an end-of-life date effective a couple of years out, probably 3 years, and start tapering down the support work they do and the free resources available immediately (try and find good Oracle support info without a paid Metalink account).

They can't just drop support.

Sure they can, the only question is how long it will take before they can do it cost-effectively. Sun's normal sales terms talk about how they can pull out of their side of a support agreement in section 5.5. Typically Sun's contracts are executed for some number of years at a time, with even their Perpetual Entitlement Contracts having a renewal date in them. So Oracle could announce an end-of-life date effective a couple of years out, probably 3 years, and start tapering down the support work they do and the free resources available immediately (try and find good Oracle support info without a paid Metalink account).

When a company says "we're no longer going to support Windows 3x or Win9x, they should MEAN IT. NO support for the software.

They'd have to be insane to do that. Only an insane OS vendor would get incompatible with the largest collection of software in the history of computing.

IBM seems to be able to move forward without breaking APIs, and their mainframe systems were first introduced in the late '60s. Sun also has an ABI guarantee system for Solaris.

Porting dtrace would be useless, Linux has pretty much catched up in that front - the only piece missing is the merge of utrace in the main kernel. In distros like Fedora, which include utrace, you already can use systemtap to probe both the kernel and userspace without problems (sure, it lacks the "final polish" of dtrace, but all the hard has been done)

Please... an attempt to copy dtrace, and not a very good one at that.

You pick on an 11 year old server! (SPARC II's, sheesh!). I can't recall any of the E series having no ECC memory in them ... sounds like a colossal error, or 3rd party memory. ;)

The HCL for the E4500 [sunsolve|bigadmin account required] and full components list shows all memory. None of that recommended stuff is non ECC. I won't say you're incorrect about the situation (we've all seen crazier stuff than this :)), but the documentation clearly shows that SUN didn't offer non ECC memory for that system.

The SUN memory installation PDF makes no mention of non ECC memory either. Dated Jan 2000, 2 years after release..

You pick on an 11 year old server! (SPARC II's, sheesh!). I can't recall any of the E series having no ECC memory in them ... sounds like a colossal error, or 3rd party memory. ;)

The HCL for the E4500 [sunsolve|bigadmin account required] and full components list shows all memory. None of that recommended stuff is non ECC. I won't say you're incorrect about the situation (we've all seen crazier stuff than this :)), but the documentation clearly shows that SUN didn't offer non ECC memory for that system.

The SUN memory installation PDF makes no mention of non ECC memory either. Dated Jan 2000, 2 years after release..

So they're a software company. Where is it written they can't branch out? In effect, their salespeople are already selling hardware, because IT application deployments are almost always hardware/software stacks. The difference is that before Oracle salespeople had to hand off the hardware purchase to a hardware partner — along with its commissions.

Oracle is claiming that Sun will add $1.5 billion a year to their profits — starting in the first year. That number may or may not be realistic, but it lacks all credibility if it's based solely on monetizing Sun's (mostly open-source) software. They could make those kinds of profits if they cut costs drastically and do a much better job of selling Sun's servers than Sun has.

Both are eminently doable. There are a lot of Sun people (sales and marketing mostly) that Oracle won't need. And Oracle's own sales force (which is bigger than all of Sun!) isn't hampered by Sun's religious belief that you can sell a SPARC system to any customer if you try hard enough. Oracle will concentrate on selling SPARC to people who actually need it (and it just so happens that most of these people run Oracle software) and stop trying to push it to customers who are committed to commodity technology. Sun has put a lot of work into developing x64 servers that many of their sales people don't seem to care about, except as an x86/Linux-to-SPARC/Solaris migration path. Once these servers are in the hands of people who don't have a stubborn attachment to 1998 market models, they might finally get some traction.

Sun is phasing out sparc, but aggressively selling UltraSparc.

The cost vs performance is very impressive, and is winning people back from x86. At least in my environment (education).

The oracle database runs really well on the ultrasparcs. We just bought a 4 cpu T2 ultrasparc server, it has 8 cores per cpu for a total of 256 threads. With an edu discount, it cost ~60,000 if I recall correctly. There is nothing remotely as powerful in the x86 realm.
http://www.sun.com/processors/UltraSPARC-T2/specs.xml

Some of Sun's PostgreSQL contributions can be found in that link.

More importantly, Sun provides PostgreSQL support on Solaris.

According to Larry Ellison, Solaris/SPARC is the leading platform for Oracle deployments.

Oracle has been lagging with Solaris/x86 support so it would have been a great opportunity for sun to do more with PostgreSQL on Solaris and increase their revenues by making a more affordable alternative to Oracle.

PostgreSQL/Solaris/ZFS/DTrace could really eat into Oracle's market if there was more effort put into it.

I think they should have done more with pgsql, and missed a big opportunity, but I wouldn't belittle their contributions.

There are solutions which meet the OP's demand, depending upon what size or scale solution he wants.

Large scale: APC makes their InfraStruXure products which can be equipped with additional hardware to deal with humidity, include UPSes, and often AC. Racks are included as well.

http://www.apc.com/products/family/index.cfm?id=258
http://www.apc.com/products/family/index.cfm?id=259

Sun's "Project Blackbox", which is a portable datacenter accomplishing the same thing -- but this is much larger.

http://www.sun.com/products/sunmd/s20/

Smaller scale:

Try the ClimateCab from Black Box, which is just a rack:

http://www.blackbox.com/Store/Detail.aspx/ClimateCab-NEMA-12-Data-Cabinet-with-12-24-Rails/RM5010A ...but with their AC/dehumidifier unit attached, you should be able to achieve what you want:

http://www.blackbox.com/Store/Detail.aspx/ClimateCab-NEMA-12-Wallmount-Cabinet-with-Air-Conditioning-Beige-Single-Hinged/RMW5130AC

Finally, consider contacting some of these companies directly! Seriously. Many of them can make custom solutions for you which could meet your needs.

But good luck doing any of this "without breaking the bank". If money is a huge concern, co-locate your boxes somewhere proper. I'm serious: do it right.

Depends on the definition of "large".
One of the leading Danish Banks, Jyske Bank (4.145 employees, 122 offices), recently announced that they would be switching their online banking system to MySQL - the announcement (in Danish) is here:

http://dk.sun.com/sunnews/press/2009/090226.jsp

I fear that we'll see a standstill in MySQL development until one of the recent forks gain enough traction for it to be accepted across Linux distributions and see an uptake in development efforts from third parties.

Oracle developing MySQL? How many open source projects does Oracle maintain today (except from a ripped off OS)? How many small and medium sized customers does Oracle have today? Exactly...

Source IDC 2008:

market share:

"Unix, mid-to-high-end servers ($17.2 billion in 2008)

IBM 37.2 pct
Sun 28.1 pct
HP 26.5 pct

"

Don't give a flying fig about Suns servers?

IIRC Solaris still has the highest market share among proprietary Unixes. And AIX ist only third after HP-UX.

And if you think about Oracle as a database company you've kind of missed the last 8 years or so. They've bought a lot of stuff and are number two behind SAP.

"IBM provides Java and Java products. "

Well I guess Sun does that too.

Regarding virtualization: XVM Server

Should be enough to keep the troll busy ;-)

Bye egghat

Just build out a normal load balancer/apache web farm, but use Sun systems with the T2/T2+ CPUs for the web servers. These systems have crypto acceleration built into the CPUs and it is amazingly fast. These servers are also pretty cheap. Here's some benchmarks.

It's usually not cheaper to demolish and build new.

You have the cost of demo, construction, zoning issues, etc. If you can find a location that has many of the characteristics you need, you're usually better off. If you find a location that isn't suitable for much else, you can usually get it cheap.

By the way, the guy in the video looked familiar. At first I thought all the network people looked and sounded the same. But it turns out he's Dean Nelson, Senior Director Global Lab & Datacenter Design Services at Sun and founder of Data Center Pulse. I remember him from this data center video.