Slashdot Mirror

A much better article is linked from this one near the bottom: what every computer scientist should know about floating-point arithmetic

As of yesterday, Google is once again using a referrer system rather than just giving you the link. So when I do a search for, say, "sun", all the search results (not just the sponsored links) provide links that look like this:

http://www.google.com/url?sa=U&start=1&q=http://ww w.sun.com/&e=10401&sig=__2EeDlnaab7EWyO2SbtTK5ClAc PE=

I assume that's some sort of referrer and tracking hash. Regardless, it's very annoying, as it screws up the link for other uses unless I first edit it.

When I used the blackboxsearch and scroogle proxies, the link was in the sane and expected form:

http://www.sun.com/

["sun" used as the test example because it's short to type and has at least one known result]

As to things that don't work, I've noticed that Mozilla sometimes ignores lines in my HOSTS file; it still *contacts* blocked sites and sometimes still downloads data from them. The problem seems to be lines that include a subdomain or a path beyond the root domain. I don't know how "normal" that is??

Side note: my everyday browser loads neither images nor javascript nor CSS, tho I do allow cookies because I use them for a number of logins (frex, Slashdot :)

Expert Eric Laforest talks about stack computers and why they are better than register-based computers. Apparently NASA uses stack computers in some of their probes.

the other is the creator of one of the most prevalent markup languages used

The other guy is an ass in a hat who likes to suck up to his management.

No contest.

You mean source is available now? (http://www.sun.com/software/communitysource/j2se/ java2/download.xml)

And they are going to open source Flash within the next few months? (http://it.slashdot.org/article.pl?sid=06/07/24/17 52212)

Well that's good news!

</sarcasm>

Yes, there are ugly hacks to keep a connection alive, but it is exactly that, a hack, and introduces problems of it's own.

There are some ugly hacks to allow the server to "push" to the client (embedded flash objects, never-closed-connections, etc.)--mostly encapsulated by the moniker COMET (get it? Ajax... Comet...)

But if you get to pick your app server, there are some ready-made solutions. The problem with traditional web servers is their IO method. It's not their fault that the HTTP spec is out-of-date, but there are already new developments on the horizon that get around the current limitations. Take a look at GlassFish, Sun's new open-source enterprise application server, and pay particular attention to NIO socket writes. The performance benefits of NIO over straight IO are astonishing, with the side-benefit that it supports server-push out-of-the-box.

At first I was just thinking to myself that: "I'm glad this guy is just a pundit and in no position to do anything about implementing such a cruel device of torture."

But, then he had to mention 'the monopoly' and suddenly things seemed plausible. Not a nice thought.

The thought has passed in that I need not concern myself with such matters since I have both FreeBSD and GNU/Linux experience (with Solaris 10 soon to follow).

I have friends that work for Sun. If you don't believe me, you can always look it up.

I personally like snow, so living in Denver is a real disappointment. We get maybe 3 big storms every decade. Most winters are warm, dry, sunny. Every so often a dusting of snow or a cold snap. It's disgusting.

I'm glad they are using JAXA technology rather than AJAX!

I didn't know the Java API for XML Aeronautics was ready yet.

You mean like Matcher.replace(String s)?

http://java.sun.com/j2se/1.5.0/docs/api/java/util/ regex/Matcher.html#replaceAll(java.lang.String)

Point, they now can offer 3 types of CPU on their blade systems. They compare with HP and DELL, but seem to "accidentally" forget SUN who deliver some very nice AMD-based blades at the moment http://www.sun.com/servers/blades/8000/specificati ons.jsp

Anycase, IBM blades were x86 already before, this won't change too much on that.

Here is the class I assume the parent is referencing: Formatter class.

Essentially what happens is you don't have C-style varargs, the JRE silently creates an array for you when you pass the arguments. This doesn't waste "gobs of heap space" like the parent says, it uses the same amount as it would using the stack. Remember, these are objects, and Java never passes objects by value -- always by refence. So each argument wastes one machine word (usually 32 bits). Whoop de fucking doo. And, since it uses references, the only allocation/deallocation is the temporary array. And in 1.5 if not previous versions, this is very very fast. With a JIT compiler you'll hardly notice it. I do agree that the decision to make the class "final" is shitty, but honestly, I don't see how subclassing it would be a huge advantage. It would be like subclassing the java.lang.String class. Sure, you could add some nifty stuff, but it's not a big deal.

As a person who earns his living off of J2EE, I know its strengths and weaknesses. I am not a fanboy, however. I am more than willing to give Java hell when it deserves it. I think string handling in general is not as well-organized or easy to use as it could be, but it is certainly capable. I rarely use sprintf() style string formatting anyway, even in C++. I find it much easier to use iostreams, which are typesafe and almost as fast as sprintf(). In Java I just use string concatenation, and the formatting classes when I need it. It isn't perfect, but it works well enough and sure isn't slow.

You might be thinking about QUAD CORE compatible processors (which will require Socket F).

Here's an 8-socket opteron now (16 cores w/dual core):

http://www.sun.com/servers/x64/x4600/index.jsp

As a starting point, I'd like to suggest designers read, "A Whole New Mind" by Daniel Pink, and check out some articles at: http://www.danpink.com/. Furthermore, I suggest visiting IDEO http://www.ideo.com/ideo.asp. Pay special attention to their "method card" deck. Lastly (for purposes of this discussion) I suggest visiting http://www.mcdonough.com/# . The common thread in all this is DESIGN. William McDonough says that the need for regulation indicates a failure in design.

The design of the product goes 'way beyond just cosmetics. There is only so much you can do with an enclosure for a PC board, but there is LOTS you can do with the system as a whole. Case modding is just a place to start. Functional design improvements are being made in everything from the input devices ( http://www.extremetech.com/article2/0,1697,1112012 ,00.asp http://www.thinkgeek.com/computing/input/ ) to really innovative interfaces ( http://en.wikipedia.org/wiki/Croquet_project http://www.sun.com/software/looking_glass/).

The IDEO method cards are different from the "Creative Whack Pack" or "Thinkertoys" cards, in that they redefine the product design domain. The jobs of the future are going to be design jobs requiring both high creativity and high technical ability. If someone in India or China can do your job as well and cheaper than you, or if a computer can do your job better and faster, your job is obsolete.

"If a subsystem component fails in such a way that the overall system continues to function as designed, no system failure has occured." Because you say so. If a system is "designed" to crash when it's hard drive fails then no system failure occurs then according to your definition. You asked for the "relative system reliability" but never defined what that was. You still haven't, yet you declare your correctness without ever demonstating that you even understand common terminology. From http://www.sun.com/products-n-solutions/hardware/d ocs/html/817-3881-11/ras.html: "Reliability, availability, and serviceability (RAS) are aspects of a system's design that affect its ability to operate continuously and to minimize the time necessary to service the system. Reliability refers to a system's ability to operate continuously without failures and to maintain data integrity. System availability refers to the ability of a system to recover to an operational state after a failure, with minimal impact. Serviceability relates to the time it takes to restore a system to service following a system failure. Together, reliability, availability, and serviceability features provide for near continuous system operation." Fact is that your original question cannot be answered because you never defined "system", never gave MTBF numbers for drives, never gave costs of SCSI or ATA drives, never offered configurations, and never provided MTBF numbers for common components. All you specified was $1000 in drives for either case. It appeared that you didn't know what you were talking about and it still does. Now, if your point was that a more useful system can be constructed with ATA drives than SCSI for the same price then I might agree with you. Reliability differences between SCSI and ATA are mainly artificial and imagined anyway. Drives should be classified as enterprise and non-enterprise, not ATA and SCSI. Protocol is meaningless. Once that is accepted, the original RAID paper makes your point well enough.

Now obviously you could do the same with natively compiled code if all phones used the same hardware but they don't. So I'm not sure how using Linux is any guarantee against fragmentation. There are lots of embedded Linux solutions that run on lots of embedded processors with lots of embedded GUIs. Unless every Linux based phone maker happened to pick (and licence) the exact solution used by Motorola, there is no possibility that apps would run on one system to the next.

Rule number one is to never apologize for not having a Comp Sci degree in an interview.

(Rule number two is that you don't want to work anywhere that strictly requires it)

I have a Liberals Arts degree and I'm a completely self-taught, working Java developer. The only thing I did do to help my job search was get a Sun Certified Java Programmer certification. Almost no experienced developers give any merit to the thing, but during my job search it was useful to "prove" to the skeptical that I really do know Java.

doesn't the process have to be frozen for a full mark-and-sweep?

I already have a machine with 8 cores sitting on my desktop... a Sun T2000... OK, it's not supposed to be desktop machine, but it's 8 cores, and 4 threads per core...

I for one hope we see a Solaris kernel in Leopard, if nothing else for ZFS support, but having better kernel tracing and swapping routines would be nice as well.

Actually, even Sun has been known to get in trouble that way. No, they don't sue themselves, but when their developers fail to follow the spec (as they have been known to do), their customer make a lot of noise.

The software is not the spec. The spec is the spec.

Because I know you're waiting with bated breath for updates, here's more info, this time from someone directly involved, with more information. Blog posting from a Sun employee

It is fair to say that down the line even when they do opensource it, Sun's version will be the defacto standard.

You're somewhat misinformed. Sun's implementation has never been a basis for determining what's "standard". That's because Sun's implementation, like every other Java implementation (and there are quite a few) is required to adhere to a written specification.

People (including everybody at Sun) often say "Java" when they mean "Sun's Java implementation". That can be misleading. When you talk about "open sourcing Java" you're really talking about open sourcing a particular implementation of Java.

Which is exactly what Sun already does with the Java Compatibility Kit. There are, in fact, a fair number of third-party Java implementations that are released on this basis. And one famous lawsuit was filed agains a certain company that released an incompatible version of Java.

However, there have always been issues with the JCK, with claims that it is too complicated and inconsistent. Also, once everybody has access to the Java Development Kit source, you can expect to see a lot of forks. Enforcing compatibility will be a lot harder than it is now, with only a few Sun licensees doing Java implementations. And it is not easy now!

It took Sun 5 years to realize that Linux, esp with IBM support was quickly passing them by; and that even their efforts with Intellectual property partnerships with Microsoft and their backing (again with Microsoft) of SCO's efforts against Linux didn't slow down IBM or Linux.

So to make parallels with Java -- Sun will Open Source Java as soon as they realize C# and Python and Ruby are kicking their ass and Java no longer is relevant.

Considering that this is a Java course we're talking about, setting up a sandbox is easy and effective; use a policy file with only the permissions the student's code needs (for simple programs that don't need file I/O, a blank policy file could even be enough). The result should be roughly as safe as running Java applets on the net and for the same reasons.

"Yes, but just try getting your dual-Opteron to run an enterprise-level database application like a billing system."
I think you might be surpised how well a dual Opteron would run a large database system.
You might want to take a look at one of these. http://www.sun.com/servers/x64/x4600/ It can be expanded to 16 cpus.
Oh and so far the Conroe CPUs seem to be beating Athlons but the Opteron is a different critter. It still has an advantage in IO thanks to it's four hypertransport channels.

...because StringBuffer.toString() will share arrays until another modification is made to the StringBuffer...

StringBuffer:

public synchronized String toString() {
return new String(value, 0, count);
}

StringBuilder:

public String toString() {
// Create a copy, don't share the array
return new String(value, 0, count);
}

I believe that the literal is turned into a new String object by the compmiler. While it may be unnecessary to add the new, it's functionally identical and will consume the same amount of memory. From Sun's website:

All string literals in Java programs, such as "abc", are implemented as instances of this class.

--trb

"will the code virus's evolution be something benign like the common cold"

To continue this analogy how about innoculating the system against future atacks. Create a processor that scrambles the microcode table. It has a run mode and an install mode. At install mode it scrambles the OP codes in the program to match the table. Any forign code attempting to run is stoped dead in its tracks.

Or how to fix Windows. Create an embedded OS that runs an emulator that provides API functions to the applications. That way dot.NET would still be able to run without the security lapses. Oh, I just remembered with Embedded Java we would have had a safe and secure Internet years ago if his billness hadn't sabotaged it on the PC.

"Because you and most other dummies are using a highly insecure operating system" ,hritcu

Not I, its SuSE Linux. Personally I don't see Linux making the huge impact it deserves until the average dummy can walk into the high street shop and buy one.

"will the code virus's evolution be something benign like the common cold"

To continue this analogy how about innoculating the system against future atacks. Create a processor that scrambles the microcode table. It has a run mode and an install mode. At install mode it scrambles the OP codes in the program to match the table. Any forign code attempting to run is stoped dead in its tracks.

Or how to fix Windows. Create an embedded OS that runs an emulator that provides API functions to the applications. That way dot.NET would still be able to run without the security lapses. Oh, I just remembered with Embedded Java we would have had a safe and secure Internet years ago if his billness hadn't sabotaged it on the PC.

"Because you and most other dummies are using a highly insecure operating system" ,hritcu

Not I, its SuSE Linux. Personally I don't see Linux making the huge impact it deserves until the average dummy can walk into the high street shop and buy one.

Linux starts IPv6 implementation on verswion 2.1.8. Current 2.2.x and 2.4.x series supports IPv6 in a stable manner. In addition to the kernel maintainers, the USAGI project is working on someextension for production quality.

zfs is administered via the zpool(1m) and zfs(1m) command line utilities. They're quite a bit more friendly than most tools for managing raid, volumes and filesystems I've encoutered so far.

zfs is administered via the zpool(1m) and zfs(1m) command line utilities. They're quite a bit more friendly than most tools for managing raid, volumes and filesystems I've encoutered so far.

How tricky it is to self-define a lightweight XML format? Use whatever element you like, and design the data structure on your own that suits you best. Why do we need to design a new language?

How is it different from calling a server to output XML and output HTML/XHTML? Which modern browser today does not support XSLT? Firefox, Internet Explorer? (Yes, Opera will support XSLT 1.0 in the coming version 9)

BTW, There are server-side XSLT processors (for very-old browsers sake). For instance, this, this and this.

And finally, what's the point to make the document semantic if the browser ignores it?

Have you seen the array of fans on the front of the system?

http://blogs.sun.com/roller/page/jonathan?entry=th e_rise_of_the_general

They are forcing the air along the gaps existing between drive columns out the back.
As you can see, the gaps between the columns are fairly narrow, so air is moving
*quickly* along the disks. Having the top cover (not shown in the picture) in place
is important for the system's cooling, so that the air taken in by the fans along the front
has no place to go except between the disk columns, and out the back.

FWIW, ZFS has N+2 redundancy: http://blogs.sun.com/roller/page/ahl?entry=double_ parity_raid_z

As to backup and replication, think zfs: http://www.sun.com/2004-0914/feature/ Lots of folks are seeing this as simply a 2 socket server with lots of disk. With zfs it's more like a huge disk farm with an open, hackable interface and nice manners at the back end.

The 12 TB config is sold at $33k, or $2.75/GB, but assembling such a server yourself is possible and can be done today for 1/3rd of this price:

• 1 x dual-Opteron mobo = 1 x $500
• 2 x Opteron 285 = 2 x $1100
• 8 x 2 GB DDR400 registred DIMM = 8 x $300
• 6 x 8-port PCI-X Marvell SATA card = 6 x $100
• 48 x 250 GB 7.2kRPM SATA disks = 48 x $110
• 1 x Chassis+PSU+Rails = 1 x $1000
• Total = $11980 or $1.00/GB

(I have actually slightly overestimated the above prices.) Of course people are going to say that such a server is not be as reliable as a Sun server, that it does not come with technical support, etc. But in most cases such arguments are invalid because you save so much money that you can afford assembling/maintaining the server and replacing faulty hardware parts yourself. Time is money, but by having saved money you can now afford time ;-) The living proof that such a model would be successful is Google: instead of buying Sun servers like most startups in their time, they built their servers themselves to save money.

Picture gallery shows 5 hot-swappable dual fans for a total of 10. And the back doesn't appear to have any (unless there are some hiding in the power supplies, but it didn't look like it.)

But real cool 360 tour complete with the ability to take out drives, fans, CPUs, power supplies.

I think the bigger news is the sun blade system. They were miising it for so long and now they have something to compete with IBM and HP blade servers. And although sun's low end servers are something similar to blades ($1300 per unit duh!) they're not quite the same.

Last update: 07-05-2006
Features (as of this writing):
Sun Java 1.4.2

Now take a look at the documentation for the current JDK (1.5.0) and pay attention to the date. Solaris 10, also aimed at the Enterprise, shipped with Java 1.5.0 right from the start. Do they think that shipping outdated software makes them more stable or something ? This is ridiculous.

Rubbish...

Bob

Huh? Many Open Source projects or 3rd party companies offer paid technical support for the project. For example, both Sun and a variety of consultants provide paid support for OpenOffice, including its support for ODF.

In short, OO.o natively supports ODF and has technical support, MS Office has an ODF plugin in development, but MS won't offer tech support for it.

Solaris has something neat that is a step better than virtualization- you can define a container that holds the running state of a system and transfer it over to another system- LIVE. It runs as native code on the machine and there's only a little overhead for maintaining the state.

http://www.sun.com/software/solaris/ds/utilization .jsp

Dude...this isn't that complicated. The point of network computing is that by sharing computing resources you need less of them and get better performance. This is because of the massive amount of wasted CPU cycles that we're all aware of. There are thin clients available today for under $300 if you want a current exact example: http://www.sun.com/sunray/sunray1/. Since almost no one buys thin clients, the price is relatively high. Typically mass production would bring the cost down.

So lets look at it this way. Say you and your 100 closest neighbors all decided to get these $300 thin clients from Sun and you all pooled your money together to buy a grid of 30 lowend pcs for $300 (the price you mentioned) each that the thin clients would connect to that all run as a grid. The total amount you spent as a group would be $39,000. This is $390 per person total, lets just make it $400 to make it a round number. It could be marginally higher due to ethernet cables etc, but this works really well if you live in an apartment complex or a condo. So, for $400 (just a little bit more expensive than the cheapest of the cheap at dell), what do you get: You get a maxiumum of 30 cpus that are shared by 100 people. Since my previous estimate was that the average user uses no more than 1 cpu hour per day, you get access to this grid. Now, lets assume you don't want to let users use more than 25% of the total computing resources for their usage bursts. That means during your peak usage (opening programs, etc) you would have access to up to 7.5 cpus at the same time. So, in this example, for about $400 (slightly over the very cheapest box available) you get a 7.5 cpu system. I don't believe you can even buy a quad cpu system for less than a couple thousand dollars (certainly not less than $1000).

I keep hearing that the US goverment favors Sun and their hard/soft -ware big time. So can I conclude that they are just stupid and (which seems pretty common) "know best how to run their business" ?

I mean, just take a look at a random outdated workstation and you'll notice it has an embedded cardreader. Solaris also has native support for this kind of security measures and whats more; even their object oriented language which is always mentioned with regards to stability and security has full support for keycard authentication. So, where did they go wrong ?

I keep hearing that the US goverment favors Sun and their hard/soft -ware big time. So can I conclude that they are just stupid and (which seems pretty common) "know best how to run their business" ?

I mean, just take a look at a random outdated workstation and you'll notice it has an embedded cardreader. Solaris also has native support for this kind of security measures and whats more; even their object oriented language which is always mentioned with regards to stability and security has full support for keycard authentication. So, where did they go wrong ?

I keep hearing that the US goverment favors Sun and their hard/soft -ware big time. So can I conclude that they are just stupid and (which seems pretty common) "know best how to run their business" ?

I mean, just take a look at a random outdated workstation and you'll notice it has an embedded cardreader. Solaris also has native support for this kind of security measures and whats more; even their object oriented language which is always mentioned with regards to stability and security has full support for keycard authentication. So, where did they go wrong ?