If the teleported person (or bystanders) would in any way not perceive the recreation to be the same self as the destroyed original, teleportation would not be seen as transportation but death-with-birth-of-a-clone.
How would the teleportee be able to tell the difference?
I kind of get where you're coming from, but I think you're focusing blame on the wrong place.
Dumping the problem into the lap of the application programmers isn't leveraging anything very effectively.
Do you want the OS to automatically parallelize my application for me? Bear in mind that the chip already does that at the register level. How can the OS do that at a higher level with the object code it's got now? You'll need a new programming language that encourages parallelizable code to be visible to an optimizer, with an object code design that includes the metadata necessary. But the language will still need to support modern programming techniques, no going back to Fortran.
This is a very simple paradigm shift, and can almost emulated by VNC on your smartphone, but not satisfactorily, and not easily, as the OS doesn't understand it has to handle the UI for both the cell phone and the 30" flat panel monitor in the same application.
And how do I control which data is presented on which of the displays and still only reference a single output device in my code? Multiple devices in one app means you need multiple screen data interfaces, and most systems handle that just fine today. You'll need a few applications that illustrate the need for a single process image to support multiple output devices simultaneously. Right now if I've got a single CPU serving a cellphone display and a workstation I can just fire up two instances of my web browser and point each one to the right display and let them work out their own data sharing strategy.
Big paradigm shifts that would completely re-orient computing are impossible in the current market.
And for good reason. I don't WANT my computing completely re-oriented. As a tool right now it's predictable, functional, fast, and I already know how to get what I need done done with it. A significant paradigm shift had better give me something a lot better.
Data management is what computers do, sport. Better, faster and more convenient data management is why there's a computer industry at all. Wake up.
Yippie hoorah! Computers manage data! So?
"Moving data between various devices is the job of applications. If the applications aren't written to interoperate and share data intelligently, there's nothing the OS can do to fix that."
Nothing... except create better ways for applications to interoperate with other applications and with local and networked peripherals.... Dumping the problem into the application developer's lap was how MS-DOS did it, not how modern operating systems do it, and certainly not how next generation OS projects will do it.
Everything you want is available now, and even packaged reasonably well so users don't have to dork with it too much. Buy a PC, plug in USB storage devices, see them on the desktop and move files around. Install a database, fire up the control center (or open a CLI) and create tables and start dumping data in. Setup a RAID array, put a DVD backup drive on it, etc. It's all there, including IPC mechanisms that applications usually ignore but some make use of in good ways (like DB2).
You misunderstood the point. I wasn't knocking Java's speed from its 1.[01].x days.
I am simply pointing out that knocking Ruby's speed now, given its abilities, is akin to dismissing Java in 1995 given what it brought to the table. And for a Java enthusiast to make a general statement about performance as a primary reason to avoid a language/platform is selective amnesia at its finest.
Many java apps, including large, complex applications like Eclipse, JBoss, etc., have been shown to run using Classpath. Kaffe + Classpath or SableVM + Classpath is pretty dang close to a complete JDK 1.4 environment and will run many (maybe most) Java programs.
*nodding head* Yup, but that's still not Certified Java(tm) and Sun won't let anyone call it Java unless it passes 100% of the compatibility tests, which are NOT free (beer or speech). Maybe Sun will change their mind someday, or maybe Kaffe/Sabre + Classpath will become the defacto JVM. But it's not at all a rosy picture. Some of the comments here are illuminating.
By now I just don't care, I've moved out of the webapp space and Java is the wrong tool for my current projects.
But there are a lot of open-source JVM's being written.
Sorta. There's lots of VM's, but none that are a) full JRE's, and b) Java certified(tm). We've got "close" to #A but nothing yet for #B until GNU Classpath is finished. That's the "Java trap" RMS is talking about: java.* and javax.* are not yet open, and until they are all Java code is pinned to the Sun implementation.
They say 'no type checking' (if it acts like a duck it is) even though it is pretty much a necessity for large or reliable systems.
I used to think that strict typing was a Good Thing, but now I don't so much, and I ESPECIALLY don't like the way Java does it. The things I don't like about it:
1. The artificial chasm between function and Object. Every function must live in one (and only one) Object. This makes both standalone methods and multimethods (Lisp CLOS) impossible, and results in overcomplicated object heirarchies where you have to think: which Object should own this function? Half the Design Patterns would be unnecessary if we used something like CLOS.
2. Every data type is an instance of Object (except for the ones that aren't (primitives) which happen to be the basis of computation) and you're stuck casting all over the place to get access to the specific function you want. I've already got the type I want in a variable declaration, why do I need to keep casting to it? And now we have autoboxing so others apparently agree with me.
3. It's SO EASY to break the typing system, who really needs it? Example 1: serialize an Object to SQL, upgrade JVM, deserialize Object and watch the app break. Example 2: toss a numeric String ("123") into a List of Integers on a JSP form and watch the app break when the backend iterates the List. Example 3: load Object from Webapp 1 (with its own ClassLoader) and try to pass it via EJB to Webapp 2 (with its own ClassLoader) and watch the app break. Again, why all the fuss from the JVM?
On a separate but related issue: why checked Exceptions? Why the arbitrary distinction between Errors and Exceptions, and why are most real errors subclasses of Exception instead of Error? I relate this to the typing system because both language features assume a certain degree of design incompetency on the part of the programmer. I've worked on large projects, and I still think that if we tailor our tools to the lowest common denominator we will generally get software that isn't as good as it should be for the time invested.
The government never tried to reulgate Internet content when everyone was using dial-up. It seems like they would have had a better argument since dial-up uses regular phone lines to to transmit data.
You must not have been around during the BBS era. Google "modem tax".
Who cares who studies it! Race and sex don't matter!... To me, it's irrelevant if fewer girls are going into engineering and CS programs.
In Meritopia where you think you live, race, gender, orientation, and ethnicity don't matter. Here on planet Earth, they do matter for the simple reason that as human beings we participate in human society and the time spent at work accounts for about 25-40% of our lives. I for one do not want that time spent in some Bizarroland where otherwise intelligent engineering types insist that the Way It Is Now really is the Way It Was Always Meant To Be. Talk about irony.
Here is one of many problems with increasing gender/race/etc. gaps in professions: it leads to stagnant work environments that undermine any semblance of merit-based promotions. Everyone in the company suffers when every manager thinks and acts the same and mentors/promotes only those people who are the physical reflection of their younger selves. This particular problem is real and easily measured: what percentage of non-white, non-male, non-straight, etc. people that graduate in profession X manage to reach the higher rungs, and how quickly do they do so compared to their white/male/straight/etc. peers?
Most of these kinds of problems have an insidious aspect that makes them hard to deal with: they can exist and have dramatic effect even when all decisions look fine at the local level. This is how person A (white, male, ME) can get a 20% salary increase in 36 months yet person B (black, male, a good friend of mine) can get only a 10% increase and be told "you need to do better" -- for work of equal difficulty and similar levels of success. I needed to do better too, but for "some reason" my manager didn't see that as a reason to withhold financial incentive. My friend later discovered that almost every black person in the facility faced similar issues year in year out. This despite the annual "diversity awareness" drives, and this was also a (ahem) Fortune-10 multi-national that we all recognize.
Since this is Slashdot, let me go further with a programming analogy:
You are responsible for maintaining an end-user application that runs on your customers' servers. Your customers insist that on many of their deployments the product is slower than it should be, sometimes unusably so. Your response is that you've never seen a problem in your development sandbox and back that up with data from several excellent profilers and debuggers. It must be PEBKAC, and management should trust you because you have a CS degree. Customers continue to complain, so someone with a Human Factors degree is brought in to examine the product. Instead of trying it out on their own development sandbox, they monitor the application directly on the customers' boxes. They sit back and watch the customers as they actually use the product, and they write a report detailing the procedures the customers are following and the actual outcomes. With this data they show that the customers are indeed seeing the problems. You immediately follow the procedures, but to your surprise you get the acceptable speedy behavior on your development box.
What do you do now? Do you argue that the outcomes didn't actually occur, that the code can't possibly do what it is in fact doing? Or do you look somewhere new? If you are familiar with the "mysterious bug that only appears for customers" then you know how easy it is for trivial, invisibile things to have large consequences later on.
Talking about society issues from an analytical perspective is often like being the developer who thinks every problem can be isolated and solved with good profiler/debugger. The problem itself is the way we think about the problem, and if we can't directly experience life outside our own upbringing (whatever that may be) then we have to take on a measure of faith what other people say. If women enrolled in CS programs say that they experience gender bias, and that it involves a lot of direct but non-verbal signals, then we have to accept that as fact and try to deal with it.
If the application were designed properly, it would not require a rewrite to provide a web interface. The business logic in your client-server application would be on the server, and you would only need to provide a web-based client interface. Perhaps you're right that there is no real need for a web interace, but arguing that it means flushing the original application either indicates that you haven't considered the situation fully or that the application is crap and deserves to be flushed.
I disagree. If you have a client/server environment (PowerBuilder, VB/MS Access, Delphi, etc.) where the server is SQL and the client has a full programming language, then there is little need to write a middle tier, and if you are budget- or time-constrained (as the GP seemed to be) it would be pretty silly to create that abstraction layer.
There are LOTS of complex applications that need the power of a full programmable environment. You can put the business logic parts (the Model part of MVC) on the server with SQL constraints and triggers (or a middle-tier server that performs the same duty), but the View and Client parts are sometimes best handled as a single piece. You need the textboxes/radiobuttons/etc on the screens to be in the right places, you need images to show at the right times, and really complicated GUI behavior (e.g. change the sub-form when the combobox value changes, dependent on data in a query and values in a different form). Until just recently most web-based application frameworks were light-years behind client/server.
Re:The real 90s versus outdated 00s software
on
Java Is So 90s
·
· Score: 1
Or WebSphere.:)
(I used to work for both in RTP circa 2000-2003.)
Re:The real 90s versus outdated 00s software
on
Java Is So 90s
·
· Score: 1
If the software was designed to the Java specification it WILL work between these two virtual machine implementations.
Most times, yes it works fine. Add EJB+DBCS+SQL and write 500 KLOCs and things might go wrong even with "100% Pure Java(tm)".
But I suspect recent JVMs have improved the situation lot. I don't hear so many horror stories these days. There was a time (Java 1.[01].x) when the JVM would just core for no apparent reason (as in run it again, on the same data and same platform, and it works fine). Later (Java 1.[0-3].x) we discovered that all AWT-based apps (like chart generation libraries) required an X11 display somewhere (that's fun to explain to customers, that your app which runs in a server room that has no CRTs in it nevertheless needs to access an X11 server, so could they please set up a virtual framebuffer server?). Then there was the whole "green threads, native threads, and JIT" suite of issues. And there will likely be a lot of problems after GNU Classpath becomes the first non-Sun-derived "Certified Java(tm)" implementation.
Anyway, YMMV. If you're happy with Java, keep using it, it'll only get better.
Re:The real 90s versus outdated 00s software
on
Java Is So 90s
·
· Score: 1
I stopped using Java professionally around JDK 1.3.2 (WebSphere 5.0x). In the J2EE/WebSphere environment, we encountered about a dozen problems due to JVM incompatibilities. Here are a few I remember off the top of my head:
1. DBCS charsets on Solaris (Sun-only JVM, 1.3.x) cored the install (InstallShield MultiPlatform).
2. DBCS charsets on AIX+Linux (IBM JVM) required a custom environment parameter to render correctly on AWT buttons. This due to an undocumented IBM "enhancement".
3. Spawning a sub-process on AIX 5L and Windows (IBM JVM) caused the sub-process to die (we saw "stackpointer=xxxxxxxx" in the outputstream) but the parent process never noticed == hung application. (Same code worked fine on Solaris, AIX 4.3.x, and Linux.)
4. We had about 20 EJBs in the application. Running a long time under Solaris (Sun JVM) caused an erroneous OutOfMemoryError: the HotSpot JIT was misbehaving. An undocumented -XX parameter could prolong the application's lifetime before the error appeared, but nothing actually fixed the problem. Sun closed the defect after two years (and confirmation from dozens of customers) with "non-reproducible".
5. On AIX (IBM JVM 1.2.2) we had to disable the JIT or it would core under heavy load (not sure why).
6. (I like this one.) Instantiating an AWT object, on Linux (IBM JVM), with the DISPLAY variable pointing to X11 server via tunnelled ssh: immediate core. Put the DISPLAY over a non-ssh link, all is well. Remove the X11 server entirely, the AWT libraries stopped working (this was before headless support) but the JVM wouldn't core.
I felt really bad for the install developer. He had to put some extra work into it so that we could run ISMP using any JVM, and then create a matrix of "operating system / charset / defect" combinations so that customers could figure out which JVM was capable of installing the product. (Once installed, the WebSphere JVM was used to run the product.) There was no way to provide a default JVM for each platform that would workaround every defect invisibly to the user. The available JVMs were IBM and Sun, 1.2.0 through 1.4.x.
I used to buy into the Java hype, now it's simply one tool of many. In some problem domains (such as large enterprise webapps) it's the best tool available, but there are many places that I see Java as inappropriate. YMMV.
I am just barely getting into Lisp, and actually enjoying it a lot. However, be prepared for anything beyond "hello world" to take a bit longer that you are normally accustomed to.:)
1) What to you do/get for a GUI? Like Algol famously leaving "IO to the library" meaning they punted on IO and this really hurt Algol, Python leaves the GUI to the library, but there are two mainline choices -- TKInter and wxWidgets/wxPython. Does Lisp in a Box have no GUI, some homebrew Smalltalk/Squeek sandboxed GUI, or does it link to TK or wx?
Most CommonLisps can reach the GUI via Tk, Xlib, GTK, wx, etc. I don't know which ones actually ship with said libraries, but there are easy to find (see below for CLiki).
2) Can you link to C/C++ to do drivers, low level stuff, reuse DLLs or SO modules, or is Lisp beyond connecting outside of the Lisp runtime? Even Java can reach outside with the much-maligned but really isn't too hard to use JNI -- I like JNI because it allows C/C++ to call Java and Java to call C/C++.
Yes, Common Lisp has a "Foreign Function Interface" that supposedly is pretty easy to use. Most implementations also provide access to OS-specific functionality via non-standard functions (i.e. "(unix:unix-getpid)" returns getpid()).
Check out the Common Lisp Wiki: http://www.cliki.net/index . There's loads of code that can get you going.
the perception is that Lisp is like some Smalltalk implementations -- so highly sandboxed to prevent this sort of thing. What is the story on this?
My personal perception of Lisp: whereas Java has essentially one mainline JVM (I include the libraries java.*/javax.* in the definition of JVM BTW, so until GNU Classpath is complete there is only one mainline JVM) and a handful of secondary JVMs (just the VM) for specialized environments, Lisp has about half a dozen mainline VMs including both commercial and open-source implementations. Each Lisp VM is cross-platform and in total they probably cover about as many platforms as the one Java VM. Code can be ported between the Lisp implementations without too much hassle. Most Lisp VMs are fast enough that they can actually be used for shell scripting (0.5s startup), and they have the same garbage collection and runtime optimization advantages as Perl, and they hover around 3MB RAM footprint like Perl.
I think the perception of Lisp as slow, RAM-hungry, and hard to interface with non-Lisp was probably true back when machines were small enough that 3MB RAM was enormous. By today's standards, a Lisp VM is quite snappy and can be (relatively) easily integrated into a non-Lisp environment.
I have chosen to learn Lisp because I want to think differently about programming, and see what people mean by "functions are first-class objects". So far it IS different but refreshing.
Re:The real 90s versus outdated 00s software
on
Java Is So 90s
·
· Score: 1
but aside from that is there a meaningful difference between how Java treats these things as objects vs. the other languages you mention?
You can't get Java to take the code from one method (or anonymous function) and stick it in a different class at runtime.
but is there a tool to automate dependency checking, to see which services need to be shut down, to actually shut them down / unload modules, and then relaunch services?
But almost everyone I talk to who complains about the "architecture" is really complaining about tools, like shells and programs (e.g., PHP, Perl, etc.), and is relatively clueless about the OS architecture (disclosure, I think the modern NT kernel is pretty well-designed, even if the default tools are sub-par.)
I'll try a stab. As others pointed out, the "Unix architecture" is still a debated term, despite various books on the subject. Fundamentally, Unix is the POSIX-like C API (fopen(), brk(), fork(), etc.). However, many Unix purists I know say the tools are also part of the architecture, e.g. sh, passwd, chsh, ls, rm, etc., since the "philosophy" of the architecture is simple processes doing their one thing well and the system making it easy to string the tools together. From this base the other layers can be stacked on -- but they remain optional to the functioning of the computer.
The "Windows architecture" seems to be less philosophical and more pragmatic (in a commercial sense.) We have the WinAPI with layers above it including programming languages (MFC, VB,.NET, etc.) and data sharing layers (OLE, COM, etc.), and the only philosophy I've seen to it is "make it easy to put data on the screen for the user". Hence spawning a process in Windows is almost as complex as firing up a full-blown GUI application with menubar, dialogs, etc. (yet spawning a thread is as fast as Unix). However, getting the GUI OUT is a harder problem. Did Microsoft finally ship a Windows server that didn't require a graphics card?
I recently had a project that I think illustrates the difference in "philosophies" on a more technical level. I had two "embedded" computers that would be deployed on a sea platform: one was a PC/104 (100MHz 486) running TinyLinux in 32MB RAM and the other was a embedded Pentium 266MHz board running Win2k workstation in 64MB RAM. Yes, apples and oranges, absolutely. The vendor who supplied the Windows-based instrument had never put one of their instruments underwater OR under external automation before; they actually thought adding an entry to the Startup folder would be sufficient to make it all work.
On the Linux PC, I already had an image from the vendor that included a kernel and BusyBox and essential filesystem (16 MB). The system would boot and drop me to a shell on COM2, with ssh server listening on eth0. I compiled separately and added to the system just the utilities I needed: rsync, rsh/rlogin, inetd, Perl, crond, etc. I also recompiled the kernel, shaving 1MB off the memory use. Last, I re-partitioned the system such that only the/dev,/var, and data directories were read-write, all else was read-only, so that in the event of power failure the system would be able to reach fsck at least. The first such system was put to sea and lasted over 8 months with zero reboots until Katrina knocked it out, all the while continually notifying a hardware watchdog circuit at 0.5 second intervals that it was still alive. The system configuration part took roughly three weeks, including testing, mainly due to IRQ issues with the proprietary hardware (USB, ethernet, PCMCIA, wireless, 12 serial ports, A/D, DIO, hardware watchdog).
The Windows PC was already set up and needed only a "little" software to integrate with the Linux PC. Like I said, the data acquisition software had never been designed for automation. I used Kermit95 to manage the serial port link between the two systems (best $50 we spent by far). However, I had to write programs in C to start the program asynchronously (it would hang Kermit95's remote host command otherwise) AND stop the program by sending a specific series of key presses and mouse clicks (because it couldn't respond to anything else like a signal or status file). I also had to download third-party software to list and kill processes (PsList, PsKill). Testing was a royal pain and getting around the vendor's naivete ultimately cost us another three weeks.
Could you elaborate on that? As someone else pointed out, DB2 does have savepoints which can bring you back to any state. And (granted this is at the JDBC driver level) I haven't seen Oracle support more than two transaction isolation levels whereas DB2 provides all four mentioned in the JDBC spec. And of course one can see how DB2 is escalating row-level locks so queries can be changed to not step on each other.
I guess my question is: what are you working with such that Oracle does much better than DB2?
If you've got so much time on your hands, I would strongly suggest studying a bit more math/engineering before you're out of high school. I had similar grades as you, started college at 16 (accelerated program), and have returned to grad school in engineering and it is way more intense (fun) than my undergrad in computer science. Thermodynamics, vector calculus, and fluid mechanics are just barely on the edge of comprehension for me, and it's delightful to be challenged at this level.
If I was 15 again, I would get one of the several Dover books on vectors, tensors, and continuum mechanics and go page-by-page through it slowly over the next year.
Don't be ridiculous. China will have absolutely no control of the Internet. As for China being the "next" empire. That is dubious, according to the CIA 15 year report, it will be the EU that will become the next superpower, economically, and militarily.
Is this the same CIA that swore Iraq had WMD? Or the CIA that swore it didn't? Could the White House, which has had a vested interest in downplaying the role of China for the last twenty years, possibly influence the CIA's assessment of China's importance?
The folks who work with the ChaiVM have informed me that it is being obsoleted and is no longer available. Sorry!
Cindy for the OV Interconnect team
I'm curious if there are *any* certified-Java(tm) implementations that don't rely on Sun code for java.* and javax.*. Let me be more specific: certified-Java(tm) J2SE runtime environments at 1.2+ level. There are plenty of JVMs and individual libraries, and GNU Classpath has like 80% coverage, good enough to run OOo on a free software stack, but I haven't heard of any competing implementations actually certified by Sun.
They have even done this for clean-room implementations by rival companies like HP.
Does HP's "clean-room implementation" use licensed code from Sun like java.* and javax.*?
Re:Burn out at work is not always work related!
on
Pay vs. Happiness
·
· Score: 1
I'll pick this post to respond to.
You seem to have little idea how much stuff costs to someone with no assets, 18 years old, and living in a city with no decent public transportation. They can't just buy some land and stick a trailer on it, or a car, or the insurance. (Do you even know how much legally-required insurance costs in the first 3 years? Try $2000/year, which I should note is 20% of annual income at minimum wage. Full coverage is $3000/year.) They might not even have a computer at all or justification to buy one and learn some skillz on it. In short, stuck at $7/hr and no place to live or way to get between there and work.
Now if you add $10,000 help from parents it's all different. After 3 years, car insurance drops to about $800/year, and with a clunker to get to/from work you can get yourself to the $10/hr job that's across town, and with enough time sure you can eventually get that land and put a trailer on it and start living much better. You can also start growing your own food -- gardening is ridiculously easy, the food is of far higher quality that grocery store, and the exercise will keep you thin and save $30/month on gym fees.
This is mostly how my wife lived. She started working 60 hrs/week when she was 14 years old and had two cars paid for and two parcels of land paid for by the time she was 23. She lived simply, bought most of her food in bulk, and restaurants were rare events. However, she had some medical bills come up that basically wiped out her savings and forced a couple thousand dollars on her credit cards -- and she HAD insurance that paid over 80%. It took both of us a while to get that cleared out.
Me, I went to college, worked five years for *big name corp*, and am back in college for a Masters in a technical discipline. Yeah, we're taking on debt for it, but I've got a passion for it and a long list of professional contacts from career #1 who can vouch for my work ethic. I don't regret the first degree or its debt, since that was the only way out of the shithole town I grew up in. My best friend did your method, skipped the school and got a good job, but after the dot-com meltdown he had to move out of his town because *no one* could find work with all the layoffs. He's back in school and almost finished now and will soon be comfortable again.
The moral of all our stories is this: shit happens, we adapt, but we don't blame the victims about it. Medical costs a lot, cars cost a lot, and someone with lots of patience, a good work ethic, and no materialism will *still* be screwed if they were born broke too.
Good for you that you lived in a town that could hire you with no college degree, and from that lowly start you busted ass and got on up. But your advice in other posts, while nice-sounding, is not possible for many, and reveals a very brittle worldview. "Watch your girlfriend for a dark side"? Huh? I'd hate to be married to someone that controlling.
Slashdot Mirror
If the teleported person (or bystanders) would in any way not perceive the recreation to be the same self as the destroyed original, teleportation would not be seen as transportation but death-with-birth-of-a-clone.
How would the teleportee be able to tell the difference?
You keep asking the same question "how" are you going to do it, and I'm asking you "who" is stopping you from doing it! *maximum encouragement voice*
The creditors expecting repayment of the existing 2 trillion dollars in consumer debt.
The landlord expecting payment for rent THIS month, not next year.
The bank that sees no point in extending a loan to someone with a measly job.
The 5% of the population that controls 90% of the wealth.
Oh wait, were you going to launch all this with the Year of Jubilee?
I kind of get where you're coming from, but I think you're focusing blame on the wrong place.
... Dumping the problem into the application developer's lap was how MS-DOS did it, not how modern operating systems do it, and certainly not how next generation OS projects will do it.
Dumping the problem into the lap of the application programmers isn't leveraging anything very effectively.
Do you want the OS to automatically parallelize my application for me? Bear in mind that the chip already does that at the register level. How can the OS do that at a higher level with the object code it's got now? You'll need a new programming language that encourages parallelizable code to be visible to an optimizer, with an object code design that includes the metadata necessary. But the language will still need to support modern programming techniques, no going back to Fortran.
This is a very simple paradigm shift, and can almost emulated by VNC on your smartphone, but not satisfactorily, and not easily, as the OS doesn't understand it has to handle the UI for both the cell phone and the 30" flat panel monitor in the same application.
And how do I control which data is presented on which of the displays and still only reference a single output device in my code? Multiple devices in one app means you need multiple screen data interfaces, and most systems handle that just fine today. You'll need a few applications that illustrate the need for a single process image to support multiple output devices simultaneously. Right now if I've got a single CPU serving a cellphone display and a workstation I can just fire up two instances of my web browser and point each one to the right display and let them work out their own data sharing strategy.
Big paradigm shifts that would completely re-orient computing are impossible in the current market.
And for good reason. I don't WANT my computing completely re-oriented. As a tool right now it's predictable, functional, fast, and I already know how to get what I need done done with it. A significant paradigm shift had better give me something a lot better.
Data management is what computers do, sport. Better, faster and more convenient data management is why there's a computer industry at all. Wake up.
Yippie hoorah! Computers manage data! So?
"Moving data between various devices is the job of applications. If the applications aren't written to interoperate and share data intelligently, there's nothing the OS can do to fix that."
Nothing... except create better ways for applications to interoperate with other applications and with local and networked peripherals.
Massive data storage: RAID + jfs/xfs/reiser4 + DB2/Oracle : hardware, OS, applications
Networking: Gigabit ethernet + IPv6 + NFS/LDAP/Kerberos/FTPd/HTTPd/X11 : hardware, OS, applications
Everything you want is available now, and even packaged reasonably well so users don't have to dork with it too much. Buy a PC, plug in USB storage devices, see them on the desktop and move files around. Install a database, fire up the control center (or open a CLI) and create tables and start dumping data in. Setup a RAID array, put a DVD backup drive on it, etc. It's all there, including IPC mechanisms that applications usually ignore but some make use of in good ways (like DB2).
What are you actually asking for?
So, comments like
:: Pot:Kettle
Java:Ruby
Are an old, out-of-date joke told badly.
You misunderstood the point. I wasn't knocking Java's speed from its 1.[01].x days.
I am simply pointing out that knocking Ruby's speed now, given its abilities, is akin to dismissing Java in 1995 given what it brought to the table. And for a Java enthusiast to make a general statement about performance as a primary reason to avoid a language/platform is selective amnesia at its finest.
This is the same excuse I have heard for decades when fans of a language try and 'justify' it's lack of performance.
:: Pot:Kettle
Java:Ruby
Many java apps, including large, complex applications like Eclipse, JBoss, etc., have been shown to run using Classpath. Kaffe + Classpath or SableVM + Classpath is pretty dang close to a complete JDK 1.4 environment and will run many (maybe most) Java programs.
*nodding head* Yup, but that's still not Certified Java(tm) and Sun won't let anyone call it Java unless it passes 100% of the compatibility tests, which are NOT free (beer or speech). Maybe Sun will change their mind someday, or maybe Kaffe/Sabre + Classpath will become the defacto JVM. But it's not at all a rosy picture. Some of the comments here are illuminating.
By now I just don't care, I've moved out of the webapp space and Java is the wrong tool for my current projects.
But there are a lot of open-source JVM's being written.
Sorta. There's lots of VM's, but none that are a) full JRE's, and b) Java certified(tm). We've got "close" to #A but nothing yet for #B until GNU Classpath is finished. That's the "Java trap" RMS is talking about: java.* and javax.* are not yet open, and until they are all Java code is pinned to the Sun implementation.
You were going well until this:
They say 'no type checking' (if it acts like a duck it is) even though it is pretty much a necessity for large or reliable systems.
I used to think that strict typing was a Good Thing, but now I don't so much, and I ESPECIALLY don't like the way Java does it. The things I don't like about it:
1. The artificial chasm between function and Object. Every function must live in one (and only one) Object. This makes both standalone methods and multimethods (Lisp CLOS) impossible, and results in overcomplicated object heirarchies where you have to think: which Object should own this function? Half the Design Patterns would be unnecessary if we used something like CLOS.
2. Every data type is an instance of Object (except for the ones that aren't (primitives) which happen to be the basis of computation) and you're stuck casting all over the place to get access to the specific function you want. I've already got the type I want in a variable declaration, why do I need to keep casting to it? And now we have autoboxing so others apparently agree with me.
3. It's SO EASY to break the typing system, who really needs it? Example 1: serialize an Object to SQL, upgrade JVM, deserialize Object and watch the app break. Example 2: toss a numeric String ("123") into a List of Integers on a JSP form and watch the app break when the backend iterates the List. Example 3: load Object from Webapp 1 (with its own ClassLoader) and try to pass it via EJB to Webapp 2 (with its own ClassLoader) and watch the app break. Again, why all the fuss from the JVM?
On a separate but related issue: why checked Exceptions? Why the arbitrary distinction between Errors and Exceptions, and why are most real errors subclasses of Exception instead of Error? I relate this to the typing system because both language features assume a certain degree of design incompetency on the part of the programmer. I've worked on large projects, and I still think that if we tailor our tools to the lowest common denominator we will generally get software that isn't as good as it should be for the time invested.
The government never tried to reulgate Internet content when everyone was using dial-up. It seems like they would have had a better argument since dial-up uses regular phone lines to to transmit data.
You must not have been around during the BBS era. Google "modem tax".
Who cares who studies it! Race and sex don't matter! ...
To me, it's irrelevant if fewer girls are going into engineering and CS programs.
In Meritopia where you think you live, race, gender, orientation, and ethnicity don't matter. Here on planet Earth, they do matter for the simple reason that as human beings we participate in human society and the time spent at work accounts for about 25-40% of our lives. I for one do not want that time spent in some Bizarroland where otherwise intelligent engineering types insist that the Way It Is Now really is the Way It Was Always Meant To Be. Talk about irony.
Here is one of many problems with increasing gender/race/etc. gaps in professions: it leads to stagnant work environments that undermine any semblance of merit-based promotions. Everyone in the company suffers when every manager thinks and acts the same and mentors/promotes only those people who are the physical reflection of their younger selves. This particular problem is real and easily measured: what percentage of non-white, non-male, non-straight, etc. people that graduate in profession X manage to reach the higher rungs, and how quickly do they do so compared to their white/male/straight/etc. peers?
Most of these kinds of problems have an insidious aspect that makes them hard to deal with: they can exist and have dramatic effect even when all decisions look fine at the local level. This is how person A (white, male, ME) can get a 20% salary increase in 36 months yet person B (black, male, a good friend of mine) can get only a 10% increase and be told "you need to do better" -- for work of equal difficulty and similar levels of success. I needed to do better too, but for "some reason" my manager didn't see that as a reason to withhold financial incentive. My friend later discovered that almost every black person in the facility faced similar issues year in year out. This despite the annual "diversity awareness" drives, and this was also a (ahem) Fortune-10 multi-national that we all recognize.
Since this is Slashdot, let me go further with a programming analogy:
You are responsible for maintaining an end-user application that runs on your customers' servers. Your customers insist that on many of their deployments the product is slower than it should be, sometimes unusably so. Your response is that you've never seen a problem in your development sandbox and back that up with data from several excellent profilers and debuggers. It must be PEBKAC, and management should trust you because you have a CS degree. Customers continue to complain, so someone with a Human Factors degree is brought in to examine the product. Instead of trying it out on their own development sandbox, they monitor the application directly on the customers' boxes. They sit back and watch the customers as they actually use the product, and they write a report detailing the procedures the customers are following and the actual outcomes. With this data they show that the customers are indeed seeing the problems. You immediately follow the procedures, but to your surprise you get the acceptable speedy behavior on your development box.
What do you do now? Do you argue that the outcomes didn't actually occur, that the code can't possibly do what it is in fact doing? Or do you look somewhere new? If you are familiar with the "mysterious bug that only appears for customers" then you know how easy it is for trivial, invisibile things to have large consequences later on.
Talking about society issues from an analytical perspective is often like being the developer who thinks every problem can be isolated and solved with good profiler/debugger. The problem itself is the way we think about the problem, and if we can't directly experience life outside our own upbringing (whatever that may be) then we have to take on a measure of faith what other people say. If women enrolled in CS programs say that they experience gender bias, and that it involves a lot of direct but non-verbal signals, then we have to accept that as fact and try to deal with it.
If the application were designed properly, it would not require a rewrite to provide a web interface. The business logic in your client-server application would be on the server, and you would only need to provide a web-based client interface. Perhaps you're right that there is no real need for a web interace, but arguing that it means flushing the original application either indicates that you haven't considered the situation fully or that the application is crap and deserves to be flushed.
I disagree. If you have a client/server environment (PowerBuilder, VB/MS Access, Delphi, etc.) where the server is SQL and the client has a full programming language, then there is little need to write a middle tier, and if you are budget- or time-constrained (as the GP seemed to be) it would be pretty silly to create that abstraction layer.
There are LOTS of complex applications that need the power of a full programmable environment. You can put the business logic parts (the Model part of MVC) on the server with SQL constraints and triggers (or a middle-tier server that performs the same duty), but the View and Client parts are sometimes best handled as a single piece. You need the textboxes/radiobuttons/etc on the screens to be in the right places, you need images to show at the right times, and really complicated GUI behavior (e.g. change the sub-form when the combobox value changes, dependent on data in a query and values in a different form). Until just recently most web-based application frameworks were light-years behind client/server.
Or WebSphere. :)
(I used to work for both in RTP circa 2000-2003.)
If the software was designed to the Java specification it WILL work between these two virtual machine implementations.
1 079&cid=14255921 .
Not always, especially in J2EE land (i.e. WebSphere/WebLogic/etc). Some examples here: http://developers.slashdot.org/comments.pl?sid=17
Most times, yes it works fine. Add EJB+DBCS+SQL and write 500 KLOCs and things might go wrong even with "100% Pure Java(tm)".
But I suspect recent JVMs have improved the situation lot. I don't hear so many horror stories these days. There was a time (Java 1.[01].x) when the JVM would just core for no apparent reason (as in run it again, on the same data and same platform, and it works fine). Later (Java 1.[0-3].x) we discovered that all AWT-based apps (like chart generation libraries) required an X11 display somewhere (that's fun to explain to customers, that your app which runs in a server room that has no CRTs in it nevertheless needs to access an X11 server, so could they please set up a virtual framebuffer server?). Then there was the whole "green threads, native threads, and JIT" suite of issues. And there will likely be a lot of problems after GNU Classpath becomes the first non-Sun-derived "Certified Java(tm)" implementation.
Anyway, YMMV. If you're happy with Java, keep using it, it'll only get better.
I stopped using Java professionally around JDK 1.3.2 (WebSphere 5.0x). In the J2EE/WebSphere environment, we encountered about a dozen problems due to JVM incompatibilities. Here are a few I remember off the top of my head:
1. DBCS charsets on Solaris (Sun-only JVM, 1.3.x) cored the install (InstallShield MultiPlatform).
2. DBCS charsets on AIX+Linux (IBM JVM) required a custom environment parameter to render correctly on AWT buttons. This due to an undocumented IBM "enhancement".
3. Spawning a sub-process on AIX 5L and Windows (IBM JVM) caused the sub-process to die (we saw "stackpointer=xxxxxxxx" in the outputstream) but the parent process never noticed == hung application. (Same code worked fine on Solaris, AIX 4.3.x, and Linux.)
4. We had about 20 EJBs in the application. Running a long time under Solaris (Sun JVM) caused an erroneous OutOfMemoryError: the HotSpot JIT was misbehaving. An undocumented -XX parameter could prolong the application's lifetime before the error appeared, but nothing actually fixed the problem. Sun closed the defect after two years (and confirmation from dozens of customers) with "non-reproducible".
5. On AIX (IBM JVM 1.2.2) we had to disable the JIT or it would core under heavy load (not sure why).
6. (I like this one.) Instantiating an AWT object, on Linux (IBM JVM), with the DISPLAY variable pointing to X11 server via tunnelled ssh: immediate core. Put the DISPLAY over a non-ssh link, all is well. Remove the X11 server entirely, the AWT libraries stopped working (this was before headless support) but the JVM wouldn't core.
I felt really bad for the install developer. He had to put some extra work into it so that we could run ISMP using any JVM, and then create a matrix of "operating system / charset / defect" combinations so that customers could figure out which JVM was capable of installing the product. (Once installed, the WebSphere JVM was used to run the product.) There was no way to provide a default JVM for each platform that would workaround every defect invisibly to the user. The available JVMs were IBM and Sun, 1.2.0 through 1.4.x.
I used to buy into the Java hype, now it's simply one tool of many. In some problem domains (such as large enterprise webapps) it's the best tool available, but there are many places that I see Java as inappropriate. YMMV.
I am just barely getting into Lisp, and actually enjoying it a lot. However, be prepared for anything beyond "hello world" to take a bit longer that you are normally accustomed to. :)
1) What to you do/get for a GUI? Like Algol famously leaving "IO to the library" meaning they punted on IO and this really hurt Algol, Python leaves the GUI to the library, but there are two mainline choices -- TKInter and wxWidgets/wxPython. Does Lisp in a Box have no GUI, some homebrew Smalltalk/Squeek sandboxed GUI, or does it link to TK or wx?
Most CommonLisps can reach the GUI via Tk, Xlib, GTK, wx, etc. I don't know which ones actually ship with said libraries, but there are easy to find (see below for CLiki).
2) Can you link to C/C++ to do drivers, low level stuff, reuse DLLs or SO modules, or is Lisp beyond connecting outside of the Lisp runtime? Even Java can reach outside with the much-maligned but really isn't too hard to use JNI -- I like JNI because it allows C/C++ to call Java and Java to call C/C++.
Yes, Common Lisp has a "Foreign Function Interface" that supposedly is pretty easy to use. Most implementations also provide access to OS-specific functionality via non-standard functions (i.e. "(unix:unix-getpid)" returns getpid()).
Check out the Common Lisp Wiki: http://www.cliki.net/index . There's loads of code that can get you going.
the perception is that Lisp is like some Smalltalk implementations -- so highly sandboxed to prevent this sort of thing. What is the story on this?
My personal perception of Lisp: whereas Java has essentially one mainline JVM (I include the libraries java.*/javax.* in the definition of JVM BTW, so until GNU Classpath is complete there is only one mainline JVM) and a handful of secondary JVMs (just the VM) for specialized environments, Lisp has about half a dozen mainline VMs including both commercial and open-source implementations. Each Lisp VM is cross-platform and in total they probably cover about as many platforms as the one Java VM. Code can be ported between the Lisp implementations without too much hassle. Most Lisp VMs are fast enough that they can actually be used for shell scripting (0.5s startup), and they have the same garbage collection and runtime optimization advantages as Perl, and they hover around 3MB RAM footprint like Perl.
I think the perception of Lisp as slow, RAM-hungry, and hard to interface with non-Lisp was probably true back when machines were small enough that 3MB RAM was enormous. By today's standards, a Lisp VM is quite snappy and can be (relatively) easily integrated into a non-Lisp environment.
I have chosen to learn Lisp because I want to think differently about programming, and see what people mean by "functions are first-class objects". So far it IS different but refreshing.
but aside from that is there a meaningful difference between how Java treats these things as objects vs. the other languages you mention?
You can't get Java to take the code from one method (or anonymous function) and stick it in a different class at runtime.
Am I the only one who knows how to use mplayer -vo sdl/vesafb/fbdev to play full-screen movies from the CLI?
but is there a tool to automate dependency checking, to see which services need to be shut down, to actually shut them down / unload modules, and then relaunch services?
# apt-get dist-upgrade
But almost everyone I talk to who complains about the "architecture" is really complaining about tools, like shells and programs (e.g., PHP, Perl, etc.), and is relatively clueless about the OS architecture (disclosure, I think the modern NT kernel is pretty well-designed, even if the default tools are sub-par.)
.NET, etc.) and data sharing layers (OLE, COM, etc.), and the only philosophy I've seen to it is "make it easy to put data on the screen for the user". Hence spawning a process in Windows is almost as complex as firing up a full-blown GUI application with menubar, dialogs, etc. (yet spawning a thread is as fast as Unix). However, getting the GUI OUT is a harder problem. Did Microsoft finally ship a Windows server that didn't require a graphics card?
/dev, /var, and data directories were read-write, all else was read-only, so that in the event of power failure the system would be able to reach fsck at least. The first such system was put to sea and lasted over 8 months with zero reboots until Katrina knocked it out, all the while continually notifying a hardware watchdog circuit at 0.5 second intervals that it was still alive. The system configuration part took roughly three weeks, including testing, mainly due to IRQ issues with the proprietary hardware (USB, ethernet, PCMCIA, wireless, 12 serial ports, A/D, DIO, hardware watchdog).
I'll try a stab. As others pointed out, the "Unix architecture" is still a debated term, despite various books on the subject. Fundamentally, Unix is the POSIX-like C API (fopen(), brk(), fork(), etc.). However, many Unix purists I know say the tools are also part of the architecture, e.g. sh, passwd, chsh, ls, rm, etc., since the "philosophy" of the architecture is simple processes doing their one thing well and the system making it easy to string the tools together. From this base the other layers can be stacked on -- but they remain optional to the functioning of the computer.
The "Windows architecture" seems to be less philosophical and more pragmatic (in a commercial sense.) We have the WinAPI with layers above it including programming languages (MFC, VB,
I recently had a project that I think illustrates the difference in "philosophies" on a more technical level. I had two "embedded" computers that would be deployed on a sea platform: one was a PC/104 (100MHz 486) running TinyLinux in 32MB RAM and the other was a embedded Pentium 266MHz board running Win2k workstation in 64MB RAM. Yes, apples and oranges, absolutely. The vendor who supplied the Windows-based instrument had never put one of their instruments underwater OR under external automation before; they actually thought adding an entry to the Startup folder would be sufficient to make it all work.
On the Linux PC, I already had an image from the vendor that included a kernel and BusyBox and essential filesystem (16 MB). The system would boot and drop me to a shell on COM2, with ssh server listening on eth0. I compiled separately and added to the system just the utilities I needed: rsync, rsh/rlogin, inetd, Perl, crond, etc. I also recompiled the kernel, shaving 1MB off the memory use. Last, I re-partitioned the system such that only the
The Windows PC was already set up and needed only a "little" software to integrate with the Linux PC. Like I said, the data acquisition software had never been designed for automation. I used Kermit95 to manage the serial port link between the two systems (best $50 we spent by far). However, I had to write programs in C to start the program asynchronously (it would hang Kermit95's remote host command otherwise) AND stop the program by sending a specific series of key presses and mouse clicks (because it couldn't respond to anything else like a signal or status file). I also had to download third-party software to list and kill processes (PsList, PsKill). Testing was a royal pain and getting around the vendor's naivete ultimately cost us another three weeks.
You aske
Could you elaborate on that? As someone else pointed out, DB2 does have savepoints which can bring you back to any state. And (granted this is at the JDBC driver level) I haven't seen Oracle support more than two transaction isolation levels whereas DB2 provides all four mentioned in the JDBC spec. And of course one can see how DB2 is escalating row-level locks so queries can be changed to not step on each other.
I guess my question is: what are you working with such that Oracle does much better than DB2?
If you've got so much time on your hands, I would strongly suggest studying a bit more math/engineering before you're out of high school. I had similar grades as you, started college at 16 (accelerated program), and have returned to grad school in engineering and it is way more intense (fun) than my undergrad in computer science. Thermodynamics, vector calculus, and fluid mechanics are just barely on the edge of comprehension for me, and it's delightful to be challenged at this level.
If I was 15 again, I would get one of the several Dover books on vectors, tensors, and continuum mechanics and go page-by-page through it slowly over the next year.
Don't be ridiculous. China will have absolutely no control of the Internet. As for China being the "next" empire. That is dubious, according to the CIA 15 year report, it will be the EU that will become the next superpower, economically, and militarily.
Is this the same CIA that swore Iraq had WMD? Or the CIA that swore it didn't? Could the White House, which has had a vested interest in downplaying the role of China for the last twenty years, possibly influence the CIA's assessment of China's importance?
Are you sure about that? http://www.hp.com/products1/unix/java/faq/index.ht ml#prod7 :
a geID=4069
Q: Are Java development tools released by HP at the same time they are released by Sun Microsystems?
A: We are a licensee of Java from Sun, and as such, must wait for them to provide code to us for our porting and tuning.
I found some more references to a JVM for embedded spaces called "Chai" or "ChaiVM", and searching HPs website see this: http://devresource.hp.com/forums/thread.jspa?mess
Paul,
The folks who work with the ChaiVM have informed me that it is being obsoleted and is no longer available. Sorry!
Cindy for the OV Interconnect team
I'm curious if there are *any* certified-Java(tm) implementations that don't rely on Sun code for java.* and javax.*. Let me be more specific: certified-Java(tm) J2SE runtime environments at 1.2+ level. There are plenty of JVMs and individual libraries, and GNU Classpath has like 80% coverage, good enough to run OOo on a free software stack, but I haven't heard of any competing implementations actually certified by Sun.
They have even done this for clean-room implementations by rival companies like HP.
Does HP's "clean-room implementation" use licensed code from Sun like java.* and javax.*?
I'll pick this post to respond to.
You seem to have little idea how much stuff costs to someone with no assets, 18 years old, and living in a city with no decent public transportation. They can't just buy some land and stick a trailer on it, or a car, or the insurance. (Do you even know how much legally-required insurance costs in the first 3 years? Try $2000/year, which I should note is 20% of annual income at minimum wage. Full coverage is $3000/year.) They might not even have a computer at all or justification to buy one and learn some skillz on it. In short, stuck at $7/hr and no place to live or way to get between there and work.
Now if you add $10,000 help from parents it's all different. After 3 years, car insurance drops to about $800/year, and with a clunker to get to/from work you can get yourself to the $10/hr job that's across town, and with enough time sure you can eventually get that land and put a trailer on it and start living much better. You can also start growing your own food -- gardening is ridiculously easy, the food is of far higher quality that grocery store, and the exercise will keep you thin and save $30/month on gym fees.
This is mostly how my wife lived. She started working 60 hrs/week when she was 14 years old and had two cars paid for and two parcels of land paid for by the time she was 23. She lived simply, bought most of her food in bulk, and restaurants were rare events. However, she had some medical bills come up that basically wiped out her savings and forced a couple thousand dollars on her credit cards -- and she HAD insurance that paid over 80%. It took both of us a while to get that cleared out.
Me, I went to college, worked five years for *big name corp*, and am back in college for a Masters in a technical discipline. Yeah, we're taking on debt for it, but I've got a passion for it and a long list of professional contacts from career #1 who can vouch for my work ethic. I don't regret the first degree or its debt, since that was the only way out of the shithole town I grew up in. My best friend did your method, skipped the school and got a good job, but after the dot-com meltdown he had to move out of his town because *no one* could find work with all the layoffs. He's back in school and almost finished now and will soon be comfortable again.
The moral of all our stories is this: shit happens, we adapt, but we don't blame the victims about it. Medical costs a lot, cars cost a lot, and someone with lots of patience, a good work ethic, and no materialism will *still* be screwed if they were born broke too.
Good for you that you lived in a town that could hire you with no college degree, and from that lowly start you busted ass and got on up. But your advice in other posts, while nice-sounding, is not possible for many, and reveals a very brittle worldview. "Watch your girlfriend for a dark side"? Huh? I'd hate to be married to someone that controlling.