The 2-party system isn't really written into the U.S. Constitution -- I guess it is something that has evolved over the years. The 2-party system seems limiting in choice (hence the Green Party, the Libertarians, and others who complain about the two main parties), but it has some merits when you consider the alternatives.
One alternative is a one-party system -- we all learned in school in the U.S. on how terrible the Soviet system was that they had only one party, and I grew up in Chicago, which with many other big cities really only had one party. Apart from the anti-Communist propaganda telling us how bad it is, what is does a political party even do when there is only one party? In Chicago, the one party was both a political party as well as a kind of social welfare system: kind of like Hamas.
In Soviet Russia (please, no "in Soviet Russia" jokes), I don't have any direct experience, but as far as I can tell it worked like some kinds of committee structures in an American university. The Party was not the government, it was not the military, it was not industry nor agriculture, and it was not a labor union, but it supervised all of those institutions to make sure that they were run according to "scientific socialist" principles. I imagine the Party was resented by people in government, military, industry, and other places just trying to do their jobs because it was a kind of oversight that a lot of people dedicated to their jobs could do without -- a lot like what takes place in universities.
The multi-party system, however, would have a Liberal Party, a Conservative Party, a Green Party, a Libertarian Party, a Labor Party, a Civil Rights Party, and so on. The problem with a multi-party system is forming a majority government -- think Israel where they have two major parties but they have to suck up to religious parties to form a government.
The two party system means that you stack the deck against minority parties to narrow it down to just two parties so one or other party is sure of having a governing majority. The two parties don't really offer much in the way of choice unless you think Coke and Pepsi represents choice. But on the other hand, the two parties compete for the center of the voting electorate, and the two parties act as critics of each other to expose gross wrong doing. Kerry and Bush are not really that different because they are all part of the same political culture, but they represent themselves as polar opposites to rally their respective core supporters, but they are careful to position themselves for the middle when they govern so having Kerry in office or Bush in office is not going to change all of that much. But that is the goal, to have a stable equalibrium of two parties competing for the center rather than the anarchy and travails of a multi-party democracy.
This notion of inspecting for quality has gotten a bad reputation with all the emphasis on lean manufacturing and emulating the Japanese success and all of that.
While "we will get quality by adding more inspectors" may be a fallacy, the idea of meeting quality goals by culling from a pool of not-so-good quality has a certain history to it. Seymour Cray did it to get parts for his early computers, S. P. Korolev did it to build a space program on a rather shoddy industrial base, and Intel and Apple have been accused of doing this to at one time or another to get their top-end clock speed processors.
While there are some things that you launch because they are high value, some things only become valuable because you put them in orbit. For example, lunch is only a couple of bucks worth of food, but at current rates it costs $5000-$15,000 per meal to bring it to the Space Station. This thing promises to bring lunch down to about $1000.
Now supposes lunch blows up on the pad. Well, the seagulls are going to have to fight over some hamburger fried in rocket fuel. I am thinking a low-reliable low-cost launcher is OK for humping supplies into orbit. On the other hand, the upper stage needs to be reliable because we don't want those things smashing into the Space Station after what happened with Mir.
I heard that Buran (means Snow Storm) was actually the name of that one orbiter and that the "system" never had a public name. It would be like calling the Shuttle "Columbia" after the first of the orbiters. But then the Shuttle never got a name (Shuttle is rather generic and Space Transportation System is rather bureaucratic).
The story told by James Harford (1997), Korelev, Wiley is that the Russians crunched the numbers on the Shuttle being a "reusable" and low-cost "space truck", and they didn't add up, so they figured there was some secret military mission to the thing so they decided they needed one of those things of their own, for whatever it was supposed to be used for when they found out.
This is only one type of error, but Clement Szeperski talks about "up calls" and related errors in inheritance hierarchies of objects.
Your basic structured, modular, procedural program has a fairly predictable flow of control. But what do you do if you have a GUI file list view and some program "underneath" deletes a file. I don't know if there is a simple way to do this in a procedural model, but in an OO system you simply have a call back to notify the file list view that the available files have changed and the file list view updates itself.
Now the call back could be implemented as the override of a virtual method to the base class, or it could be a Java style action listener, or it could be a call to a function referenced by a function pointer in a table as with Gnome. In any event, you start to get interesting side effects and interactions when you do this. Anytime you invoke a method of an object that calls out to anything, there is no way to insure that it doesn't call back in to that object to a method that changes the state of the object out from under you as it will.
OO solves one problem, that is provides elegant expressions of ideas that would be a rats nest of procedure hierarchies, state variables, and polling loops, but introduces another problem: "Hey, where did that method call come from?"
Yeah, yeah, we can talk about all the limitations: the indent-level delimiting is OK until your editor mixes spaces and tabs in the indents, reference counting plus cycle removal instead of a "real" garbage collector, no value types.
I am intrigued with the idea that Lisp was the answer to everything -- strong typing but dynamic typing, eval-loop command line, and all of that. Now Lisp took a hit in the aftermath of the Lisp machine days, but with computers and Moore's law and advances in compiler and language technology, maybe it is time for the idea to come back.
Lisp, they say (Paul Graham, et al), was the victim of worse is better, and while a SUN workstation, Unix, and C were no where as "elegant" as Lisp and a Lisp-M, they were a heck of a lot cheaper, and people felt they could solve much the same problems with that setup and more. Now saying C is a Lisp replacement is a stretch, but saying Java is a Lisp replacement is getting a little closer (GC, reflection).
How about Python as a worse-is-better substitute for Lisp? Even Paul Graham admits it is getting close.
Python is Free as in Speech, Free as in Beer, a unified language with no spun-off dialects. Java has SUN controlling it, C# has Microsoft owning it.
If Python is to become the heir to Lisp, it is going to need arrays of value types, compilers (yeah, yeah, I know about various array classes and Psyco, but they are works in varying degrees of progress). But speaking as a Delphi programmer who has done some work in C++, Java, and C#, I think Python has potential and is something I am going to be looking at as it evolves.
I agree with you 100% on the matters surrounding flying the Shuttle again.
There is a matter regarding developing a robot to service the Hubble. Part of the party line on the manned space program is the limitation of robots. While the Hubble optics error was a dark moment for NASA, the Hubble repair mission with a lot of complex, long space walks was a "jewel in the crown" of the manned space program, and if they can get a robot to do the repair, it would kind of puncture that concept.
On the other hand, given both the cost and risk of manned space flight, oh forget the risk because people are willing to take risks, simply the cost, I am beginning to agree with Freeman Dyson that instead of one complex, man-servicable Hubble, NASA should have built a series of throw-away Hubbles with specific missions launched on expendable rockets. What is a Shuttle flight, anywhere from 300 mil to a cool billion depending on accounting? Couldn't you build a whole new telescope and launch it for that money?
I am of the mind that asking technical questions are fair interview questions -- I mean what are you supposed to ask in an interview for a technical job?
On the other hand, most of us mortals don't store all the details of API's in our heads. Back in the Stone Age we used manuals and in the Information Age we use the SUN Web site. If your interview objective is to see how someone would get the information to solve the problem, that is fine, but if your objective is to see if that person already has some narrow set of information, you are going to exclude some capable people.
I am mainly a Delphi developer (I should say a Delphi component developer), and my Java experience is only 4 months old, and gee, my Java experience is limited to using JNI to allow a Delphi ActiveX component to invoke an extension module written in Java and using a class loader so that extension module can be reloaded while the ActiveX component is still running.
I don't know the answer to your question about Java collection objects without looking it up, although I have enough sense to know that you have to use Object wrappers for value types in collections and then have to cast those objects back to their original types when you pull Object references out of collections -- I know that from "wasting" time reading Slashdot.
The Japanese attacked Pearl Harbor, the treated POW's with cruelty, but the one characteristic relevant to the current discussion is the use of suicide tactics (the Kamikaze or "divine wind" as the typhoon that swept away the foreign invader).
Besides the horrific casualties the Kamikaze was inflicting on U.S. ships, at a stage where the war was for the most part lost for Japan, the notion that surrender was dishonorable, that suicide bombers would be used in large scale to defend the homeland suggested the normal pattern of surrender in the face of military defeat was not going to work.
The Trade Center bombing had both elements of suicide attack as well as mass casualties -- they were on the Pearl Harbor level, but they could have been on the Nagasaki level. My own Near Eastern paranoid-style thinking lead me to believe that taking President Bush to Omaha was no accident -- it was to signal that the attack rose to the level that could provoke a nuclear response. The only time we were provoked to use atomic weapons in the past was in response to suicide tactics.
The terrorists in Madrid blew up several train cars (fact), the Aznar government suggested the type of explosives implicated ETA (fact -- not that ETA was involved but that the government was saying that), the voters in Spain voted out Aznar's successor and voted in Mr. Shoemaker (Zapetero -- fact), who promptly pledged to make good on his campaign promise to pull Spanish soldiers out of Iraq (fact), and the terrorists drew some conclusions from all of this (not factual, largely speculation, but there are communications pointing to their intentions).
I am trying to parse your words carefully. Your concern is with "Bush supporters." You didn't actually say that Bush supporters are liars, so unlike Senator Kerry, you do have an out, but you did say that these Bush supporters made "an accusation... inconsistent with the facts" and these same Bush supporters "never thought twice about lying in the past", which suggests that as a class, Bush supporters are habitual liars and are lying about the Spanish election results. Didn't actually accuse someone of being a Communist but described the habits and customs of water birds.
The terrorists out there keep their own counsel, so any discussion of how they interpret the Spanish election on my part is not in the realm of fact, it is speculation. And how can you accuse someone of lying when they are engaged in speculation and theorizing and being open about it? You have an opinion on what the Spanish election means, I have an opinion on that same topic, and if we differ, that means one of us is lying?
I gather you see something illegitimate in saying "vote for Kennedy, he will be tough on Communism" because by implication Nixon would be soft on Communism? Who was tougher on Communism was an issue in 1960, and the whole business about the Missile Gap (which proved to be non-existent from U-2 and Corona/Discoverer intelligence but Eisenhower couldn't talk about it but showed to Senator Kennedy) was arguing that the Democrats were going to be tougher on Communism than Eisenhower, who wasn't properly minding the store.
Or by your reasoning, you should be mad at the Democratic party for appeasing the sentiment that being against the Iraq war is appeasing the terrorists. The Democratic Party primary voters largely rejected Howard Dean and turned to John Kerry, who has "street cred" on matters of war as a war veteran and has positioned himself as critical of the handling of the war without being in favor of abandoning the war.
One part of byte code is portability. The second part is security -- you know, the original applet concept of running programs downloaded off the Web in a secure sandbox.
I think that both Java and C# are influenced by the idea that byte codes can be better sandboxed and given controlled level of privileges.
I am beginning to think that what brings Java to its knees is class loading.
Unlike Python or other scripting languages, every time you run a Java program it has to load each class that first time you create an object instance, and it does some kind of analysis on the byte codes of that class to see that it is OK and not kind of a Trojan Horse program. The idea is that Java is supposed to be secure by doing all that checking.
That checking comes at a cost that Java programs are pokey to load. Think of it this way -- Windows is pokey to load, but you load it once when you start your computer. Java is a kind of OS within an OS, and it is pokey to load every time you run an app.
I heard that the Mac has better Java on it, and they keep one image of the JVM around so once you load a class it stays loaded for the entire OS session instead of loading every time you say java myprog.
The other thing that probably brings Java to its knees is Swing on account that they tried to write Swing more object-oriented than thou and more document-view than thou and that you have gobs of little classes -- both a class loading startup overhead as well as overhead of creating and GC'ing object instances.
While people talk about "no shrink-wrap" Java apps, not only is Matlab very Java friendly, Matlab is partly implemented in Java. Matlab is a special case in that it is one of the few commercial shrink-wrap apps where having Windows/Unix/Linux versions is an important part of their business model, and I think they are migrating Matlab piecemeal over to Java internally. They say the GUI is done in Java -- if you go Files Open to open an editor window for a Matlab script file (.M file), it takes about 3 seconds to load, even on a late model P4. So you can say the MathWorks guys don't know how to program Swing correctly, or you could say Swing has some serious pokeyness in terms of loading and creating stuff.
But apart from class loading and Swing, I am not finding much that is pokey about Java.
I have tried native integration in both Java and CLR, and I argue that each has their merits.
CLR on the surface may be more "seemless", but a lot of it depends on decorating C# function prototypes of attribute information, which is documented in Microsoft circular-hyper-link fashion (each doc entry is "go link to here" which tells you "go link to there" and you end up where you started without learning anything). If the attributes don't do what you want you can go nuts.
Java JNI on the surface is a lot more wordy and clunky, but you program the interface by writing code rather than depend on some square-bracket attributes to do the automagic thing. There is a lot you have to read to get it to work, but SUN has a document that explains the whole thing with chapter, paragraph, and verse rather than hyper link Hell. Yeah, it would be nice if you didn't have to write some much code, especially on the C/C++ side, but it does the job, and since you are writing code to control it, it seems it is more programmable.
I have taken to liking the JNI that much (I am more a Delphi/C++ programmer and am just starting on Java) that I was thinking of migrating apps salami style (one cut at a time) by implementing all the non-GUI parts in Java and calling those parts from a slimmed-down native GUI part -- I might port the native GUI part to different OS's or might eventually port that to Java when I get a handle on Java GUI stuff.
But guess what, I would be taking a functioning Windows app and turn it into something that requires the JVM, and it would have been nicer in an Alternate Universe where the JVM came with Windows (heck, it would be nice if the CLR came with Windows instead of requiring an install).
Anyway, people think of program written in Java calling some low level GUI natives (like with SWT). My thinking is to write the GUI parts native but put the "business logic" in Java -- the JNI goes in both directions and lets you do that. Of course SUN's idea is Java Everywhere for Every Stupid Last Thing, while MS's idea is Java, huh?
What I am also trying to say is that instead of putting effort into things like Mono, which is tied to Microsoft and all that entails, why not put the same effort into advancing Python?
Python doesn't have the corporate fetters of either C# or Java, and while it is nowhere near the execution speed (especially for numeric work), it seems to have growth potential, especially if array data types could be standardized.
I apologize -- I came across as whining. Allow me to tell you what I am trying to do.
I am interested in teaching Electrical Engineering students how to write software to implement signal processing algorithms. I want them actually writing code rather than running some canned engineering package that designs all the filters for them and shows pretty pictures.
I wrote a canned program that shows pretty pictures of signals run through filters with the results showing up on spectrum plots where the students could write extension modules in C++ to implement their filters. A typical extension modules is derived from a base class that buffers the signal input and output from the filter, a filter is typically written by overriding a couple of virtual methods -- one to calculate the filter coefficients and another to apply those filter coefficients to the next buffer of data.
The students have been told that it is valuable for an engineer to have some C++ experience, but even this sheltered, structured use of C++ makes it hard to get students to take my course. C++ has a clunky edit-compile-test cycle, it is a lot of effort to get a clean compile, and then more often than not a student written C++ program crashes and is hard to debug because it is an embedded object in another program (yeah, yeah, I have a test framework).
The engineering students regard C++ as cod liver oil -- they would all rather use Matlab. They use Matlab in all their other courses. Back in the days engineers would struggle to get a FORTRAN program to compile and then struggle to get it debugged. No one wants to bother with implementing engineering calculations at that level, it is all Matlab now. Matlab has an "eval" loop style of programming, good numeric libraries, good plot libraries, reasonable error message feedback, excellent self-discovery and help capability.
I am not a Stallmanesque zealot and a don't begrudge MathWorks making a profit, but in the olden days interesting software came out of the academy (P-system, BSD, Lisp) and these days we seem to be a trade school for the preferred commercial packages. Universities are supposed to be at the leading edge. Why aren't we using Lisp if we wanting something interactive and developing engineering packages for it (there is a thermodyamic modelling package written in Lisp called CyclePad by the way)?
Outfits like the National Labs (Sandia, Livermore) and the Space Telescope Science Institute are doing what the universities ought to be doing -- they are looking into Python as the next Lisp. It is a good effort and I look forward to it bearing fruit.
All I am trying to say is that if Python is going to become an "infix Lisp" for science and engineering, there are a number of things (such as NumArray) that are a work in progress but not fully there yet.
Using Python as glue and doing the heavy lifting in C/C++ extension modules is OK, but it doesn't solve the following problem. NumArray over here has this neat Python class implemented in C++ for efficiency, and there is this plot package over there that doesn't know about NumArray, and how am I going to get the data from here to over there without diving into C/C++ sources of third party packages to extend them to know about NumArray?
There is a lot of interest in using Python as a numerical, infix Lisp, but there needs to be more interoperability of these new data types. I am hoping this will come with time, but it is not quite there yet.
I am starting to get sold on Python as an alternative to Java, C#, VB, and whatever as a way of developing GUI apps. I see at as the best attempt at bringing a Lisp-like setup (functions assignable to variables, dynamic typing) and I suppose I can live with the warts (white-space delimited -- it makes it readable, but it stumbles when you switch editors with different ideas on tabs and spaces).
My concern is that Python is still a work in progress as a Java replacement. Notable lacking is an array type (list of elements of all the same type making it more efficient than the list-of-objects approach in both memory and speed). How is a person supposed to do graphics by computing rasters without a good array type? I believe Lisp was originally everything-is-a-tree-data-structure, but they must have added proper array types in the Lisp machine days.
But what about Numeric? Well yes, you can use C extensions to define whatever kind of Python container object to serve as a proper array, but that container object becomes this atomistic entity that requires copying all the elements in and out to Python lists to interface with just about everything else. And Numeric is a work-in-progress because it is being replaces with NumArray which is in some state of beta test. And are these things standardized enough that you can say (like in Matlab), here is an array of pixels I have created, plot them as a bitmap image?
Software is a commodity in the way a gallon of gas is a commodity -- if to run my car I had to pump in exactly 10 gallons of gas, and then I had to add exactly 10 ounces of a detergent additive, and there were several incompatible brands of detergent additive on the market, and everytime I wanted to fill the tank I had to read labels and figure out if I had the right kind.
I am thinking that OSS is a commodity in the sense that computer hardware was a commodity in the pre-PC days of CP-M and the S-100 bus. Jerry Pournelle had this mantra "iron is expensive, silicon is cheap" that a person would "invest" in a boat anchor cabinet, a good power supply, and an S-100 backplane and then plug in boards with memory, processors, and peripherals and upgrade the silicon while keeping the boat anchor iron.
This putting a computer together from parts required someone who knew what they were doing, and Pournelle plugged the idea of "system integrators", dudes who would in essence sell you generic hardware, but in their markup they were selling you a service of knowing what hardware was compatible with what and where to get the drivers for everything.
For all the talk of Linux weenies and lusers, the average Linux distro really is not an end user product, but it seems that the Linux savy could pull together pieces parts of software and put together systems tailored to the requirements of specific customers. You know, open and free software, but the money is to be made in providing services, and the service is being the propeller head who knows what software is out there and what works with what, and what configuration tweaks will make a customer happy.
The PC kind of changed Pournelle's model. The silicon was cheap, but the iron (cabinet, motherboard, power supply) started coming from Free China and later from Not-Quite-As-Free China and it became cheap, and with the business model of Dell, it is pretty much cheaper just to replace the whole system than poke around with doing your own upgrades.
As far as the software, the software has kind of moved away from this mix and match model as well. Sure, a Windows install may be as hard as some Linux distro installs, but who even installs software -- you buy the computer with Windows, Office, and networking already installed from Dell. So I guess the system integration has become a mass market instead of a cottage industry.
I am thinking that for Linux to catch on, there has to be some patron, some "angel", some big player to do the system integration and sell ready-to-run systems to the mass market. Is it Wal-Mart and Lindows? Is it SUN and their "Java workstation?"
I think the way it played out is that Kirk and Spock figured out that Joan Collins had to get flattened for America to get in the war to fight Hitler, and they used the Time Portal to get McCoy out of the picture so this could happen, but then Kirk had somehow taken a fancy to Joan Collins in all this but had to stand and watch while all this takes place, so Roddenberry didn't completely ruin the notion of Kirk having to suffer for love.
My favorite part of this is Spock calling vacuum tube electronics "bear skins and stone knives." I saw the episode as a kid and didn't understand what he was trying to do. Much later I figured that the tricorder lacked a video display, but they needed to see the fast forward images in the Portal to know where McCoy was, and Spock was using 1930's technology to build a video monitor to retrieve the readout from the tricorder of the Portal images. So the tricorder lacked a flat panel display. Sheesh!
The baseball bat story sounds like an urban legend. But then Dick Wolf's "Law and Order" had a good yarn about cybercrime.
This young white dude, computer programmer family man is found shot dead. In his house. With an exotic WW-II German pistol.
The crack team of Briscoe and Green do their leg work, and they come across an old black dude who is somehow connected to the young white dude in a financial scam. The cops think "salt and pepper" team and one crime guy turning on his partner.
The D.A.'s office goes to work on this and what finally emerges is 1) the black dude is an upstanding citizen and a war hero (hence the WW-II war trophy gun), 2) the white dude lost his job and turned to cybercrime to support his family, 3) the black dude had no connection to the white dude apart from that the white dude picked a random victim to scam for identity theft, 4) the black dude, finding all of his net worth was scammed, used his charm and his war hero "street smarts" to get the ISP to give up the address of the guy who ripped him off.
Like many Law and Order episodes, the whole thing is Hollywood and little connection to real cops, victims, and criminals. On the other hand, the writers are not that creative and probably use some news stories as a jumping off point.
Like the better of the Law and Order episodes, it makes you think. What right did the white dude have to pick a random victim and take all his money? What right did the black dude have to get his revenge as murder? What right does society have to put an 80 year old man (the black dude) away for seeking his own justice?
But to me the moral of this episode along with the baseball bat incident is what goes around comes around. No, I don't think threatening a 15 year old with a baseball bat is a good idea, and the 15 year old and parents can get in a lather about their "rights." But a 15 year old with access to a computer (and the parents of said 14 year old) could appreciate that if one could hack into someone's server, owner of such server could hack back to track you down, and such owner could be meaner, tougher, better connected, or in a much worse mood than you. And somehow going after people who threaten 15 year olds with bats under those circumstances doesn't seem like the highest priority for the courts.
Didn't he write that Star Trek episode where DeForest Kelly saves Joan Collins from getting flattened by a street car, and because she is an influential peace activist, Sadda^H^H^H^H^HHitler takes over the world and the Enterprise vanishes because the Federation never happens, Spock bitches and moans about 1930's tech in building a video monitor to read images from his tricorder (I guess the tricorder doesn't have a display -- and Spock is awful testy for someone who doesn't have emotion).
I heard that Ellison's original script had some low-level Enterprise crew doing drugs which leads to this time-travel incident, but the Hollywood suits took that out because the Enterprise crew are all Eagle Scouts and that would never happen in the Star Trek universe, so they rewrote the script to have McCoy accidently inject himself with a dangerous (legit) medication when the Enterprise hits a space-time "air pocket" (you know, one of those things that tosses everyone out of the seats that they are not belted into).
If you write scripts for Hollywood, having someone else do a rewrite for any of a number of pandering reasons is part of the landscape, and as far as making the Enterprise crew Boy Scouts who would never use drugs, whoever is supervising a series has to keep the episodes consistent with a vision or else it turns into Superman Comics where Superman keeps getting more powers in each episode that they have to hit some kind of reset button.
Getting back on topic, Ellison may not be a nut case, but he has a track record of pissing into the wind on matters of principle that turn out to be no-win situations.
Lets see. You put some tritium in the center of a levitated-core composite HEU/Pu implosion device to get a little extra shot of neutrons to "boost" the yield -- most of the yield is from fission, but the fusion gives it a few extra neutrons to get it going before it flies apart. Of course the yield of such a fission implosion device is limited so you channel the flash of x-rays on to some Styrofoam to implode this large tube of lithium deuteride.
Of course that doesn't do too much. So you put a plutonium rod in the middle as a "spark plug." That gets some fusion going from the flood of neutrons, but most of what that does is just give more neutrons which would otherwise fly off and not contribute to the explosive force unless you surround the works with U-238, which fissions from the flood of neutrons.
And people call this a fusion device. This is relevant to the discussion of table-top fusion because even uncontrolled fusion is hard to get -- the original Teller idea of the "Super" which would be just sticking an A-bomb next to a tub of deuterium simply doesn't work, and it is a Good Thing because otherwise A-bomb explosions could start a fusion chain reaction in the Earth.
Multiplatform unfortunately has a lowest-common-denominator effect. Think of it -- if all the platforms all had the same features but were just called a little differently, they would all be in reality the same platform.
The Audacity spectrogram (I suppose it is just an add on and not the primary focus) is a dithered multi-color mess that scrolls like molasses.
Windows offers 1) hardware assisted scroll with ScrollWindowEx, 2) vertical retrace synch, two features that go a long way to making a good scrolling waveform display/spectrum analyser that I don't see in and of the platform portable tool kits. Those features, by the way, were added to Windows when 2-D scrolling games were a big thing (ScrollWindowEx I believe was part of WinG and vertical synch is part of IDirectDraw). Windows is a little weak in the latency department, but running everything on really fast processors is kind of taking care of that.
What is out there that is an improvement on Windows in this regard? I heard BeOS had very low latencies (I had heard the same about OS/2), but both those systems are pretty much past history if you want to reach a lot of people.
Slashdot Mirror
One alternative is a one-party system -- we all learned in school in the U.S. on how terrible the Soviet system was that they had only one party, and I grew up in Chicago, which with many other big cities really only had one party. Apart from the anti-Communist propaganda telling us how bad it is, what is does a political party even do when there is only one party? In Chicago, the one party was both a political party as well as a kind of social welfare system: kind of like Hamas.
In Soviet Russia (please, no "in Soviet Russia" jokes), I don't have any direct experience, but as far as I can tell it worked like some kinds of committee structures in an American university. The Party was not the government, it was not the military, it was not industry nor agriculture, and it was not a labor union, but it supervised all of those institutions to make sure that they were run according to "scientific socialist" principles. I imagine the Party was resented by people in government, military, industry, and other places just trying to do their jobs because it was a kind of oversight that a lot of people dedicated to their jobs could do without -- a lot like what takes place in universities.
The multi-party system, however, would have a Liberal Party, a Conservative Party, a Green Party, a Libertarian Party, a Labor Party, a Civil Rights Party, and so on. The problem with a multi-party system is forming a majority government -- think Israel where they have two major parties but they have to suck up to religious parties to form a government.
The two party system means that you stack the deck against minority parties to narrow it down to just two parties so one or other party is sure of having a governing majority. The two parties don't really offer much in the way of choice unless you think Coke and Pepsi represents choice. But on the other hand, the two parties compete for the center of the voting electorate, and the two parties act as critics of each other to expose gross wrong doing. Kerry and Bush are not really that different because they are all part of the same political culture, but they represent themselves as polar opposites to rally their respective core supporters, but they are careful to position themselves for the middle when they govern so having Kerry in office or Bush in office is not going to change all of that much. But that is the goal, to have a stable equalibrium of two parties competing for the center rather than the anarchy and travails of a multi-party democracy.
While "we will get quality by adding more inspectors" may be a fallacy, the idea of meeting quality goals by culling from a pool of not-so-good quality has a certain history to it. Seymour Cray did it to get parts for his early computers, S. P. Korolev did it to build a space program on a rather shoddy industrial base, and Intel and Apple have been accused of doing this to at one time or another to get their top-end clock speed processors.
Now supposes lunch blows up on the pad. Well, the seagulls are going to have to fight over some hamburger fried in rocket fuel. I am thinking a low-reliable low-cost launcher is OK for humping supplies into orbit. On the other hand, the upper stage needs to be reliable because we don't want those things smashing into the Space Station after what happened with Mir.
The story told by James Harford (1997), Korelev, Wiley is that the Russians crunched the numbers on the Shuttle being a "reusable" and low-cost "space truck", and they didn't add up, so they figured there was some secret military mission to the thing so they decided they needed one of those things of their own, for whatever it was supposed to be used for when they found out.
Your basic structured, modular, procedural program has a fairly predictable flow of control. But what do you do if you have a GUI file list view and some program "underneath" deletes a file. I don't know if there is a simple way to do this in a procedural model, but in an OO system you simply have a call back to notify the file list view that the available files have changed and the file list view updates itself.
Now the call back could be implemented as the override of a virtual method to the base class, or it could be a Java style action listener, or it could be a call to a function referenced by a function pointer in a table as with Gnome. In any event, you start to get interesting side effects and interactions when you do this. Anytime you invoke a method of an object that calls out to anything, there is no way to insure that it doesn't call back in to that object to a method that changes the state of the object out from under you as it will.
OO solves one problem, that is provides elegant expressions of ideas that would be a rats nest of procedure hierarchies, state variables, and polling loops, but introduces another problem: "Hey, where did that method call come from?"
I am intrigued with the idea that Lisp was the answer to everything -- strong typing but dynamic typing, eval-loop command line, and all of that. Now Lisp took a hit in the aftermath of the Lisp machine days, but with computers and Moore's law and advances in compiler and language technology, maybe it is time for the idea to come back.
Lisp, they say (Paul Graham, et al), was the victim of worse is better, and while a SUN workstation, Unix, and C were no where as "elegant" as Lisp and a Lisp-M, they were a heck of a lot cheaper, and people felt they could solve much the same problems with that setup and more. Now saying C is a Lisp replacement is a stretch, but saying Java is a Lisp replacement is getting a little closer (GC, reflection).
How about Python as a worse-is-better substitute for Lisp? Even Paul Graham admits it is getting close.
Python is Free as in Speech, Free as in Beer, a unified language with no spun-off dialects. Java has SUN controlling it, C# has Microsoft owning it.
If Python is to become the heir to Lisp, it is going to need arrays of value types, compilers (yeah, yeah, I know about various array classes and Psyco, but they are works in varying degrees of progress). But speaking as a Delphi programmer who has done some work in C++, Java, and C#, I think Python has potential and is something I am going to be looking at as it evolves.
There is a matter regarding developing a robot to service the Hubble. Part of the party line on the manned space program is the limitation of robots. While the Hubble optics error was a dark moment for NASA, the Hubble repair mission with a lot of complex, long space walks was a "jewel in the crown" of the manned space program, and if they can get a robot to do the repair, it would kind of puncture that concept.
On the other hand, given both the cost and risk of manned space flight, oh forget the risk because people are willing to take risks, simply the cost, I am beginning to agree with Freeman Dyson that instead of one complex, man-servicable Hubble, NASA should have built a series of throw-away Hubbles with specific missions launched on expendable rockets. What is a Shuttle flight, anywhere from 300 mil to a cool billion depending on accounting? Couldn't you build a whole new telescope and launch it for that money?
On the other hand, most of us mortals don't store all the details of API's in our heads. Back in the Stone Age we used manuals and in the Information Age we use the SUN Web site. If your interview objective is to see how someone would get the information to solve the problem, that is fine, but if your objective is to see if that person already has some narrow set of information, you are going to exclude some capable people.
I am mainly a Delphi developer (I should say a Delphi component developer), and my Java experience is only 4 months old, and gee, my Java experience is limited to using JNI to allow a Delphi ActiveX component to invoke an extension module written in Java and using a class loader so that extension module can be reloaded while the ActiveX component is still running.
I don't know the answer to your question about Java collection objects without looking it up, although I have enough sense to know that you have to use Object wrappers for value types in collections and then have to cast those objects back to their original types when you pull Object references out of collections -- I know that from "wasting" time reading Slashdot.
I guess I would fail your interview.
Besides the horrific casualties the Kamikaze was inflicting on U.S. ships, at a stage where the war was for the most part lost for Japan, the notion that surrender was dishonorable, that suicide bombers would be used in large scale to defend the homeland suggested the normal pattern of surrender in the face of military defeat was not going to work.
The Trade Center bombing had both elements of suicide attack as well as mass casualties -- they were on the Pearl Harbor level, but they could have been on the Nagasaki level. My own Near Eastern paranoid-style thinking lead me to believe that taking President Bush to Omaha was no accident -- it was to signal that the attack rose to the level that could provoke a nuclear response. The only time we were provoked to use atomic weapons in the past was in response to suicide tactics.
I am trying to parse your words carefully. Your concern is with "Bush supporters." You didn't actually say that Bush supporters are liars, so unlike Senator Kerry, you do have an out, but you did say that these Bush supporters made "an accusation ... inconsistent with the facts" and these same Bush supporters "never thought twice about lying in the past", which suggests that as a class, Bush supporters are habitual liars and are lying about the Spanish election results. Didn't actually accuse someone of being a Communist but described the habits and customs of water birds.
The terrorists out there keep their own counsel, so any discussion of how they interpret the Spanish election on my part is not in the realm of fact, it is speculation. And how can you accuse someone of lying when they are engaged in speculation and theorizing and being open about it? You have an opinion on what the Spanish election means, I have an opinion on that same topic, and if we differ, that means one of us is lying?
I gather you see something illegitimate in saying "vote for Kennedy, he will be tough on Communism" because by implication Nixon would be soft on Communism? Who was tougher on Communism was an issue in 1960, and the whole business about the Missile Gap (which proved to be non-existent from U-2 and Corona/Discoverer intelligence but Eisenhower couldn't talk about it but showed to Senator Kennedy) was arguing that the Democrats were going to be tougher on Communism than Eisenhower, who wasn't properly minding the store.
Or by your reasoning, you should be mad at the Democratic party for appeasing the sentiment that being against the Iraq war is appeasing the terrorists. The Democratic Party primary voters largely rejected Howard Dean and turned to John Kerry, who has "street cred" on matters of war as a war veteran and has positioned himself as critical of the handling of the war without being in favor of abandoning the war.
I think that both Java and C# are influenced by the idea that byte codes can be better sandboxed and given controlled level of privileges.
Unlike Python or other scripting languages, every time you run a Java program it has to load each class that first time you create an object instance, and it does some kind of analysis on the byte codes of that class to see that it is OK and not kind of a Trojan Horse program. The idea is that Java is supposed to be secure by doing all that checking.
That checking comes at a cost that Java programs are pokey to load. Think of it this way -- Windows is pokey to load, but you load it once when you start your computer. Java is a kind of OS within an OS, and it is pokey to load every time you run an app.
I heard that the Mac has better Java on it, and they keep one image of the JVM around so once you load a class it stays loaded for the entire OS session instead of loading every time you say java myprog.
The other thing that probably brings Java to its knees is Swing on account that they tried to write Swing more object-oriented than thou and more document-view than thou and that you have gobs of little classes -- both a class loading startup overhead as well as overhead of creating and GC'ing object instances.
While people talk about "no shrink-wrap" Java apps, not only is Matlab very Java friendly, Matlab is partly implemented in Java. Matlab is a special case in that it is one of the few commercial shrink-wrap apps where having Windows/Unix/Linux versions is an important part of their business model, and I think they are migrating Matlab piecemeal over to Java internally. They say the GUI is done in Java -- if you go Files Open to open an editor window for a Matlab script file (.M file), it takes about 3 seconds to load, even on a late model P4. So you can say the MathWorks guys don't know how to program Swing correctly, or you could say Swing has some serious pokeyness in terms of loading and creating stuff.
But apart from class loading and Swing, I am not finding much that is pokey about Java.
CLR on the surface may be more "seemless", but a lot of it depends on decorating C# function prototypes of attribute information, which is documented in Microsoft circular-hyper-link fashion (each doc entry is "go link to here" which tells you "go link to there" and you end up where you started without learning anything). If the attributes don't do what you want you can go nuts.
Java JNI on the surface is a lot more wordy and clunky, but you program the interface by writing code rather than depend on some square-bracket attributes to do the automagic thing. There is a lot you have to read to get it to work, but SUN has a document that explains the whole thing with chapter, paragraph, and verse rather than hyper link Hell. Yeah, it would be nice if you didn't have to write some much code, especially on the C/C++ side, but it does the job, and since you are writing code to control it, it seems it is more programmable.
I have taken to liking the JNI that much (I am more a Delphi/C++ programmer and am just starting on Java) that I was thinking of migrating apps salami style (one cut at a time) by implementing all the non-GUI parts in Java and calling those parts from a slimmed-down native GUI part -- I might port the native GUI part to different OS's or might eventually port that to Java when I get a handle on Java GUI stuff.
But guess what, I would be taking a functioning Windows app and turn it into something that requires the JVM, and it would have been nicer in an Alternate Universe where the JVM came with Windows (heck, it would be nice if the CLR came with Windows instead of requiring an install).
Anyway, people think of program written in Java calling some low level GUI natives (like with SWT). My thinking is to write the GUI parts native but put the "business logic" in Java -- the JNI goes in both directions and lets you do that. Of course SUN's idea is Java Everywhere for Every Stupid Last Thing, while MS's idea is Java, huh?
Python doesn't have the corporate fetters of either C# or Java, and while it is nowhere near the execution speed (especially for numeric work), it seems to have growth potential, especially if array data types could be standardized.
I am interested in teaching Electrical Engineering students how to write software to implement signal processing algorithms. I want them actually writing code rather than running some canned engineering package that designs all the filters for them and shows pretty pictures.
I wrote a canned program that shows pretty pictures of signals run through filters with the results showing up on spectrum plots where the students could write extension modules in C++ to implement their filters. A typical extension modules is derived from a base class that buffers the signal input and output from the filter, a filter is typically written by overriding a couple of virtual methods -- one to calculate the filter coefficients and another to apply those filter coefficients to the next buffer of data.
The students have been told that it is valuable for an engineer to have some C++ experience, but even this sheltered, structured use of C++ makes it hard to get students to take my course. C++ has a clunky edit-compile-test cycle, it is a lot of effort to get a clean compile, and then more often than not a student written C++ program crashes and is hard to debug because it is an embedded object in another program (yeah, yeah, I have a test framework).
The engineering students regard C++ as cod liver oil -- they would all rather use Matlab. They use Matlab in all their other courses. Back in the days engineers would struggle to get a FORTRAN program to compile and then struggle to get it debugged. No one wants to bother with implementing engineering calculations at that level, it is all Matlab now. Matlab has an "eval" loop style of programming, good numeric libraries, good plot libraries, reasonable error message feedback, excellent self-discovery and help capability.
I am not a Stallmanesque zealot and a don't begrudge MathWorks making a profit, but in the olden days interesting software came out of the academy (P-system, BSD, Lisp) and these days we seem to be a trade school for the preferred commercial packages. Universities are supposed to be at the leading edge. Why aren't we using Lisp if we wanting something interactive and developing engineering packages for it (there is a thermodyamic modelling package written in Lisp called CyclePad by the way)?
Outfits like the National Labs (Sandia, Livermore) and the Space Telescope Science Institute are doing what the universities ought to be doing -- they are looking into Python as the next Lisp. It is a good effort and I look forward to it bearing fruit.
All I am trying to say is that if Python is going to become an "infix Lisp" for science and engineering, there are a number of things (such as NumArray) that are a work in progress but not fully there yet.
Using Python as glue and doing the heavy lifting in C/C++ extension modules is OK, but it doesn't solve the following problem. NumArray over here has this neat Python class implemented in C++ for efficiency, and there is this plot package over there that doesn't know about NumArray, and how am I going to get the data from here to over there without diving into C/C++ sources of third party packages to extend them to know about NumArray?
There is a lot of interest in using Python as a numerical, infix Lisp, but there needs to be more interoperability of these new data types. I am hoping this will come with time, but it is not quite there yet.
My concern is that Python is still a work in progress as a Java replacement. Notable lacking is an array type (list of elements of all the same type making it more efficient than the list-of-objects approach in both memory and speed). How is a person supposed to do graphics by computing rasters without a good array type? I believe Lisp was originally everything-is-a-tree-data-structure, but they must have added proper array types in the Lisp machine days.
But what about Numeric? Well yes, you can use C extensions to define whatever kind of Python container object to serve as a proper array, but that container object becomes this atomistic entity that requires copying all the elements in and out to Python lists to interface with just about everything else. And Numeric is a work-in-progress because it is being replaces with NumArray which is in some state of beta test. And are these things standardized enough that you can say (like in Matlab), here is an array of pixels I have created, plot them as a bitmap image?
I am thinking that OSS is a commodity in the sense that computer hardware was a commodity in the pre-PC days of CP-M and the S-100 bus. Jerry Pournelle had this mantra "iron is expensive, silicon is cheap" that a person would "invest" in a boat anchor cabinet, a good power supply, and an S-100 backplane and then plug in boards with memory, processors, and peripherals and upgrade the silicon while keeping the boat anchor iron.
This putting a computer together from parts required someone who knew what they were doing, and Pournelle plugged the idea of "system integrators", dudes who would in essence sell you generic hardware, but in their markup they were selling you a service of knowing what hardware was compatible with what and where to get the drivers for everything.
For all the talk of Linux weenies and lusers, the average Linux distro really is not an end user product, but it seems that the Linux savy could pull together pieces parts of software and put together systems tailored to the requirements of specific customers. You know, open and free software, but the money is to be made in providing services, and the service is being the propeller head who knows what software is out there and what works with what, and what configuration tweaks will make a customer happy.
The PC kind of changed Pournelle's model. The silicon was cheap, but the iron (cabinet, motherboard, power supply) started coming from Free China and later from Not-Quite-As-Free China and it became cheap, and with the business model of Dell, it is pretty much cheaper just to replace the whole system than poke around with doing your own upgrades.
As far as the software, the software has kind of moved away from this mix and match model as well. Sure, a Windows install may be as hard as some Linux distro installs, but who even installs software -- you buy the computer with Windows, Office, and networking already installed from Dell. So I guess the system integration has become a mass market instead of a cottage industry.
I am thinking that for Linux to catch on, there has to be some patron, some "angel", some big player to do the system integration and sell ready-to-run systems to the mass market. Is it Wal-Mart and Lindows? Is it SUN and their "Java workstation?"
My favorite part of this is Spock calling vacuum tube electronics "bear skins and stone knives." I saw the episode as a kid and didn't understand what he was trying to do. Much later I figured that the tricorder lacked a video display, but they needed to see the fast forward images in the Portal to know where McCoy was, and Spock was using 1930's technology to build a video monitor to retrieve the readout from the tricorder of the Portal images. So the tricorder lacked a flat panel display. Sheesh!
This young white dude, computer programmer family man is found shot dead. In his house. With an exotic WW-II German pistol.
The crack team of Briscoe and Green do their leg work, and they come across an old black dude who is somehow connected to the young white dude in a financial scam. The cops think "salt and pepper" team and one crime guy turning on his partner.
The D.A.'s office goes to work on this and what finally emerges is 1) the black dude is an upstanding citizen and a war hero (hence the WW-II war trophy gun), 2) the white dude lost his job and turned to cybercrime to support his family, 3) the black dude had no connection to the white dude apart from that the white dude picked a random victim to scam for identity theft, 4) the black dude, finding all of his net worth was scammed, used his charm and his war hero "street smarts" to get the ISP to give up the address of the guy who ripped him off.
Like many Law and Order episodes, the whole thing is Hollywood and little connection to real cops, victims, and criminals. On the other hand, the writers are not that creative and probably use some news stories as a jumping off point.
Like the better of the Law and Order episodes, it makes you think. What right did the white dude have to pick a random victim and take all his money? What right did the black dude have to get his revenge as murder? What right does society have to put an 80 year old man (the black dude) away for seeking his own justice?
But to me the moral of this episode along with the baseball bat incident is what goes around comes around. No, I don't think threatening a 15 year old with a baseball bat is a good idea, and the 15 year old and parents can get in a lather about their "rights." But a 15 year old with access to a computer (and the parents of said 14 year old) could appreciate that if one could hack into someone's server, owner of such server could hack back to track you down, and such owner could be meaner, tougher, better connected, or in a much worse mood than you. And somehow going after people who threaten 15 year olds with bats under those circumstances doesn't seem like the highest priority for the courts.
I heard that Ellison's original script had some low-level Enterprise crew doing drugs which leads to this time-travel incident, but the Hollywood suits took that out because the Enterprise crew are all Eagle Scouts and that would never happen in the Star Trek universe, so they rewrote the script to have McCoy accidently inject himself with a dangerous (legit) medication when the Enterprise hits a space-time "air pocket" (you know, one of those things that tosses everyone out of the seats that they are not belted into).
If you write scripts for Hollywood, having someone else do a rewrite for any of a number of pandering reasons is part of the landscape, and as far as making the Enterprise crew Boy Scouts who would never use drugs, whoever is supervising a series has to keep the episodes consistent with a vision or else it turns into Superman Comics where Superman keeps getting more powers in each episode that they have to hit some kind of reset button.
Getting back on topic, Ellison may not be a nut case, but he has a track record of pissing into the wind on matters of principle that turn out to be no-win situations.
Ha, for a real babe magnet you need a bumper stickers which say "do my own laundry", "cook more than hamburgers", and "make more than minimum wage."
Of course that doesn't do too much. So you put a plutonium rod in the middle as a "spark plug." That gets some fusion going from the flood of neutrons, but most of what that does is just give more neutrons which would otherwise fly off and not contribute to the explosive force unless you surround the works with U-238, which fissions from the flood of neutrons.
And people call this a fusion device. This is relevant to the discussion of table-top fusion because even uncontrolled fusion is hard to get -- the original Teller idea of the "Super" which would be just sticking an A-bomb next to a tub of deuterium simply doesn't work, and it is a Good Thing because otherwise A-bomb explosions could start a fusion chain reaction in the Earth.
So they haven't yet implemented garbage collection -- what do you do, keep putting in more RAM sticks?
The Audacity spectrogram (I suppose it is just an add on and not the primary focus) is a dithered multi-color mess that scrolls like molasses.
Windows offers 1) hardware assisted scroll with ScrollWindowEx, 2) vertical retrace synch, two features that go a long way to making a good scrolling waveform display/spectrum analyser that I don't see in and of the platform portable tool kits. Those features, by the way, were added to Windows when 2-D scrolling games were a big thing (ScrollWindowEx I believe was part of WinG and vertical synch is part of IDirectDraw). Windows is a little weak in the latency department, but running everything on really fast processors is kind of taking care of that.
What is out there that is an improvement on Windows in this regard? I heard BeOS had very low latencies (I had heard the same about OS/2), but both those systems are pretty much past history if you want to reach a lot of people.
What are you, some kind of Commie Lisp programmer?