So, Libranet offers a variant of Debian for download for $15, appealing to people who have high speed Internet connections and the smarts to make a CD. Meanwhile, there are a handful of established distributions available for free download and/or on CD, easy enough for complete novices to install by putting a CD into the drive.
Sure, they can probably do this; it seems to be in compliance with the GPL. RedHat and others may even try to go down the same road (and with for-pay updates, they are). But my guess at this point is that their business model isn't going to work. But, hey, that's part of a free market: people try differnt things. Some of them work and some of them don't. If they figure how to add sufficient value to their services, they'll make it, and we'll all win.
I don't think the GPL says that you can redistribute the ISO images. I think it merely says that you can get the sources for the GPL'ed binaries that are on the ISO images (there may be others). You can then use those to make your own ISO images, which are very unlikely to be identical to the original ISO images.
That's the way a market economy works: if you know you can get something comparable free/cheaper, you take it rather than the more expensive product. Doing anything else would be irrational, for a Windows or MacOS user as much as for a Linux user.
If people do pay for stuff where a comparable free product exists, it's for opportunity cost: somebody paying for some commercial system may not know that some particular free system would solve his problems just as well, or it may be too costly to actually get a hold of (e.g. download) the free product.
Also, you can doodle pictures, which are then automatically converted to images, which can them be resized and placed elsewhere. [... more old stuff deleted...]
How nice. Microsoft is, again, delivering 10-20 year old technology developed by other companies and university research projects, after they killed off all their competitors with fake demos and other shenanigans. And because the memory of the average Microsoft employee or customer in their 20's doesn't extend that far, they even get the reputation for having invented this kind of thing.
You can handwrite notes, and the software can spell-check, in handwriting!
Like, totally cool, dude. Will wonders never cease? You don't have to disable the spell checker if you use a different input method? Never mind that handwriting recognition already makes extensive use of spell checking techniques to translate people's illegible scribbles into words.
One might even be able to get Linux running on it, only sans the neato software.
Oh, I wouldn't be so pessimistic. The research projects that developed these technologies long before Microsoft may be old, but they weren't done on punch cards or paper tapes either. You can probably still find source code for some of this stuff on the Internet.
As for handwriting recognition, there are several projects already to deliver state-of-the-art handwriting recognition for Linux.
Objective-C's mechanism was considered for Java but dropped in favor of nested classes, which fulfill the same function and have other advantages.
Proxying in Objective-C is convenient but not safe, while proxying in Java is less convenient but safe. Which is better? Hard to tell. I would err on the side of safety.
Incidentally, both Objective-C features can be implemented efficiently on top of Java's current runtime; the issue of "C++-style method binding" is a red herring.
AMD's strategy makes a lot more sense to me than Intel's. Intel simply delivered the best processors they could under the constraint of backwards compatibility, which turned out to be good enough for the market even though other companies had faster and cheaper processors.
With 64bit processors, the roles are reversed. Intel's 64bit architecture is a radical break with the past. Code generation for it is completely different from previous Intel processors, and even harder than for other RISC machines. AMD, OTOH, is delivering something backwards compatible with some 64bit instructions.
If I wanted a nice 64bit processor and didn't care about backwards compatibility, I'd get a Sun Blade or an Alpha today. But I'm using a PC because of the widespread adoption of the processor. 64bit support in a PC to me is a convenient add-on feature, but I don't want to throw out my existing environment over it. Altogether, Intel's 64bit processor doesn't look attractive to me at all.
Well, I dislike the idea, and I think we may well end up being forced to buy this stuff under the DMCA. However, I think there are two potential benefits.
First, I don't like the trash that the commercial media and content providers are putting out anyway. The less easy it is for people to get a copy of the latest Disney movie, the better as far as I'm concerned. And I don't think that these devices can be so closed that only commercial content can be delivered: you should still be able to deliver free content freely.
Second, the hardware support for this may end up being usable for securing our own data cryptographically. So far, people haven't encrypted their disks because it's too expensive and because of export controls. This kind of hardware may well put hardware-based cryptography on every desktop.
It's not like the government is trying to force hardware manufacturers to do this. If Intel starts making chips that have copyright protection built-in, it only opens up a market to a chip manufacturer that won't.
If most PCs have this and can play content this way, hardare that doesn't support it could be considered having no substantial use other than to circumvent copyright protections, in violation of the DMCA.
So, yes, once this kind of hardware is widely available, the government may well end up forcing consumers to buy this stuff and hardware manufacturers to stop producing anything else.
(1) The degree to which QM helps with computation is an open question. (2) The authors don't claim that QM is relevant to their problem. (3) QM hasn't been "proven"; in fact, it's pretty much accepted that QM cannot be a complete and correct physical theory. (4) Having a working computer for the six node case tells you nothing about complexity or whether the approach works for larger systems (and using any analog system algorithmically itself is fraught with problems when it comes to defining "solve"). (5) The article itself says that their computer grows exponentially in size, so what they actually came up with is an exponentially sized circuit to solve an NP hard problem in less than an exponential number of steps, which is not the same as solving NP complete problems in polynomial time.
I haven't found much difference in similarly-priced same-generation machines. And performance is variable between applications: one processor may beat another on integer performance but be behind on floating point and graphics. Within 30% or so, you might as well flip a coin unless you spend a lot of time benchmarking.
One consideration is that GNU C is probably not as good for the PPC as it is for the Pentium, so published benchmarks may not quite apply. Also, you can't get the Apple without an OS, so you pay a little extra for that. Furthermore, there is quite a bit of s/w that doesn't exist for Linux PPC and would be difficult to port.
In short, for running Linux, I'd go with a Pentium. I'd consider buying the Mac with MacOSX as an alternative, though.
Peer-to-peer sharing used to be the norm on the Internet: most hosts on the network participated equally in services like FTP, IRC, talk, mail, and USENET.
The current view of the world came when we got slow, intermittent dial-up connections, firewalls, Windows, and ISPs. But most of all, it came when companies like ICQ and Hotmail wanted to derive large amounts of money by funneling traffic through their sites.
So, where are we going? Content providers want central control over distribution and service providers want to tie users to their services. Many ISPs write into their TOS that you aren't allowed to run servers because they are afraid of the content providers, don't want to provide the bandwidth anyway, and want to charge much higher fees to supposedly commercial servers.
Altogether, the outlook looks pretty bleak to me: the Internet is at risk of turning into a medium where the majority of information is provided by those with money and power. There is one bright spot, though: while less and less relative bandwidth is P2P, there is a lot more bandwidth than 10-20 years ago. So, while in relative terms, almost all audio and video may come from big companies with commercial agendas, there is more bandwidth than ever available for peer-to-peer distribution of text and image content. And while "access rights management" on audio, video, and e-book formats may be in our future, plain ASCII and images are likely not to fall under that either. I hope that's all we need to keep the web available for important non-commercial social functions.
The abstract claims that their system solves NP complete problems in polynomial time, which is clearly wrong even given the little information they provide in the abstract.
Well, I managed to get a hold of the full article. It turns out that their circuits are exponential size and the largest problem they can solve even using optimistic assumptions is of size 20 (if that). So, they didn't even manage to solve the problem in polynomial time using fluid dynamics, they are merely confused about the meaning of the term "NP complete" and "polynomial time" (which refer to complexity on a Turing machine, not circuit depth). In their introduction, they also make lots of mistakes in the definition of NP completeness.
These people don't know what they are doing, and they don't seem to have found anything interesting as far as I can tell. Just ignore it. As far as I know, not all articles in PNAS are peer reviewed; this one should have been.
You don't have to know much about physics to see that this is wrong. No matter what kind of abstractions you find convenient in your discipline, computation ultimately is implemented in terms of particles and their interactions. Particles have a finite size and finite interaction times, and you can't outrun an exponential with a classical system. You probably can't even outrun it with a quantum system.
Such a network may solve an NP complete problem quickly under an idealized model of fluid dynamics. However, the real world does not obey idealized fluid dynamics. It doesn't obey it in theory, and it doesn't obey it in practice. Fluid dynamics is a convenient approximation that breaks down at some point.
It's am embarrassment to see this kind of nonsense come out of Harvard. There are lots of people there who should know better. This isn't a problem with some obscure point in the physics of computation, it shows a pretty fundamental misunderstanding of physical theory by a physicist.
First, we don't know what the worst case complexity of NP complete problems is. Furthermore, NP completeness is about worst case complexity, which does not necessarily measure practicality.
But let's assume that some computation really does take more than polynomial time. Or, let's even assume it takes a really high polynomial, like N^1000 steps for a problem of size 10, as you yourself suggest. Can physics help us?
Not really. Let's say you want to solve a problem of size 10. That's 10^1000 steps. Let's take a good macroscopic number of atoms, say 10^24. Let's say that all these atoms even interact simultaneously and that each interaction is a computational step, that's 10^48 steps. Let's assume that each step takes 10^-15 seconds. So, we can now perform 10^63 computations per second. Now, 10^1000/10^63 is still 10^937, or 10^928 years, or 10^919 billion years. I'm not willing to wait around for that answer. And, so far, it appears that even quantum computing isn't going to help us out of this one.
The US record on dealing with crime is very poor compared to other western countries. The US has a very high murder rate, the US prison system is in shambles and doesn't even attempt rehabilitation, the US prison system has been cited for human rights violations by AI and the UN, the US has high recidivism rates, sensitive private records about citizens are available from the US government to business just for the asking, and US police may impose many kinds of penalties without a trial.
It does not seem to me like a good thing to me if a country with such a poor record is trying to organize "crime fighting" in democracies that have a much better record on crime and crime fighting.
I wonder how much this is due to lack of cooperation by the local phone company. I had Northpoint DSL installed. Their installers were pretty prompt, but PacBell screwed up left and right. Given that PacBell wants this market for itself, why should they help Northpoint? Even if they don't want to serve a particular area now, they probably would much rather get the customers later than have them sign up with another company.
I would trust a company that takes its responsibility as a certificate authority seriously. VeriSign's responsibility seems to extend no further than to charging your credit card. There are companies whose management looks into the future and spends the necessary money today to earn and keep a reputation. VeriSign doesn't seem to be one of them, and I don't have much hope for them.
Maybe an established insurance company that both issues certificates and insures companies against abuses of certificates has a sufficient incentive to act more prudently.
You have to separate the company from the nerds like you and me.
Working for Microsoft is a choice. You don't have to do it. There are (still) lots of other jobs around. If you are waiting for options to vest and think they are going to be worth something, you should be able to laugh all the way to the bank and put up with the complaints.
but there are a lot of cool OS and game programmers
You can argue whether the stuff that Microsoft produces deserves the commercial success it has. But in terms of technology, Microsoft is about as "cool" as a MacDonald's Happy Meal. Microsoft produces software for the masses, using mostly technology that was roughly state of the art in 1980. Oh, but I will agree: they are getting pretty cool when it comes to graphic design.
Guys, Microsoft is not nearly as evil as you think it is.
Sure they are: the company E-mail and Gates's testimony that have come out have confirmed people's worst suspicions. Microsoft is a big company with but one goal: profit by any means that are even marginally legal. The fact that there are lots of other companies that are just as bad doesn't excuse Microsoft.
While Obj-C doesn't have garbage collection, the Foudation kit does.
The Foundation Kit really only has a some utility classes for helping you keep track of pointers. That's nice, but it's not quite the same.
but once you've learned to retain your objects you very seldom have pointers pointing wrong,
Well, the value of safety isn't so much for one's own code, it's to make it easier and safer to use other people's code. Reuse becomes easier because you have to worry less about the memory management policies that another library may use. It becomes safer because a "bad apple" won't usually make your whole program crash.
One example is when an object doesn't respond to a method call. Then a special function is called with a structure describing the call and the object can do whatever it wishes.
Like many other features in Objective-C and NeXTStep (libraries, browsers, UI), that actually comes directly from Smalltalk (doesNotUnderstand:). More recent languages like Python also have it.
It's a real shame that Objective-C wasn't used more when it came out; adding garbage collection and a bit more runtime safety to it is not all that hard, and with a larger user community, that would probably have happened by now.
NeXTStep managed to package Smalltalk-like convenience with the efficiency of C and a UNIX-like operating system. That was great a decade or two ago. In hindsight, I think the industry made a big mistake not going that way (but Stepstone's and NeXT's lousy corporate strategy didn't help). Even today, I think the GNUstep programming model is much more convenient than Gtk+ or Qt.
But does that mean I want to use GNUstep? Not really anymore. I view Java, Java2D, and Swing as the direct descendent of NeXTStep and OpenStep. Java as a language has many of the same dynamic features as Objective-C, but Java adds runtime safety and garbage collection, two really important features. Java2D has an imaging model very similar to DPS. And Swing is a mature, powerful toolkit that has borrowed a lot from its predecessors. In terms of size and performance, a Java installation is probably smaller than a full OpenStep (or GNUstep) installation, and Java performs well enough on modern hardware.
So, I understand the appeal of GNUstep. Some things are still easier in it than in Java/Swing. But, on balance, I think Java is a pretty decent successor, and problems are getting ironed out. If you find Java programming itself too cumbersome, you can always use some of the more dynamic languages built on top of Java, like Jython.
The main driving force behind the InterfaceBuilder and the OpenStep API's is ObjectiveC's deferred method call linking. This language is what enables the juiciest part of the IB, the one people seldom talk about: the possibility to build object graphs by plugging objects into other objects' outlets,
Java gives you the same functionality. In fact, you can build a set of GUI objects and serialize them if you want to.
And yes, Java comes painfully close, but its C++ legacy perpetrates enough brain damage to cancel the benefits of the ObjectiveC influence.
Lots of people claim that, but repeating it over and over again doesn't make it true. Yes, the languages are different, but in terms of expressiveness, they don't seem much different. In fact, even Smalltalk has been implemented on top of the JVM. So, can you be more specific? What is it that you think you can't do in Java that you can do in Objective-C?
Berlin is a Vector-based GUI. [...] I don't think there are any other GUI's out there with that capability
DisplayPostscript and Swing, for example.
Many X11 toolkits provide rescalable primitives if you want to use them. However, it is generally not correct to use vector scaling for changing window or widget sizes (e.g., displaying a font at 12pt and at 24pt is not just scaling).
You and all the other "X has (won/caught up)" folks miss the important point about opensource OSes - that there is no coerced standard, and little barrier to porting. Stuff like Berlin can grow in an "incubator" environment of skilled developers and early-adopters, and take its time in winning over ports and converts.
Yet, many open source projects do die, sometimes because they use old technology, sometimes because they are trying to solve a problem that is already well solved by other packages, sometimes because they aren't very well executed, etc. I'm merely predicting that Berlin will like fizzle out; you are welcome to make a different prediction.
One of the most important things to me is that X is not resolution independant. The X protocol postulates that everything has to be drawn pixel accurate, even over the network.
X doesn't require pixel accurate drawing for everything, it requires it only for a well-defined set of its primitives. Other parts of the X protocol, including scalable fonts, 3D, and antialiased drawing, are not bound by that.
One could change that, of course, in say X12, but it would break all backward compatibility. Or at least you would have nice resizable applications alongside with ugly fixed size application.
If the whole world consisted of desktop applications, yes, picking g raphics model like that in DisplayPostscript, Berlin, or Quartz is simple and convenient. However, the whole world does not consist of desktop applications. You need pixel-accurate drawing for low-resolution screens, and you need display-specific bitmaps for efficient network transport. This was important 20 years ago on almost all machines, but it is still important in industrial and commercial applications.
Since X11 can already give you both, I don't see a problem. If you stick with a toolkit that is based on a desktop view of the world (Gtk+ and Qt both are, and both use scalable fonts and antialiasing), you get the applications you want. The fact that the same window system works with other stuff as well is an added bonus.
GIMP is the best Linux image editor. KDevelop is the best Linux IDE. Reconcile that Lusenix boy.
So? What's your point? Be happy that you can at least run them on the same screen. If one were written for BeOS and the other for Berlin, you couldn't even do that.
Slashdot Mirror
Sure, they can probably do this; it seems to be in compliance with the GPL. RedHat and others may even try to go down the same road (and with for-pay updates, they are). But my guess at this point is that their business model isn't going to work. But, hey, that's part of a free market: people try differnt things. Some of them work and some of them don't. If they figure how to add sufficient value to their services, they'll make it, and we'll all win.
I don't think the GPL says that you can redistribute the ISO images. I think it merely says that you can get the sources for the GPL'ed binaries that are on the ISO images (there may be others). You can then use those to make your own ISO images, which are very unlikely to be identical to the original ISO images.
If people do pay for stuff where a comparable free product exists, it's for opportunity cost: somebody paying for some commercial system may not know that some particular free system would solve his problems just as well, or it may be too costly to actually get a hold of (e.g. download) the free product.
How nice. Microsoft is, again, delivering 10-20 year old technology developed by other companies and university research projects, after they killed off all their competitors with fake demos and other shenanigans. And because the memory of the average Microsoft employee or customer in their 20's doesn't extend that far, they even get the reputation for having invented this kind of thing.
You can handwrite notes, and the software can spell-check, in handwriting!
Like, totally cool, dude. Will wonders never cease? You don't have to disable the spell checker if you use a different input method? Never mind that handwriting recognition already makes extensive use of spell checking techniques to translate people's illegible scribbles into words.
One might even be able to get Linux running on it, only sans the neato software.
Oh, I wouldn't be so pessimistic. The research projects that developed these technologies long before Microsoft may be old, but they weren't done on punch cards or paper tapes either. You can probably still find source code for some of this stuff on the Internet.
As for handwriting recognition, there are several projects already to deliver state-of-the-art handwriting recognition for Linux.
Proxying in Objective-C is convenient but not safe, while proxying in Java is less convenient but safe. Which is better? Hard to tell. I would err on the side of safety.
Incidentally, both Objective-C features can be implemented efficiently on top of Java's current runtime; the issue of "C++-style method binding" is a red herring.
With 64bit processors, the roles are reversed. Intel's 64bit architecture is a radical break with the past. Code generation for it is completely different from previous Intel processors, and even harder than for other RISC machines. AMD, OTOH, is delivering something backwards compatible with some 64bit instructions.
If I wanted a nice 64bit processor and didn't care about backwards compatibility, I'd get a Sun Blade or an Alpha today. But I'm using a PC because of the widespread adoption of the processor. 64bit support in a PC to me is a convenient add-on feature, but I don't want to throw out my existing environment over it. Altogether, Intel's 64bit processor doesn't look attractive to me at all.
First, I don't like the trash that the commercial media and content providers are putting out anyway. The less easy it is for people to get a copy of the latest Disney movie, the better as far as I'm concerned. And I don't think that these devices can be so closed that only commercial content can be delivered: you should still be able to deliver free content freely.
Second, the hardware support for this may end up being usable for securing our own data cryptographically. So far, people haven't encrypted their disks because it's too expensive and because of export controls. This kind of hardware may well put hardware-based cryptography on every desktop.
If most PCs have this and can play content this way, hardare that doesn't support it could be considered having no substantial use other than to circumvent copyright protections, in violation of the DMCA.
So, yes, once this kind of hardware is widely available, the government may well end up forcing consumers to buy this stuff and hardware manufacturers to stop producing anything else.
(1) The degree to which QM helps with computation is an open question. (2) The authors don't claim that QM is relevant to their problem. (3) QM hasn't been "proven"; in fact, it's pretty much accepted that QM cannot be a complete and correct physical theory. (4) Having a working computer for the six node case tells you nothing about complexity or whether the approach works for larger systems (and using any analog system algorithmically itself is fraught with problems when it comes to defining "solve"). (5) The article itself says that their computer grows exponentially in size, so what they actually came up with is an exponentially sized circuit to solve an NP hard problem in less than an exponential number of steps, which is not the same as solving NP complete problems in polynomial time.
One consideration is that GNU C is probably not as good for the PPC as it is for the Pentium, so published benchmarks may not quite apply. Also, you can't get the Apple without an OS, so you pay a little extra for that. Furthermore, there is quite a bit of s/w that doesn't exist for Linux PPC and would be difficult to port.
In short, for running Linux, I'd go with a Pentium. I'd consider buying the Mac with MacOSX as an alternative, though.
The current view of the world came when we got slow, intermittent dial-up connections, firewalls, Windows, and ISPs. But most of all, it came when companies like ICQ and Hotmail wanted to derive large amounts of money by funneling traffic through their sites.
So, where are we going? Content providers want central control over distribution and service providers want to tie users to their services. Many ISPs write into their TOS that you aren't allowed to run servers because they are afraid of the content providers, don't want to provide the bandwidth anyway, and want to charge much higher fees to supposedly commercial servers.
Altogether, the outlook looks pretty bleak to me: the Internet is at risk of turning into a medium where the majority of information is provided by those with money and power. There is one bright spot, though: while less and less relative bandwidth is P2P, there is a lot more bandwidth than 10-20 years ago. So, while in relative terms, almost all audio and video may come from big companies with commercial agendas, there is more bandwidth than ever available for peer-to-peer distribution of text and image content. And while "access rights management" on audio, video, and e-book formats may be in our future, plain ASCII and images are likely not to fall under that either. I hope that's all we need to keep the web available for important non-commercial social functions.
Well, I managed to get a hold of the full article. It turns out that their circuits are exponential size and the largest problem they can solve even using optimistic assumptions is of size 20 (if that). So, they didn't even manage to solve the problem in polynomial time using fluid dynamics, they are merely confused about the meaning of the term "NP complete" and "polynomial time" (which refer to complexity on a Turing machine, not circuit depth). In their introduction, they also make lots of mistakes in the definition of NP completeness.
These people don't know what they are doing, and they don't seem to have found anything interesting as far as I can tell. Just ignore it. As far as I know, not all articles in PNAS are peer reviewed; this one should have been.
Such a network may solve an NP complete problem quickly under an idealized model of fluid dynamics. However, the real world does not obey idealized fluid dynamics. It doesn't obey it in theory, and it doesn't obey it in practice. Fluid dynamics is a convenient approximation that breaks down at some point.
It's am embarrassment to see this kind of nonsense come out of Harvard. There are lots of people there who should know better. This isn't a problem with some obscure point in the physics of computation, it shows a pretty fundamental misunderstanding of physical theory by a physicist.
But let's assume that some computation really does take more than polynomial time. Or, let's even assume it takes a really high polynomial, like N^1000 steps for a problem of size 10, as you yourself suggest. Can physics help us?
Not really. Let's say you want to solve a problem of size 10. That's 10^1000 steps. Let's take a good macroscopic number of atoms, say 10^24. Let's say that all these atoms even interact simultaneously and that each interaction is a computational step, that's 10^48 steps. Let's assume that each step takes 10^-15 seconds. So, we can now perform 10^63 computations per second. Now, 10^1000/10^63 is still 10^937, or 10^928 years, or 10^919 billion years. I'm not willing to wait around for that answer. And, so far, it appears that even quantum computing isn't going to help us out of this one.
It does not seem to me like a good thing to me if a country with such a poor record is trying to organize "crime fighting" in democracies that have a much better record on crime and crime fighting.
I wonder how much this is due to lack of cooperation by the local phone company. I had Northpoint DSL installed. Their installers were pretty prompt, but PacBell screwed up left and right. Given that PacBell wants this market for itself, why should they help Northpoint? Even if they don't want to serve a particular area now, they probably would much rather get the customers later than have them sign up with another company.
Maybe an established insurance company that both issues certificates and insures companies against abuses of certificates has a sufficient incentive to act more prudently.
Working for Microsoft is a choice. You don't have to do it. There are (still) lots of other jobs around. If you are waiting for options to vest and think they are going to be worth something, you should be able to laugh all the way to the bank and put up with the complaints.
You can argue whether the stuff that Microsoft produces deserves the commercial success it has. But in terms of technology, Microsoft is about as "cool" as a MacDonald's Happy Meal. Microsoft produces software for the masses, using mostly technology that was roughly state of the art in 1980. Oh, but I will agree: they are getting pretty cool when it comes to graphic design.
Sure they are: the company E-mail and Gates's testimony that have come out have confirmed people's worst suspicions. Microsoft is a big company with but one goal: profit by any means that are even marginally legal. The fact that there are lots of other companies that are just as bad doesn't excuse Microsoft.
The Foundation Kit really only has a some utility classes for helping you keep track of pointers. That's nice, but it's not quite the same.
Well, the value of safety isn't so much for one's own code, it's to make it easier and safer to use other people's code. Reuse becomes easier because you have to worry less about the memory management policies that another library may use. It becomes safer because a "bad apple" won't usually make your whole program crash.
Like many other features in Objective-C and NeXTStep (libraries, browsers, UI), that actually comes directly from Smalltalk (doesNotUnderstand:). More recent languages like Python also have it.
It's a real shame that Objective-C wasn't used more when it came out; adding garbage collection and a bit more runtime safety to it is not all that hard, and with a larger user community, that would probably have happened by now.
But does that mean I want to use GNUstep? Not really anymore. I view Java, Java2D, and Swing as the direct descendent of NeXTStep and OpenStep. Java as a language has many of the same dynamic features as Objective-C, but Java adds runtime safety and garbage collection, two really important features. Java2D has an imaging model very similar to DPS. And Swing is a mature, powerful toolkit that has borrowed a lot from its predecessors. In terms of size and performance, a Java installation is probably smaller than a full OpenStep (or GNUstep) installation, and Java performs well enough on modern hardware.
So, I understand the appeal of GNUstep. Some things are still easier in it than in Java/Swing. But, on balance, I think Java is a pretty decent successor, and problems are getting ironed out. If you find Java programming itself too cumbersome, you can always use some of the more dynamic languages built on top of Java, like Jython.
Java gives you the same functionality. In fact, you can build a set of GUI objects and serialize them if you want to.
Lots of people claim that, but repeating it over and over again doesn't make it true. Yes, the languages are different, but in terms of expressiveness, they don't seem much different. In fact, even Smalltalk has been implemented on top of the JVM. So, can you be more specific? What is it that you think you can't do in Java that you can do in Objective-C?
DisplayPostscript and Swing, for example.
Many X11 toolkits provide rescalable primitives if you want to use them. However, it is generally not correct to use vector scaling for changing window or widget sizes (e.g., displaying a font at 12pt and at 24pt is not just scaling).
Yet, many open source projects do die, sometimes because they use old technology, sometimes because they are trying to solve a problem that is already well solved by other packages, sometimes because they aren't very well executed, etc. I'm merely predicting that Berlin will like fizzle out; you are welcome to make a different prediction.
X doesn't require pixel accurate drawing for everything, it requires it only for a well-defined set of its primitives. Other parts of the X protocol, including scalable fonts, 3D, and antialiased drawing, are not bound by that.
One could change that, of course, in say X12, but it would break all backward compatibility. Or at least you would have nice resizable applications alongside with ugly fixed size application.
If the whole world consisted of desktop applications, yes, picking g raphics model like that in DisplayPostscript, Berlin, or Quartz is simple and convenient. However, the whole world does not consist of desktop applications. You need pixel-accurate drawing for low-resolution screens, and you need display-specific bitmaps for efficient network transport. This was important 20 years ago on almost all machines, but it is still important in industrial and commercial applications.
Since X11 can already give you both, I don't see a problem. If you stick with a toolkit that is based on a desktop view of the world (Gtk+ and Qt both are, and both use scalable fonts and antialiasing), you get the applications you want. The fact that the same window system works with other stuff as well is an added bonus.
So? What's your point? Be happy that you can at least run them on the same screen. If one were written for BeOS and the other for Berlin, you couldn't even do that.