By "evil" I presume you mean "bad for privacy," but this is a non-issue if you control your own personal data repository on your own server in your own basement. To fully reap the benefits of ubiquitous internet access, all your data must be somewhat centralized. (For a large business, this centralization may, of course, be logical not physical) So my take would be "centralize the data, distribute the processing."
Aha, so you advocate exactly the same kind of logical fabric that I do. You just call it 'ubiquitous access' and you're not quite sure how it will play out at the tech level.
It's the function of the tech layer (the physical computation layer) to provide storage (with associated risk) and computation (with associated load). Everything else, like logical location of data and even privacy, can be provided in a higher layer. Since everything that the tech layer provides is merely necessary (but bad) we don't want to centralize this layer (ie, centralize risk and load) but distribute it. It's only the higher layer that we want to centralize.
In particular, the DNS / Web way of addressing resources based on their physical tech location is something we want to do away with. Though whatever does away with DNS will have to provide substantially more advantages than the minimal ones provided by URNs. It will have to be a revolutionary step up, not an evolutionary step forward.
Dynamic as in "what you see is what you want/need" rather than static GUIs that somebody designed with specific use cases in mind. (not that there isn't a place for both..)
In general, you can go the scripted route, or you can go the direct manipulation route. DM ought to be automated for consistency. And it's generally useful for it to be reflexive, though how much reflexion doesn't blow the user's mind is debatable.
I'm always open to new ideas, but what currently exists to replace it? With the amount investment into XML family tools at this point and the amount of unity XML related projects are bringing, a successor would have to be something dramatically better to be worth throwing everything out and starting over.
There's only one solution, Kill The Web. If you kill the web, you take away the entire foundation upon which XML is based and you can create a new foundation however you please. This is actually feasible even in the short term (5-10 years). All it takes is a small team of programmers creating the next revolutionary step up. And this is believable if you appreciate that in design terms, software systems are suboptimal in the 'order of magnitude' range, and that there is a pool of severely underutilized design talent that you can tap into.
At this point I would enjoy continuing this conversation more privately. My email address is p.r.o.m.e.t.e.u.s.5.7.@y.a.h.o.o.c.a. Pity slashdot doesn't have double-blind private exchange.
My regards, I'm rather enjoying this meeting of minds.:)
I may not agree with any of your conclusions but I do recognize that you understand the issues involved. Kudos for that. Having dealt with the "XML is the nectar of the gods, how dare you criticize it!" attitude, I do appreciate your having something meaningful to say.
Regarding PUI, I'll take your word for it. It was just strange that you considered the desktop metaphor to have been perfectly valid and useful for 15 years (from 25 years ago to 10 years ago). That's a damnably long period for anything to survive in a rapidly changing field.
It's surprising that you'd consider Star Trek to have anything to offer in terms of providing a model for technology and interface design. Star Trek is well behind the known state of the art *today*. Or more accurately, the state of the dreck. Star Trek interfaces are completely AI driven. It's the only way anything makes sense.
Regarding your concrete proposals: * ubiquitous access is a function of the capabilities and costs of technology, not any design issues at all. * people today already have seamless interoperability if they stick with Apple. So broken systems are broken, yawn. * centralized data stores are evil, what we should aim for is a secure logical fabric on top of a highly distributed tech layer. Somewhat like Ameoba aimed for but much better. * highly associative / semantically rich? You invented this. The only evidence Star Trek has of it is its AI functions. Star Trek doesn't provide it as a model separate from AI, although it's certainly desirable and feasible. * diverse client hardware? You invented this too. * dynamic user interfaces? Again, something you invented, though there it's not even clear what you mean, nevermind whether it's a good idea.
Regarding XML, the flaw in your reasoning is your assumption that XML has valid uses. XML is a truly horrific way of marshalling and distributing even textual data, even so-called documents. But to see that you have to think of documents as they *could* be (eg, think Xanadu) and not the crap we're currently dealing with which HTML created and XML aims to perpetuate.
As for RPC, you haven't lived until you've worked in a language where RPC is done transparently.
As for separating code from data, this is what object-classes do. Objects contain the data and classes contain the code. Perfect separation! Without the uglification and horrendous waste that XML is heir too. Without all the hype either.
The point here is that what requires an entire "language" with XML is just a big yawn with a high-level object-oriented language. So if you don't like Lisp then go with Smalltalk, Opentalk is excellent. Either way, it's a complete non-issue if you bother to do it right. So no, there will never be a "language" that marshalls arbitrary data and text elegantly... because it will be a small component of a real language and because its user-visibility will be nil.
It's practically a tautology, you can never have a beautiful solution to a problem you've done away with.
You should entertain the notion once in a while that a person formed their opinion from intensive study instead of merely parroting what everybody else says. The assumption that everyone else thinks and acts exactly like you do is a comfortable one but it is false.
So for instance, "full fledged AI" obviously refers to natural language processing, something we patently do not have. It may also refer to human ADULT equivalent AI instead of the human-toddler equivalent which Cyc achieved or the various idiot savant expert systems floating around.
And since we broached this subject when someone brought up having a CONVERSATION with their computer, I dismiss your "more intelligent usage" as so much irrelevant gobbledygook. And it is gobbledygook since you're trying to pass off better designed applications as some kind of revolution in interface design. Nice bit of sophistry there.
Regarding XML, there you truly showcase the mindless parroting that you pass off as original thought and considered opinion.
First off, the whole "content vs presentation" is a complete non-issue. It has absolutely nothing to do with what's wrong with XML. What's wrong with XML is simply that SGML was invented to deal with textual documents, precisely for the purposes of typesetting. Or, not to put too fine a point on it, *automated* typesetting. SGML was invented so that a computer could automatically typeset a TEXTUAL document based on abstract considerations. You may like to think that SGML was invented for "marking up" any arbitrary data but you would be wrong. This is not something you can just wave off, the fact that SGML was invented to deal with TEXT in a TYPESETTING environment is absolutely crucial to its nature. And once one understands the nature of SGML, the problem-space it was highly-optimized to solve, it becomes obvious that it is absolutely the wrong solution for formatting arbitrary binary objects.
But to go back to your contention that presentation is the result of "browser wars", with the implicit claim that presentation really doesn't matter and that it's not something that people should have bothered with... exactly what kind of idiot are you? You do realize that those browser wars only occured because people WANTED to control presentation, yes? You also realize that this desire by people to control presentation was PERFECTLY OBVIOUS, or should have been at any rate, to the inventors of HTML, yes? Once you accept these facts, it is obvious that the deficiency wasn't on the part of Netscape and Microsoft, but on the part of HTML itself.
The fact that you can't follow Naggum's arguments, which are perfectly coherent, clear and elementary, means you really don't understand the issues. In which case, why the hell are you even talking about this? Your rambling about processing being cheap (yes exactly, so why should a solution invented for when processing was expensive be retained?) and XML being "self-describing" (an absurd propagandistic notion) only confirm that you know and understand nothing of the issues involved.
As I previously said, XML is entirely the wrong solution to format arbitrary binary objects. The proper solution for this problem is S-expressions, which were invented two decades before SGML ever was, never mind XML. No doubt, S-expressions are entirely off of your radar. Since you don't understand the issues involved, you'll have to accept my considered opinion that XML is a horrific solution to a problem that was solved more than *four decades* before XML ever came on the scene.
Just want to correct an error you made. The desktop metaphor aka the "GUI" aka the PUI aka the Xerox PARC User Interface invented for the Alto, is 25 years old, not 10.
Also, the so-called Star Trek interface you paint is really full-fledged AI and so is stupid as a goal of serious interface design.
Finally, XML is a step backwards. A HUGE step backwards. What was good for typesetting documents is really horrific for arbitrary objects. Erik Naggum pointed out as much more many years ago in this article: http://groups.google.com/group/comp.lang.lisp/msg/ a0e340b526c4a07
You could have voice commands, and they would be INCREDIBLY useful, flexible, powerful and fast if done right. We already have the technology to do that, it's called voice recognition.
But conversation is a completely different problem, a completely different tech; natural language processing. In order to have a conversation with a machine, you'd pretty much need to solve the AI problem. Good luck.
The detractors of voice commands are hide-bound narrow-minded idiots. One of their favourite tactics is to confuse voice commands with conversation then claim "see, natural language processing isn't possible yet so you can't make use of voice recognition".
There's absolutely nothing in this article that hasn't been known for the last 10 years. There are no insights. There is nothing remotely novel, interesting or useful. This article is trash.
The author is worse. Tristan gets it wrong every time. He makes at least one basic error in every page of the article. For example,
* trees are great because they impose an order found in nature? Yeah, that's just great if you're a biologist. Guess everyone else is a second-class citizen.
* DAGs are great because they're acyclic? Wrong. Acyclic has absolutely nothing to do with it, it's a completely useless property.
Then he mumbles some kind of crap about why trees are so supposedly great. Yuck. Next page:
* Human perception of the world is 2D? Bullshit! Human perception of the world is 2.5D, which is in between 2D and 3D.
Then he mumbles some kind of crap about the human brain compensating for the 2D visual system. Complete bullshit.
The rest is just more of the same. He describes "Physical Simulation" as if it weren't entirely discredited (though that's never stopped programmers from doing something). And he doesn't mention the achilles' heel of "complex information visualization" schemes; their non-generality.
He conflates manual placement of objects with spatial memory. Bullshit! The reason every useful computer provides an auto-arrange feature for folders is because humans are notorious for misplacing and losing stuff. It is in fact NOT AT ALL USEFUL to give users the ability to (mis)place stuff at will. And you don't even need an expensive useability study to prove this because it can be proved from straightforward analysis.
And Tristan finishes his article by mumbling some crap about how the desktop might be optimal. As if! The reason it's persisted unchanged for 20 years is programmers' well-documented conservatism and rampant lack of imagination.
This author doesn't know shit.
Slashdot Mirror
Aha, so you advocate exactly the same kind of logical fabric that I do. You just call it 'ubiquitous access' and you're not quite sure how it will play out at the tech level.
It's the function of the tech layer (the physical computation layer) to provide storage (with associated risk) and computation (with associated load). Everything else, like logical location of data and even privacy, can be provided in a higher layer. Since everything that the tech layer provides is merely necessary (but bad) we don't want to centralize this layer (ie, centralize risk and load) but distribute it. It's only the higher layer that we want to centralize.
In particular, the DNS / Web way of addressing resources based on their physical tech location is something we want to do away with. Though whatever does away with DNS will have to provide substantially more advantages than the minimal ones provided by URNs. It will have to be a revolutionary step up, not an evolutionary step forward.
Dynamic as in "what you see is what you want/need" rather than static GUIs that somebody designed with specific use cases in mind. (not that there isn't a place for both..)
In general, you can go the scripted route, or you can go the direct manipulation route. DM ought to be automated for consistency. And it's generally useful for it to be reflexive, though how much reflexion doesn't blow the user's mind is debatable.
I'm always open to new ideas, but what currently exists to replace it? With the amount investment into XML family tools at this point and the amount of unity XML related projects are bringing, a successor would have to be something dramatically better to be worth throwing everything out and starting over.
There's only one solution, Kill The Web. If you kill the web, you take away the entire foundation upon which XML is based and you can create a new foundation however you please. This is actually feasible even in the short term (5-10 years). All it takes is a small team of programmers creating the next revolutionary step up. And this is believable if you appreciate that in design terms, software systems are suboptimal in the 'order of magnitude' range, and that there is a pool of severely underutilized design talent that you can tap into.
At this point I would enjoy continuing this conversation more privately. My email address is p.r.o.m.e.t.e.u.s.5.7.@y.a.h.o.o.c.a. Pity slashdot doesn't have double-blind private exchange.
My regards, I'm rather enjoying this meeting of minds. :)
I may not agree with any of your conclusions but I do recognize that you understand the issues involved. Kudos for that. Having dealt with the "XML is the nectar of the gods, how dare you criticize it!" attitude, I do appreciate your having something meaningful to say.
Regarding PUI, I'll take your word for it. It was just strange that you considered the desktop metaphor to have been perfectly valid and useful for 15 years (from 25 years ago to 10 years ago). That's a damnably long period for anything to survive in a rapidly changing field.
... because it will be a small component of a real language and because its user-visibility will be nil.
It's surprising that you'd consider Star Trek to have anything to offer in terms of providing a model for technology and interface design. Star Trek is well behind the known state of the art *today*. Or more accurately, the state of the dreck. Star Trek interfaces are completely AI driven. It's the only way anything makes sense.
Regarding your concrete proposals:
* ubiquitous access is a function of the capabilities and costs of technology, not any design issues at all.
* people today already have seamless interoperability if they stick with Apple. So broken systems are broken, yawn.
* centralized data stores are evil, what we should aim for is a secure logical fabric on top of a highly distributed tech layer. Somewhat like Ameoba aimed for but much better.
* highly associative / semantically rich? You invented this. The only evidence Star Trek has of it is its AI functions. Star Trek doesn't provide it as a model separate from AI, although it's certainly desirable and feasible.
* diverse client hardware? You invented this too.
* dynamic user interfaces? Again, something you invented, though there it's not even clear what you mean, nevermind whether it's a good idea.
Regarding XML, the flaw in your reasoning is your assumption that XML has valid uses. XML is a truly horrific way of marshalling and distributing even textual data, even so-called documents. But to see that you have to think of documents as they *could* be (eg, think Xanadu) and not the crap we're currently dealing with which HTML created and XML aims to perpetuate.
As for RPC, you haven't lived until you've worked in a language where RPC is done transparently.
As for separating code from data, this is what object-classes do. Objects contain the data and classes contain the code. Perfect separation! Without the uglification and horrendous waste that XML is heir too. Without all the hype either.
The point here is that what requires an entire "language" with XML is just a big yawn with a high-level object-oriented language. So if you don't like Lisp then go with Smalltalk, Opentalk is excellent. Either way, it's a complete non-issue if you bother to do it right. So no, there will never be a "language" that marshalls arbitrary data and text elegantly
It's practically a tautology, you can never have a beautiful solution to a problem you've done away with.
You should entertain the notion once in a while that a person formed their opinion from intensive study instead of merely parroting what everybody else says. The assumption that everyone else thinks and acts exactly like you do is a comfortable one but it is false.
... exactly what kind of idiot are you? You do realize that those browser wars only occured because people WANTED to control presentation, yes? You also realize that this desire by people to control presentation was PERFECTLY OBVIOUS, or should have been at any rate, to the inventors of HTML, yes? Once you accept these facts, it is obvious that the deficiency wasn't on the part of Netscape and Microsoft, but on the part of HTML itself.
So for instance, "full fledged AI" obviously refers to natural language processing, something we patently do not have. It may also refer to human ADULT equivalent AI instead of the human-toddler equivalent which Cyc achieved or the various idiot savant expert systems floating around.
And since we broached this subject when someone brought up having a CONVERSATION with their computer, I dismiss your "more intelligent usage" as so much irrelevant gobbledygook. And it is gobbledygook since you're trying to pass off better designed applications as some kind of revolution in interface design. Nice bit of sophistry there.
Regarding XML, there you truly showcase the mindless parroting that you pass off as original thought and considered opinion.
First off, the whole "content vs presentation" is a complete non-issue. It has absolutely nothing to do with what's wrong with XML. What's wrong with XML is simply that SGML was invented to deal with textual documents, precisely for the purposes of typesetting. Or, not to put too fine a point on it, *automated* typesetting. SGML was invented so that a computer could automatically typeset a TEXTUAL document based on abstract considerations. You may like to think that SGML was invented for "marking up" any arbitrary data but you would be wrong. This is not something you can just wave off, the fact that SGML was invented to deal with TEXT in a TYPESETTING environment is absolutely crucial to its nature. And once one understands the nature of SGML, the problem-space it was highly-optimized to solve, it becomes obvious that it is absolutely the wrong solution for formatting arbitrary binary objects.
But to go back to your contention that presentation is the result of "browser wars", with the implicit claim that presentation really doesn't matter and that it's not something that people should have bothered with
The fact that you can't follow Naggum's arguments, which are perfectly coherent, clear and elementary, means you really don't understand the issues. In which case, why the hell are you even talking about this? Your rambling about processing being cheap (yes exactly, so why should a solution invented for when processing was expensive be retained?) and XML being "self-describing" (an absurd propagandistic notion) only confirm that you know and understand nothing of the issues involved.
As I previously said, XML is entirely the wrong solution to format arbitrary binary objects. The proper solution for this problem is S-expressions, which were invented two decades before SGML ever was, never mind XML. No doubt, S-expressions are entirely off of your radar. Since you don't understand the issues involved, you'll have to accept my considered opinion that XML is a horrific solution to a problem that was solved more than *four decades* before XML ever came on the scene.
Just want to correct an error you made. The desktop metaphor aka the "GUI" aka the PUI aka the Xerox PARC User Interface invented for the Alto, is 25 years old, not 10.
/ a0e340b526c4a07
Also, the so-called Star Trek interface you paint is really full-fledged AI and so is stupid as a goal of serious interface design.
Finally, XML is a step backwards. A HUGE step backwards. What was good for typesetting documents is really horrific for arbitrary objects. Erik Naggum pointed out as much more many years ago in this article: http://groups.google.com/group/comp.lang.lisp/msg
You could have voice commands, and they would be INCREDIBLY useful, flexible, powerful and fast if done right. We already have the technology to do that, it's called voice recognition.
But conversation is a completely different problem, a completely different tech; natural language processing. In order to have a conversation with a machine, you'd pretty much need to solve the AI problem. Good luck.
The detractors of voice commands are hide-bound narrow-minded idiots. One of their favourite tactics is to confuse voice commands with conversation then claim "see, natural language processing isn't possible yet so you can't make use of voice recognition".
There's absolutely nothing in this article that hasn't been known for the last 10 years. There are no insights. There is nothing remotely novel, interesting or useful. This article is trash. The author is worse. Tristan gets it wrong every time. He makes at least one basic error in every page of the article. For example, * trees are great because they impose an order found in nature? Yeah, that's just great if you're a biologist. Guess everyone else is a second-class citizen. * DAGs are great because they're acyclic? Wrong. Acyclic has absolutely nothing to do with it, it's a completely useless property. Then he mumbles some kind of crap about why trees are so supposedly great. Yuck. Next page: * Human perception of the world is 2D? Bullshit! Human perception of the world is 2.5D, which is in between 2D and 3D. Then he mumbles some kind of crap about the human brain compensating for the 2D visual system. Complete bullshit. The rest is just more of the same. He describes "Physical Simulation" as if it weren't entirely discredited (though that's never stopped programmers from doing something). And he doesn't mention the achilles' heel of "complex information visualization" schemes; their non-generality. He conflates manual placement of objects with spatial memory. Bullshit! The reason every useful computer provides an auto-arrange feature for folders is because humans are notorious for misplacing and losing stuff. It is in fact NOT AT ALL USEFUL to give users the ability to (mis)place stuff at will. And you don't even need an expensive useability study to prove this because it can be proved from straightforward analysis. And Tristan finishes his article by mumbling some crap about how the desktop might be optimal. As if! The reason it's persisted unchanged for 20 years is programmers' well-documented conservatism and rampant lack of imagination. This author doesn't know shit.