I'm not sure that the rules are the same throughout Europe, but they tend to be significantly nastier here than in the US. Often there is no equivalent of Chapter 11. There have been a fair few suggestions recently to make the EU rules (a) the same as each other and (b) more like the US ones.
Not in the context of your comment. Scheme, as a language, is defined as properly tail recursive, and defines iterative constructs in terms of tail-recursive ones, whereas most other Lisp implementations (specifically Common Lisp and Emacs Lisp) are not defined this way, and do not do so. Since implementations may not eliminate tail calls, you definitely need to use iterative constructs if you want your code not to blow up in your face.
In fact Common Lisp has one of the most powerful iterative constructs I know - the LOOP macro.
They might leap, they might not. They might not be able to (conflicting options or features or not want to or it might be difficult or any number of other problems might arise.
Go and look at the roots of the Xemacs / FSF Emacs saga for how hard it can be.
I think they've finally blown all their feet off. The big machines are certainly the interesting bit, but their products are now even more of a tangled mess than they were before.
The combined company has (among others):
Alpha and alpha-based machines. Alpha is officially dead. Tru64 presumably will die with it. Anyway people won't buy too many of these boxes. OpenVMS also runs here, but this also must be of only legacy interest by now.
PA-RISC and the HP-UX machines based on it. PA-RISC is also rumoured to be dead, although it's taking a while to go away.
Big x86 boxes from both companies, running Windows.
Itanium, which is meant to replace all of these. Except, that, oh dear, it doesn't really exist yet: it's horribly, horribly late (4 years plus), and no-one really knows if it will succeed, especially after everyone else has eaten the big-64bit-commerical-machine market.
(Oh, and they also have nicely overlapping desktop and laptop ranges too at the low-end.)
So what are they going to try and sell you? They have *three* processor families, all officially to be replaced by something that doesn't work yet, two Unices, VMS, Windows, and maybe Linux. On top of this they need to unify the groups of people making these things, including finishing the digestion of DEC.
This is just horrible. If you go to Sun, you know they're going to sell you a big SPARC box running Solaris. If you go to IBM you know they wrote the book on big machines and reliability. If you go to HPaq, it's probably going to be because you want a printer or a commodity PC, because the rest of their range is just completely confused.
But. They have PC businesses which will overlap horribly and are anyway in a low-margin business they do not dominate. They have at least a couple of overlapping Unixes as well as, I guess, Linux on overlapping hardware platforms. They have a dying CPU (alpha) a supposed-to-be-moribund one (PA-RISC), a big investment in a third CPU which hasn't done very well so far and might well fail (IA64), and a CPU they don't control (x86). They do have some good printers.
I can't work out what I would buy from HP/Compaq other than printers. If I wanted high-end kit I'd go to Sun or IBM: Sun have a coherent story about Unix and are getting good at support, IBM wrote the book on high-end machines and support. If I wanted consultancy I'd go to IBM or maybe Sun or someone platform-independent. If I wanted test-equipment... oh they sold that division already.
An 8-16 CPU machine is not high end. A 60-100 CPU database machine such as an E10k or E15k is high end, and I suspect that really a lot of Sun's profit comes from this area. Sun's only serious competition here is IBM (well, perhaps better to phrase it as Sun are now competing with IBM in this area, as it's historically been IBM's baby) and I suspect the boxes cost kind of the same amount.
Someone did an interesting sum the other day: last year Sun sole ~100,000 cobalt boxes world wide. Maybe they made $200 per box, so that's $20,000,000 profit. In the same period they sold ~100 E10ks in the UK, at maybe $2,000,000 profit per machine. So *in the UK* they made $20,000,000 on E10K sales: the same as they made on cobalts world-wide. You can do similar sums for other low-end boxes.
I think this misses the point. Locust has about 600 users, who pay L3/month. That's L1800/month income. The zero-point cost for a billing system is probably more than this. Billing in arrears would also expose Locust to the risk of default which they don't currently have. Whether or not Locust *wants* to charge per message it probably cannot do so and remain anything like it is at present: it either would need to grow enormously to justify the overhead of the billing system, or it would need to crank up the base charge by a large factor.
Really, billing is viable if it's either flat-rate, in advance (like locust currently is), or if you're huge (like Orange). So what Orange are *actually* saying is that they just aren't interested in small operations: unless you can afford a billing system they don't care.
This is actually very pertinent to the net: the reason the net as we know it today exists is because people could do small-scale experiments without having to worry about billing. You could write a system to send messages from one machine to another without having to stress away about charging per message, and suddenly you've invented email or news, and later on you could invent some crappy little SGML-based networked hypertext system and arrogantly call it the `world wide web' when it only ran at CERN anyway, and you could do these kinds of experiments because you weren't getting billed per packet, and so you didn't have to worry about passing on that cost to the users. An argument that's often heard in the UK is that the net took off better in the US because *local calls are free*.
The moment you have to worry about billing you're in an entirely different place. You have to make business cases and worry about risk of default. Worst of all you have to have a billing system which means nightmare database hell and lots of paper and so on. If your basic transaction is very small, like an SMS or an email, or a packet, you stand a serious chance of your costs being completely dominated by the billing overhead. I don't know the figures for telcos but I bet a really large chunk of their profit is eaten by the billing system.
I think there's really a lot of evidence that having to worry about billing simply stifles a lot of innovation. Of course, Orange can say `well, so what?', and that is their right. But I think its a catastrophically dumb decision, because they (and the other telcos) really need to foster innovation, because no one
really has much idea where to go next. They've kind of done voice, since everyone now has a mobile phone (maybe not yet in the US, but Europe is pretty saturated). SMS was this thing that no-one saw coming that has been hugely successful, but it works fine on 2G networks. So they've now spent enormous money on 3G and they really have no idea what to do with it - video is not going to be that interesting, neither is the web, and no one knows what is, really).
But they can't see this - they're so panic stricken because they've spent this huge amount of money, that they are obsessing away about making everything they do profitable and trying to rake in money from SMS traffic (which, really, must be a big money earner: people send billions and billions of SMSs), instead of actually thinking a bit and allowing some lunatic in a basement to play with some idea without having to buy Oracle to do the billing. Of course they probably don't want to allow a huge company to do this, but that's easily arranged by just throttling the bandwidth that Locust (say) can have: then they can't grow beyond an experiment.
What is saddest of all is that they are missing a huge trick here. The problem is that billing costs don't scale down. But the telcos already have a billing system, they are large so the costs aren't too bad, and they have this really big stick to beat their customers with: pay up or we cut off your phone. So the obvious thing for them to do is to get into bed with the little innovative people to provide a billing service, which they can do at small extra cost, and which would enable innovation to procede without the crippling overhead. Even better, you only get the service if you're a customer of the telco concerned, because they need to be able to bill you, so you probably move your phone to them too. This probably isn't right for locust-as-it-stands, but some kind of semi-locust type thing could do this, if only the telcos had half a brain between them.
--tim
Re:It's a damn scooter
on
This is IT?
·
· Score: 1
I'm a college student, and until last year I lived about a 15 minute walk away from campus. That meant I spent half an hour to an hour every day walking between class and home. This thing would have therefore saved me 10-20 minutes per day of walking time-- not a huge savings, but non-trivial. In addition, I make short trips around campus that take 5-10 minutes of walking. This thing could cut those times in half at least.
Of course those 10-20 minutes a day might also be thought of as getting some exercise, which you now won't get. So you'll end up fatter and less fit, more likely to have heart trouble, and more likely to die younger.
Oh, I know why he's a billionaire thank you: because selling lowest-common-denominator solutions in large numbers is a good way of getting rich. There's this guy Henry Ford who worked that out.
There's a weird thing about software: everyone thinks that `popular' implies `good' when all it actually implies is `good at being popular'. Should we apply this logic to, say, drugs?
(And fortunately I'm not a graduate student or any kind of academic)
The principle is called "designing for the common case". Sure, it can be useful to have active windows not on top when you're copying text from one window to another, but what about when you're not? Usually, when you want to activate a window, you want to bring it to the front at the same time, and you want to be able to do so with a single click anywhere in the window. Microsoft sets itself up this way because they've done actual work to find out whether it's better or not, rather than ad hoc theorising.
I think it's actually called designing for the lowest common denominator. People who are not really used to doing complicated things with computers find it confusing to be able to have the non-topmost window active. They also find it confusing if just moving the mouse over a window makes it active rather than having to click in it. Microsoft sell to this market.
But, you know, some people do very complicated things with computers and spend a lot of time using them. Some people may have two or more editor windows on their screen with just enough text in them so they can keep track of what they are doing, and another partly shaded window with, say a shell, or the buttons that crank up the compiler. It is a real pain to those people that moving the focus to one of these partly shaded windows so they can type C-P RET or whatever mucks around with the stacking order, because they do this every few seconds.
This whole `designing for the common case' story is a myth that UI `professionals' like to spread, because they can't quite understand that people use computers in different ways. Do you think racing cars are designed the same way as the thing you drive to work? Why not? Maybe the people who use racing cars do different things with them and have different requirements, and they are also a lot more competent at making a car go along. Well, now maybe people who program a lot do different things with their computers than people who just surf the web all day, and myabe they are more confident at driving the machine too, and don't need some bogus `desktop metaphor' to help them along.
Lisp is a language (family of languages, perhaps) which strongly encourage people to think about languages - one common approach to solving problems in Lisp is to build up the language gradually until it matches the problem. This is why macros ans ysntactic flexibility are so important to Lisp. So Lisp people tend to be people who are interested in language, and interested in being able to express themselves clearly in the language. I think this correlates with being articulate in human languages, too.
The above quoted lisp sexp could mean almost anything or nothing, depending on the context in which that sexp occurs. (Is it data? Is it code? In what evaluation context will it be processed?). Java, C, and most other languages at least give you more distinct contextual tokens to guide you in your understanding.
This is a good point, but it's true largely because, in these languages, there are many less possibilities, so it's possible to provide much more syntactic help: you pretty much know that what you are looking at is code, and in fact code in the base language. You don't really have to worry if
{ for (int i = 0;...)... ;... }
is code or data or code in some other language, you know what it is.
But things aren't like that in Lisp, because there's such extraordinary syntactic flexibility. What is this?
Code? Data? In fact its code, in one of the many html-generating lisp systems. Note whan I say it's code, I mean that: this is not some chunk of data that is parsed to produce HTML at runtime, this is expanded at compile time into a bunch of calls to emit appropriate tags and so on, and it can be freely mixed with `ordinary' Lisp code.
This is actually data, it's a typical (slightly cleaned up, real ones tend to have various string grut in them) parse tree from a system which parses TML, which is a simpler HTML/XML syntax.
Sure, Lisp could have all sorts of `better' support for writing ordinary code, but then all this stuff would be much harder, the same way it's hard in, say, C. In fact there have been many more-conventional syntaxes for Lisp systems over the years. None of them have really caught on because the syntactic flexibility turns out to matter, and Lisp's minimal-commitment syntax is what gives you that.
While they definitely did have overlapping windows, at least some systems which ran on these machines had strange restrictions as to output and stacking order.
I *think* the restriction was that, in order to make things quick, output happened to the physical screen bitmap, so any window that got output needed to be at the top of the stacking order, so it was fully-exposed. I forget what happened if a window was not at the top of the stack, but I think it got automatically raised when output happened, and this was mildly annoying if you had a lot of windows open.
I suspect that this is what is being misremembered by the original poster - the apple people probably wanted to allow output to non-fully-exposed windows and this was hard.
Really now, just post your text to Usenet or Gnutella and it'll get distributed across a vast network of computers and output to every kind of medium from hard disks, CDs, tapes, flash memory, etc..
Well, I can't find a lot of usenet stuff from the early 90s, and last time I looked neither could deja/google. That's 6-10 years: not too good.
The issue is not how easy it is to keep things, but how easy it is to destroy them.
For film, keeping images is easy - just keep the film - but destroying them is quite hard - you have to physically destroy the film. Destroying selectively is even harder - you have to cut the film up and keep only some frames. The balance is such that almost everybody keeps everything - film is pretty compact and stores well, so most people just keep all the images they take. Even in a less-compact form like mounted slides, I keep everything - even if I wanted to throw stuff away I don't have the energy.
For digital media, destruction is trivial: just a keystroke or button-press will destroy an image or a set of images for ever. It's so easy to do this that it seems likely that almost everyone will do it, and a lot of images which are otherwise kept by default will be destroyed, because it's so easy.
An example that people are familiar with is email: how many people keep all their mail? I bet not many - most people probably delete almost everything, because it's so easy to do - it's easy to keep it - storage for mail really is cheap now - but it's also easy to destroy it. (I decided to keep all my mail a couple of years ago, not because I think it matters, but because I got frustrated because I kept finding that I'd deleted stuff I actually needed, but I only made it work by using procmail to file copies of everything in a safe place, and I bet I'm in a fairly small minority.) And what about news, who keeps that?
(You are also, I think, wrong in some current details: I doubt digital media is cheaper than film yet, given that 20 good-quality scans from 6x7 negs are over 400Mb. This might be cheaper than film in the form of CD or cheapo disk storage, it is certainly much more expensive in the form of storage you can put in a camera. Of course this will change over time, so it's not really an important point.)
Well, first of all they need to be able to recognise the pad. It could be, say, the low-order bits on a commercial CD or DVD, or any number of other sources of near-random information. It could be quite hard to recognise the pad. The index isn't too hard either, use the current Unix time or something to index into it (OK you can only send one bit a second this way, but we're not talking about vast amounts of information flying around). Since there's just huge amounts of commercial digitised analogue stuff around there's a lot to choose from, and of course you don't even need to have the thing physically at any given time, since you can just buy another one.
Apart from the correction you made yourself (CMUCL is actively maintained, and CLISP beats Python), there's an underlying truth in what you say: yes, the best CL implementations are probably commercial.
But that's nothing new. The best C implementations are also commercial. gcc may run everywhere, but it typically doesn't compete too well in terms of performance with a vendor's compiler. Similarly, I guess, for almost all other languages other than those (like perl or Python) defined by a single implementation.
Correct me if I'm wrong, but the actual typeface and point size of the individual letters in a plain text file are identical. So if there are 26 letters in the alphabet, and we count the insertion of spaces and punctuation, there are finitely many possible combinations of letters and spaces, seeing as the notes themselves are undifferentiated in any aspect but character set and character code. And, compared to the total number of combinations of possible images (which for all intents and purposes is infinite), this number is rather small. Seen in this light, the argument that some of these texts are protected works of creativity, capable of being--nay, supposed to be--charged for seems absolutely absurd!
Erm, which dot-coms were making billions a year ago? Which dot-coms were making any money at all a year ago? Of course some were, but I doubt any were making billions.
What I think you meant to say is that the speculative bubble has burst, and people are no longer able to make money on share-price rises without actually turning any kind of profit. I guess that the people who either were taken in by the con or weren't, but didn't sell fast enough to walk away with cash, are feeling pretty bad just the same though.
Slashdot Mirror
I'm not sure that the rules are the same throughout Europe, but they tend to be significantly nastier here than in the US. Often there is no equivalent of Chapter 11. There have been a fair few suggestions recently to make the EU rules (a) the same as each other and (b) more like the US ones.
Not in the context of your comment.
Scheme, as a language, is defined as properly tail recursive, and defines iterative constructs in terms of tail-recursive ones, whereas most other Lisp implementations (specifically Common Lisp and Emacs Lisp) are not defined this way, and do not do so. Since implementations may not eliminate tail calls, you definitely need to use iterative constructs if you want your code not to blow up in your face.
In fact Common Lisp has one of the most powerful iterative constructs I know - the LOOP macro.
You should talk to some investment bankers or large telcos sometime.
And you can, of course do this without the OS going down, right? Because it's costing you a million dollars an hour while it's down?
And you can get people to bring you replacement parts at 4am, for a 5 year old system?
And, oh yes, you've solved the problems associated with running big transactional databases on farms of PCs?
Last year, Sun sold ~100,000 cobalts world wide, making maybe ~$100 per box. That's ~$10,000,000.
Last year they sold ~100 E10ks in the UK, making ~$1,000,000 per system. That's ~$100,000,000. In the UK.
No, they aren't `about to drop their last products', whatever that means. Don't be an idiot.
They might leap, they might not. They might not be able to (conflicting options or features or not want to or it might be difficult or any number of other problems might arise.
Go and look at the roots of the Xemacs / FSF Emacs saga for how hard it can be.
Yeah, sorry, I meant `exist as a viable commercial architecture' by `exist yet'.
The combined company has (among others):
So what are they going to try and sell you? They have *three* processor families, all officially to be replaced by something that doesn't work yet, two Unices, VMS, Windows, and maybe Linux. On top of this they need to unify the groups of people making these things, including finishing the digestion of DEC.
This is just horrible. If you go to Sun, you know they're going to sell you a big SPARC box running Solaris. If you go to IBM you know they wrote the book on big machines and reliability. If you go to HPaq, it's probably going to be because you want a printer or a commodity PC, because the rest of their range is just completely confused.
But. They have PC businesses which will overlap horribly and are anyway in a low-margin business they do not dominate. They have at least a couple of overlapping Unixes as well as, I guess, Linux on overlapping hardware platforms. They have a dying CPU (alpha) a supposed-to-be-moribund one (PA-RISC), a big investment in a third CPU which hasn't done very well so far and might well fail (IA64), and a CPU they don't control (x86). They do have some good printers.
I can't work out what I would buy from HP/Compaq other than printers. If I wanted high-end kit I'd go to Sun or IBM: Sun have a coherent story about Unix and are getting good at support, IBM wrote the book on high-end machines and support. If I wanted consultancy I'd go to IBM or maybe Sun or someone platform-independent. If I wanted test-equipment... oh they sold that division already.
They are a dead company.
Well, Sun and IBM think it's a good thing, anyway. I guess they must be right...
An 8-16 CPU machine is not high end. A 60-100 CPU database machine such as an E10k or E15k is high end, and I suspect that really a lot of Sun's profit comes from this area. Sun's only serious competition here is IBM (well, perhaps better to phrase it as Sun are now competing with IBM in this area, as it's historically been IBM's baby) and I suspect the boxes cost kind of the same amount.
Someone did an interesting sum the other day: last year Sun sole ~100,000 cobalt boxes world wide. Maybe they made $200 per box, so that's $20,000,000 profit. In the same period they sold ~100 E10ks in the UK, at maybe $2,000,000 profit per machine. So *in the UK* they made $20,000,000 on E10K sales: the same as they made on cobalts world-wide. You can do similar sums for other low-end boxes.
--tim
I think this misses the point. Locust has about 600 users, who pay L3/month. That's L1800/month income. The zero-point cost for a billing system is probably more than this. Billing in arrears would also expose Locust to the risk of default which they don't currently have. Whether or not Locust *wants* to charge per message it probably cannot do so and remain anything like it is at present: it either would need to grow enormously to justify the overhead of the billing system, or it would need to crank up the base charge by a large factor.
Really, billing is viable if it's either flat-rate, in advance (like locust currently is), or if you're huge (like Orange). So what Orange are *actually* saying is that they just aren't interested in small operations: unless you can afford a billing system they don't care.
This is actually very pertinent to the net: the reason the net as we know it today exists is because people could do small-scale experiments without having to worry about billing. You could write a system to send messages from one machine to another without having to stress away about charging per message, and suddenly you've invented email or news, and later on you could invent some crappy little SGML-based networked hypertext system and arrogantly call it the `world wide web' when it only ran at CERN anyway, and you could do these kinds of experiments because you weren't getting billed per packet, and so you didn't have to worry about passing on that cost to the users. An argument that's often heard in the UK is that the net took off better in the US because *local calls are free*.
The moment you have to worry about billing you're in an entirely different place. You have to make business cases and worry about risk of default. Worst of all you have to have a billing system which means nightmare database hell and lots of paper and so on. If your basic transaction is very small, like an SMS or an email, or a packet, you stand a serious chance of your costs being completely dominated by the billing overhead. I don't know the figures for telcos but I bet a really large chunk of their profit is eaten by the billing system.
I think there's really a lot of evidence that having to worry about billing simply stifles a lot of innovation. Of course, Orange can say `well, so what?', and that is their right. But I think its a catastrophically dumb decision, because they (and the other telcos) really need to foster innovation, because no one
really has much idea where to go next. They've kind of done voice, since everyone now has a mobile phone (maybe not yet in the US, but Europe is pretty saturated). SMS was this thing that no-one saw coming that has been hugely successful, but it works fine on 2G networks. So they've now spent enormous money on 3G and they really have no idea what to do with it - video is not going to be that interesting, neither is the web, and no one knows what is, really).
But they can't see this - they're so panic stricken because they've spent this huge amount of money, that they are obsessing away about making everything they do profitable and trying to rake in money from SMS traffic (which, really, must be a big money earner: people send billions and billions of SMSs), instead of actually thinking a bit and allowing some lunatic in a basement to play with some idea without having to buy Oracle to do the billing. Of course they probably don't want to allow a huge company to do this, but that's easily arranged by just throttling the bandwidth that Locust (say) can have: then they can't grow beyond an experiment.
What is saddest of all is that they are missing a huge trick here. The problem is that billing costs don't scale down. But the telcos already have a billing system, they are large so the costs aren't too bad, and they have this really big stick to beat their customers with: pay up or we cut off your phone. So the obvious thing for them to do is to get into bed with the little innovative people to provide a billing service, which they can do at small extra cost, and which would enable innovation to procede without the crippling overhead. Even better, you only get the service if you're a customer of the telco concerned, because they need to be able to bill you, so you probably move your phone to them too. This probably isn't right for locust-as-it-stands, but some kind of semi-locust type thing could do this, if only the telcos had half a brain between them.
--tim
Of course those 10-20 minutes a day might also be thought of as getting some exercise, which you now won't get. So you'll end up fatter and less fit, more likely to have heart trouble, and more likely to die younger.
I'd rather walk.
Oh, I know why he's a billionaire thank you: because selling lowest-common-denominator solutions in large numbers is a good way of getting rich. There's this guy Henry Ford who worked that out.
There's a weird thing about software: everyone thinks that `popular' implies `good' when all it actually implies is `good at being popular'. Should we apply this logic to, say, drugs?
(And fortunately I'm not a graduate student or any kind of academic)
I think it's actually called designing for the lowest common denominator. People who are not really used to doing complicated things with computers find it confusing to be able to have the non-topmost window active. They also find it confusing if just moving the mouse over a window makes it active rather than having to click in it. Microsoft sell to this market.
But, you know, some people do very complicated things with computers and spend a lot of time using them. Some people may have two or more editor windows on their screen with just enough text in them so they can keep track of what they are doing, and another partly shaded window with, say a shell, or the buttons that crank up the compiler. It is a real pain to those people that moving the focus to one of these partly shaded windows so they can type C-P RET or whatever mucks around with the stacking order, because they do this every few seconds.
This whole `designing for the common case' story is a myth that UI `professionals' like to spread, because they can't quite understand that people use computers in different ways. Do you think racing cars are designed the same way as the thing you drive to work? Why not? Maybe the people who use racing cars do different things with them and have different requirements, and they are also a lot more competent at making a car go along. Well, now maybe people who program a lot do different things with their computers than people who just surf the web all day, and myabe they are more confident at driving the machine too, and don't need some bogus `desktop metaphor' to help them along.
I think it helps, yes.
Lisp is a language (family of languages, perhaps) which strongly encourage people to think about languages - one common approach to solving problems in Lisp is to build up the language gradually until it matches the problem. This is why macros ans ysntactic flexibility are so important to Lisp. So Lisp people tend to be people who are interested in language, and interested in being able to express themselves clearly in the language. I think this correlates with being articulate in human languages, too.
This is a good point, but it's true largely because, in these languages, there are many less possibilities, so it's possible to provide much more syntactic help: you pretty much know that what you are looking at is code, and in fact code in the base language. You don't really have to worry if
is code or data or code in some other language, you know what it is.
But things aren't like that in Lisp, because there's such extraordinary syntactic flexibility. What is this?
Code? Data? In fact its code, in one of the many html-generating lisp systems. Note whan I say it's code, I mean that: this is not some chunk of data that is parsed to produce HTML at runtime, this is expanded at compile time into a bunch of calls to emit appropriate tags and so on, and it can be freely mixed with `ordinary' Lisp code.
Or this:
This is actually data, it's a typical (slightly cleaned up, real ones tend to have various string grut in them) parse tree from a system which parses TML, which is a simpler HTML/XML syntax.
Sure, Lisp could have all sorts of `better' support for writing ordinary code, but then all this stuff would be much harder, the same way it's hard in, say, C. In fact there have been many more-conventional syntaxes for Lisp systems over the years. None of them have really caught on because the syntactic flexibility turns out to matter, and Lisp's minimal-commitment syntax is what gives you that.
While they definitely did have overlapping windows, at least some systems which ran on these machines had strange restrictions as to output and stacking order.
I *think* the restriction was that, in order to make things quick, output happened to the physical screen bitmap, so any window that got output needed to be at the top of the stacking order, so it was fully-exposed. I forget what happened if a window was not at the top of the stack, but I think it got automatically raised when output happened, and this was mildly annoying if you had a lot of windows open.
I suspect that this is what is being misremembered by the original poster - the apple people probably wanted to allow output to non-fully-exposed windows and this was hard.
Well, I can't find a lot of usenet stuff from the early 90s, and last time I looked neither could deja/google. That's 6-10 years: not too good.
I think you've completely missed the point.
The issue is not how easy it is to keep things, but how easy it is to destroy them.
For film, keeping images is easy - just keep the film - but destroying them is quite hard - you have to physically destroy the film. Destroying selectively is even harder - you have to cut the film up and keep only some frames. The balance is such that almost everybody keeps everything - film is pretty compact and stores well, so most people just keep all the images they take. Even in a less-compact form like mounted slides, I keep everything - even if I wanted to throw stuff away I don't have the energy.
For digital media, destruction is trivial: just a keystroke or button-press will destroy an image or a set of images for ever. It's so easy to do this that it seems likely that almost everyone will do it, and a lot of images which are otherwise kept by default will be destroyed, because it's so easy.
An example that people are familiar with is email: how many people keep all their mail? I bet not many - most people probably delete almost everything, because it's so easy to do - it's easy to keep it - storage for mail really is cheap now - but it's also easy to destroy it. (I decided to keep all my mail a couple of years ago, not because I think it matters, but because I got frustrated because I kept finding that I'd deleted stuff I actually needed, but I only made it work by using procmail to file copies of everything in a safe place, and I bet I'm in a fairly small minority.) And what about news, who keeps that?
(You are also, I think, wrong in some current details: I doubt digital media is cheaper than film yet, given that 20 good-quality scans from 6x7 negs are over 400Mb. This might be cheaper than film in the form of CD or cheapo disk storage, it is certainly much more expensive in the form of storage you can put in a camera. Of course this will change over time, so it's not really an important point.)
Well, first of all they need to be able to recognise the pad. It could be, say, the low-order bits on a commercial CD or DVD, or any number of other sources of near-random information. It could be quite hard to recognise the pad. The index isn't too hard either, use the current Unix time or something to index into it (OK you can only send one bit a second this way, but we're not talking about vast amounts of information flying around). Since there's just huge amounts of commercial digitised analogue stuff around there's a lot to choose from, and of course you don't even need to have the thing physically at any given time, since you can just buy another one.
But that's nothing new. The best C implementations are also commercial. gcc may run everywhere, but it typically doesn't compete too well in terms of performance with a vendor's compiler. Similarly, I guess, for almost all other languages other than those (like perl or Python) defined by a single implementation.
Correct me if I'm wrong, but the actual typeface and point size of the individual letters in a plain text file are identical. So if there are 26 letters in the alphabet, and we count the insertion of spaces and punctuation, there are finitely many possible combinations of letters and spaces, seeing as the notes themselves are undifferentiated in any aspect but character set and character code. And, compared to the total number of combinations of possible images (which for all intents and purposes is infinite), this number is rather small. Seen in this light, the argument that some of these texts are protected works of creativity, capable of being--nay, supposed to be--charged for seems absolutely absurd!
Erm, which dot-coms were making billions a year ago? Which dot-coms were making any money at all a year ago? Of course some were, but I doubt any were making billions. What I think you meant to say is that the speculative bubble has burst, and people are no longer able to make money on share-price rises without actually turning any kind of profit. I guess that the people who either were taken in by the con or weren't, but didn't sell fast enough to walk away with cash, are feeling pretty bad just the same though.
I think it's at least 43! The 40th anniversary was 1998.