More to the point, no mention I could see of the ZX81, which must have been easily the most popular PC in the world for a while (its sales dwarfed anything Acorn produced, and Britain at this time had far higher percentage domestic PC ownership, largely thanks to Sinclair).
The main reason for mentioning Acorn is not market share in the PC market, but because it led to the creation of the ARM processor which has much of the embedded market pretty much sewn up. But of course ARM is a British company too.
In other words, it's another one of those "if it wasn't made by an American it never existed" articles.
The fact that OpenOffice (an especially poor choice of OSS poster child, but whatever) is even within an order of magnitude of Office (with literally hundreds of developers and tens of millions of dollars behind it) is simply astonishing.
Not really: most of the development happened when it was a commercial product (which had a fairly large niche market in Germany, AFAIR). For me, the damning thing about the whole OO saga from the OSS point of view is how little truly revolutionary has happened since Star Office went open source.
And before all the OSS groupies throw a hissy fit, have a look here for Linus totally agreeing with the statement
One explanation for why the Linux model has worked best with developer-type software - Web servers, compilers, the OS itself - seems to be that in these areas, there is much intersection between the developer and user bases. End-users contribute, actively participating in the community. In other areas - office software such as professional wordprocessors - the Linux model has had much less success. (StarOffice doesn't count as a "Linux model" creation, since it is proprietary and backed by completely commercial software.) Isn't this because in such markets end-users tend to be completely passive consumers?
The quantity of mechanical energy which could be produced in the atmosphere is 6000 times greater than the mechanical energy produced by humans.
Well, yes, but is that either surprising or remotely useful? I mean, the sun produces millions of times more energy than we could ever possibly use (at source), but harnessing that energy is the tricky bit. The oceans contain massive amounts of water, but that is unlikely to prevent global water shortages in some of our lifetimes. And if we extracted all the available energy from moving air, worldwide, do we think this might have some interesting side effects?
You seem to be arguing with someone else... I didn't say porn should be banned, and that isn't what the OP is about either. The OP is about making sources of porn easier to detect by technical means. If you were looking for porn, I guess that could be a good thing...
To put it another way, how is this banning porn any more than anti-spam measures such as SPEWS ban spammers? In both cases, it's about enabling people to make choices - the categorisation itself is neutral (and, if you don't think the categories are reliable or useful, you can ignore them).
There are one or two people who would like to achieve this without unplugging their computers long term. One example: I ran an Internet awareness day in our village a couple of years ago, with clean linux terminals, and one of the kids experimenting with Google managed to fill her screen with hardcore porn in about 30 second flat (to her considerable and audible distress). Sure, plenty of kids go looking, but that isn't what happened in this case.
True, but just getting US-generated and US-hosted porn under control, as well as porn passing through US-owned ISPs, would account for quite a lot of sites, and an awful lot of the sites that tend to pop up in Google. America is regularly cited as one of the obstacles to dealing with Internet porn - if it took any steps, however technically incompetent, to address the issue, it would make an enormous difference.
I realise that restricting access to porn may not be a subject dear to the heart of all/.ers, but I have the impression that most of the rest of this thread is going to boil down to "no-one can do a thing about porn, la la la la I can't hear you", when the reality is that a lot of people around the world would like to see the present situation changed, and, one way or another, sooner or later, that will result in legislation. And if a solution is finally imposed, it may well turn out to be as draconian as the French government's anti-nazi legislation, which has been successfully imposed on Yahoo.
Well, RISC OS did launch a mom and pop chip designer, run by people with no experience of processor design, called Acorn Risc Machines, or ARM (the A changed in meaning along the way), and whose chip design now appears in more devices worldwide than Intel processors. Has linux turned the hardware world upside down yet?
The really ironic thing is that RISC OS was supposed to be Un*x, but whichever American university Acorn subcontracted to write it (California?) dropped the ball. It's a pity, as the original spec would have been OS X on RISC hardware in the mid 80s.
I used my RISC PC only the other day to do some vector graphics work that is still a pain with Linux. In many ways it's still a wonderful platform, but, realistically, it has no mainstream future whatever the licencing arrangements, because there will never be enough people writing for it.
This has to be the worst thing that could possibly happen
Not according to Nick Page, who made most of the stuff. As he pointed out in the BBC article, it's dreadful news, but nothing compared to the awfulness of most of the other items of news today.
You're right, it was a Symbolics, but pre-Genera. My understanding was that most of the OS-end of things was common with the MIT version - we beta tested a Texas Instruments clone of the Symbolics that was remarkably similar, for example. I think my general point stands that comparing this codebase with the JVM is a rather odd thing to do.
For a start, it's source code. It compiled for a processor designed specifically to run Lisp (36-bits anyone?), so the object code was quite compact.
Second, it did stuff in the 80s that are still hard to do today. For example, how many development systems allow you to locate a piece of windowing behaviour that you like, anywhere in the OS, and import that behaviour to your own code with a few mouse clicks? Or drop into microcode when necessary? The JVM is smaller because it does a whole lot less.
Third, the machine I used cost £1M at the time, so I don't think they were worried about the odd extra RAM chip. Lisp uses lots of RAM when it runs, no matter how compact the code.
They were fantastic machines - in the year Apple released their first laser printer and IBM were talking seriously about a windowing environment, these guys were using a 68000 just to bootstrap the main processor, and there weren't a lot of other 24-bit displays around, let alone full-blown rendering packages to take advantage of them.
If the huge middle class you refer to can afford these $200 PCs, enough of them could afford $400 PCs to add a nought to the current 13 million, unless the Indian income bell curve is a very odd shape indeed. The quality of infrastructure makes a difference, as does the pressure to have a PC in the first place. Most technology has to reach a threshold at which you are suddenly out of the loop if you don't have it. I'm old enough to remember this happening in the space of 3 months with fax machines in France. To put it another way, it's no fun having an email address at home if you have no-one to write to. If India has hit that threshold for PCs, the difference between $200 and $400 won't make a massive difference. If it hasn't hit that threshold, the price won't make a massive difference either.
The PC penetration in India is very low, at a measly 13 million, due to the high costs involved."
Doesn't the fact that many of the population are struggling to feed their families or dying of easily-curable diseases, and that many more are illiterate or live in places without a reliable electricity supply, let alone an Internet connection, have something to do with it too? The Indians who are on the receiving end of outsourcing are a small minority. Also, how interesting is a laptop if you rarely leave your village, compared with, say, a cybercafe (costing typically a few cents an hour)?
I'm waiting for SPEWS to start escalating to blocking all machines on the same 32-bit range as a known spammer, just to make really sure they get him. It would make about as much sense as their current policy.
The Telewest customers I know have no other broadband provider to swap to, so pressure on Telewest through this block is zero. All this kind of thing does is penalise innocent users and make a few self-righteous slashdotters feel more self-righteous - unless it's their mail that is being bounced of course.
I was on the receiving end of SPEWS "let's kick a million people for the hell of it" blocks a while back. It was seriously annoying in principle, but, in practice, I think we had 3 mails bounced among the tens of thousands our servers handle, which says something about how many people in the real world use SPEWS. I guess most of them have realised that a random number generator is about as selective.
Why are these cheap entry-level systems always targeted at the "Third World", rather than poor people here in the US?
Because, for many users, this sort of technology just cannot deliver the user experience they want.
I've spent 3.5 years running a cybercafe in France that sounds remarkably like their proposed setup - 10 diskless terminals connected to a fast Linux server. For many things it's fine. But try watching a realplayer video over a remote X session and watch the network saturate. This proposal uses gigabit cards. OK, that will help a bit, but you still hit a wall with many applications. Note that whatever whizzy (read "expensive") switch gear you have, all the packets either start from or go to the same NIC on the same server, and, ultimately, that's your bottleneck.
If you have 200 terminals, as the article suggests, that means 5Mbit/sec uncontended bandwidth per terminal, assuming your gigabit setup will run smoothly at 100% load. 2Mbit/sec sounds more likely to me. Try running X over a flakey 802.11b connection and you may spot the problem.
This will work fine for WP and text-based browsing (the size of OO is a red herring as it lives on the server, not the client). But any kind of large bitmapped image, let alone animation, will kill it.
Yes, yes, I know, this is not X, it's sending pixel images. So it's doing more or less what Citrix does. Try opening an image of random pixels full-screen over a Citrix session and watch the system hang for several seconds.
There's no way around the basic facts. Networks are much slower than hard discs. You can compress a lot of images very efficiently, and you can optimise your compression to handle GUI furniture and so on, but arbitary graphical data doesn't compress well, and the time taken to send it is simply the size of the compressed image divided by the network bandwidth. And the much-touted dumbness of the terminal radically reduces your options for context-sensitive compression.
In other words, it's a low-price solution offering a low-comfort user experience. The assumption tends to be that "poor people" are simple souls who will settle for basic services. Apart from being somewhat patronising, that assumption just doesn't tie up with my experience. Poverty tends to correlate with limited education and limited experience of computer systems. Poor people expect everything to "just work", and are not going to be pleased to learn that they can't use certain sites because of some technical consideration.
Incidentally, my experience suggests that they are also more likely to have trouble using any software other than whatever software they have used before (inevitably Windows), so expect lots of support calls about switching from MSN to GAIM, for example. And, yes, Africans do use Instant Messaging a lot if they can, as it's much much cheaper than international phone calls.
Usually you don't have a terminal service just for the sake of it but for accessing some kind of application or the other.
That's a very un*x-eye view of things. There are plenty of companies connecting large numbers of thin terminals to TS or Citrix servers to run their entire desktop, just because of the intrinsic benefits of thin clients (simpler maintenance, easier to tie down).
Others have mentioned specific benefits of TS, which include persistent sessions and local printers. Another advantage I noticed myself is that X really is a dog over slow Internet connections, whereas TS is useable over dial-up and very useable over ISDN.
Now, sure, there are fixes for all the above problems with X. (And I do have some experience in this area, having used LTSP in a commercial setting for 3.5 years.) But using the Windows solution that works very nicely and out of the box isn't necessarily a dumb move.
If the app also runs on *ix or has appropriate replacements there then run it there and use one of the suggestions above
Even when the application exists for both platforms, my limited experience is that commercial apps for Un*x tend to be a lot more expensive. This was certainly true of Adobe Acrobat last time I looked - for Linux you had to buy the server version. Most Windows desktop apps will work in client-server mode over TS, so, while you may need multiple licences, that may still be cheaper than buying a un*x server pack that is priced for 500 users.
Another example I came across recently was page layout utilities for Canon colour copiers - free PC and Mac drivers, the Linux version costs $1500 for exactly the same functionality. You can get a copy of Server 2003 and a few seat licences for that kind of money, especially if you are a school or an NGO.
There are also a cluster of problems connected with how satellite connections handle character data. The satellite hop is converted from IP packets to something streaming, and then the packets are rebuilt at the other end. That process (the details of which I don't pretend to understand) works very well for streaming data, and very badly for a couple of typed characters a second.
The worst case I have come across is ssh, which can lag by 5 seconds on occasions, which I think is either because some of the handshaking takes several exchanges of packets or because the Earth station waits in the hope that a few more packets will come along to fill its "bucket". FTPing a load of small files is pretty bad too, because the handshaking for each file takes several seconds.
I'd always assumed it was the base station in Benelux. I didn't get the impression that there was a whole lot of packet manipulation going on in the satellite, but I may well be wrong.
We're not talking about TV-type satellite broadcasts here. Security is one of the alleged benefits of satellite broadband. It goes straight from the ISP to the customers, and you usually have a lot of dedicated hardware on both ends of the connection that make strong encryption easy enough to implement. I would have thought that just demultiplexing out the signal you want would take a fair bit of work. The weakest link would be between the film studio and the uplink, but my guess is that they'll put a 5m dish in the film studio parking lot.
What, when the speed of light increases a bit? I'm typing this on a terrestial satellite connection, and that's about the latency you get.
You can do VoIP over this sort of connection, you just have to get used to the fact that there's about a second's lag, and, if you can hear yourself on the remote speakers, a 2-second echo.
Here's part of my traceroute...
3 firewall (192.168.1.1) 3.067 ms 7.553 ms 12.684 ms
4 172.31.254.253 (172.31.254.253) 4.062 ms 5.199 ms 6.160 ms
5 172.16.128.19 (172.16.128.19) 631.013 ms 583.867 ms 1562.693 ms
6 172.16.64.61 (172.16.64.61) 1658.335 ms 1537.707 ms 1099.666 ms
7 172.16.100.1 (172.16.100.1) 1156.839 ms 1302.108 ms 1372.232 ms
8 iamrt01p.idc.aramiska.net (172.16.101.254) 794.488 ms 528.926 ms 569.041 ms
9 213.181.59.142 (213.181.59.142) 602.372 ms 606.007 ms 732.337 ms
10 cbu-pcbru11-pecbru21-pos31.car.belbone.be (213.181.59.198) 707.791 ms 782.617 ms 619.338 ms
11 so-6-1-2-bcr1.bru.cw.net (206.24.147.189) 743.572 ms 759.283 ms 732.355 ms
12 so-5-1-0-dcr2.par.cw.net (195.2.10.34) 906.991 ms 770.496 ms 628.550 ms
13 as0-dcr1.par.cw.net (195.2.10.161) 904.086 ms 658.604 ms 1547.670 ms
14 so-0-0-0-dcr1.was.cw.net (195.2.10.117) 670.472 ms 822.322 ms 698.954 ms
15 dcr1-so-2-0-0.Washington.savvis.net (206.24.238.17) 620.367 ms 654.039 ms 711.492 ms
16 dcr2-loopback.SanFranciscosfo.savvis.net (206.24.210.100) 848.040 ms 795.551 ms 879.444 ms
17 bhr1-pos-0-0.SantaClarasc8.savvis.net (208.172.156.198) 1360.829 ms 994.331 ms 758.175 ms
18 csr1-ve240.SantaClarasc8.savvis.net (66.35.194.34) 1306.824 ms 1061.993 ms 861.418 ms
19 66.35.212.174 (66.35.212.174) 1148.671 ms 1076.630 ms 696.751 ms
20 slashdot.org (66.35.250.150) 983.516 ms 1103.163 ms 778.011 ms
Last time I installed some plugin or other with IE (shockwave I think), Yahoo promptly took over the menu bar, and I've yet to work out how to undo the damage. So now this will happen to Firefox too? And this is something to celebrate?
But for 99.99% of computer users, it's academic whatever the OS. I've been running a Linux cybercafe for 3.5 years, and in that time I reckon we've had maybe 3 people through the door who might have been able to fiddle with our OS without having to reinstall afterwards (if I had let them). Indeed, if we're talking about kernel-level fiddling, what percentage of/.ers routinely take their patches to bits to see how they work?
If I hack a piece of proprietary code, assuming I have the means to do so, I invalidate my warranty and might get sued. If I hack a piece of OSS, I won't get sued, but I probably still invalidate the equivalent of my warranty - it's usually called "forking".
A real example, admittedly a long way from kernels. One of our customers was a long-time user of a proprietary forum package, written in Perl. When we took over hosting his site, we had a look at the code, and found some enormous security problems, which we fixed, which of course meant that we then couldn't run the usual upgrade procedures. Said customer has now moved his site over to phpBB. We found one or two problems with that, fixed them, and, hey presto, we have identical issues as soon as we want to install a "mod" or any of the official patches. The only difference is that we're legally stuffed rather than illegally stuffed:) (Much as it pains me to say this, phpBB is a much better bit of code than the perl thing they had before, and support is far better, so we've won on the exchange, but if you want off-the-shelf fixes you're still better off not touching the code yourself).
It sounds like a really bizarre argument to me. If the general population wants to browse source code on a Saturday night (and is this true even where MIT staff live?) they don't need their governments' computers to be running it, they just need to download some source code. What might be of interest to the general population is better access to what is stored using the government's OS, but "open data" is not quite the same thing.
Incidentally, I learned to program on an Acorn too, and still use my RiscPC for certain tasks. 10-character filenames weren't that ergonomic though, and of course Acorn and their largest developer could never agree what look and feel to go for anyway.
As for Perl, I'm a philosopher, not a perl programmer, and my knowledge of programming languages stops somewhere around Pascal or BASIC. But I would expect that if you tried to call an undefined variable, it wouldn't "know" what you were telling it, and would be unable to proceed.
By default it will make up a value and carry on. In a lot of situations you force perl not to do this sort of thing, because it's better for the program to crash than to improvise, but Perl does improvisation alarmingly well. And I can think of real-world situations where it would be good if people crashed rather than improvising too:)
As for your foo example, all you seem to be saying is that the office lackey has a different interface. If the lackey has never come across that bit of lexis before, he'll "crash" too. If he doesn't, I wouldn't want him touching anything in my office! If the function is around somewhere in Perl, then, depending how it is set up, there is a fair chance of it finding it. If you give it the wrong number of arguments, it will cope. If you give it the wrong type of arguments, it will cope. If you call the function in a way that throws away the result, it might complain but it will cope.
The original subject was not whether or not AI is a viable project (although your argument sounds a lot like the sort of thing Dreyfuss says about the limitations of AI), it was the extent to which human language skills relate to programming. And, while Perl is clearly not identical to French, I'm saying that the skills involved in dealing with French are more transferable to dealing with Perl than you want to admit. And I'd also say that speaking fluent English doesn't in any way guarantee that the speaker will ever be able to communicate fluently in French, let alone Chinese.
Slashdot Mirror
... we'd still be relying on SPEWS to bully innocent bystanders into bullying ISPs into shutting down spammers after the event.
More to the point, no mention I could see of the ZX81, which must have been easily the most popular PC in the world for a while (its sales dwarfed anything Acorn produced, and Britain at this time had far higher percentage domestic PC ownership, largely thanks to Sinclair). The main reason for mentioning Acorn is not market share in the PC market, but because it led to the creation of the ARM processor which has much of the embedded market pretty much sewn up. But of course ARM is a British company too. In other words, it's another one of those "if it wasn't made by an American it never existed" articles.
The fact that OpenOffice (an especially poor choice of OSS poster child, but whatever) is even within an order of magnitude of Office (with literally hundreds of developers and tens of millions of dollars behind it) is simply astonishing.
Not really: most of the development happened when it was a commercial product (which had a fairly large niche market in Germany, AFAIR). For me, the damning thing about the whole OO saga from the OSS point of view is how little truly revolutionary has happened since Star Office went open source.
And before all the OSS groupies throw a hissy fit, have a look here for Linus totally agreeing with the statement
The quantity of mechanical energy which could be produced in the atmosphere is 6000 times greater than the mechanical energy produced by humans.
Well, yes, but is that either surprising or remotely useful? I mean, the sun produces millions of times more energy than we could ever possibly use (at source), but harnessing that energy is the tricky bit. The oceans contain massive amounts of water, but that is unlikely to prevent global water shortages in some of our lifetimes. And if we extracted all the available energy from moving air, worldwide, do we think this might have some interesting side effects?
You seem to be arguing with someone else... I didn't say porn should be banned, and that isn't what the OP is about either. The OP is about making sources of porn easier to detect by technical means. If you were looking for porn, I guess that could be a good thing...
To put it another way, how is this banning porn any more than anti-spam measures such as SPEWS ban spammers? In both cases, it's about enabling people to make choices - the categorisation itself is neutral (and, if you don't think the categories are reliable or useful, you can ignore them).
It's possible to live a porn-free live
There are one or two people who would like to achieve this without unplugging their computers long term. One example: I ran an Internet awareness day in our village a couple of years ago, with clean linux terminals, and one of the kids experimenting with Google managed to fill her screen with hardcore porn in about 30 second flat (to her considerable and audible distress). Sure, plenty of kids go looking, but that isn't what happened in this case.
International adult sites not subject to US laws
True, but just getting US-generated and US-hosted porn under control, as well as porn passing through US-owned ISPs, would account for quite a lot of sites, and an awful lot of the sites that tend to pop up in Google. America is regularly cited as one of the obstacles to dealing with Internet porn - if it took any steps, however technically incompetent, to address the issue, it would make an enormous difference.
I realise that restricting access to porn may not be a subject dear to the heart of all /.ers, but I have the impression that most of the rest of this thread is going to boil down to "no-one can do a thing about porn, la la la la I can't hear you", when the reality is that a lot of people around the world would like to see the present situation changed, and, one way or another, sooner or later, that will result in legislation. And if a solution is finally imposed, it may well turn out to be as draconian as the French government's anti-nazi legislation, which has been successfully imposed on Yahoo.
Well, RISC OS did launch a mom and pop chip designer, run by people with no experience of processor design, called Acorn Risc Machines, or ARM (the A changed in meaning along the way), and whose chip design now appears in more devices worldwide than Intel processors. Has linux turned the hardware world upside down yet?
The really ironic thing is that RISC OS was supposed to be Un*x, but whichever American university Acorn subcontracted to write it (California?) dropped the ball. It's a pity, as the original spec would have been OS X on RISC hardware in the mid 80s.
I used my RISC PC only the other day to do some vector graphics work that is still a pain with Linux. In many ways it's still a wonderful platform, but, realistically, it has no mainstream future whatever the licencing arrangements, because there will never be enough people writing for it.
This has to be the worst thing that could possibly happen Not according to Nick Page, who made most of the stuff. As he pointed out in the BBC article, it's dreadful news, but nothing compared to the awfulness of most of the other items of news today.
You're right, it was a Symbolics, but pre-Genera. My understanding was that most of the OS-end of things was common with the MIT version - we beta tested a Texas Instruments clone of the Symbolics that was remarkably similar, for example. I think my general point stands that comparing this codebase with the JVM is a rather odd thing to do.
For a start, it's source code. It compiled for a processor designed specifically to run Lisp (36-bits anyone?), so the object code was quite compact.
Second, it did stuff in the 80s that are still hard to do today. For example, how many development systems allow you to locate a piece of windowing behaviour that you like, anywhere in the OS, and import that behaviour to your own code with a few mouse clicks? Or drop into microcode when necessary? The JVM is smaller because it does a whole lot less.
Third, the machine I used cost £1M at the time, so I don't think they were worried about the odd extra RAM chip. Lisp uses lots of RAM when it runs, no matter how compact the code.
They were fantastic machines - in the year Apple released their first laser printer and IBM were talking seriously about a windowing environment, these guys were using a 68000 just to bootstrap the main processor, and there weren't a lot of other 24-bit displays around, let alone full-blown rendering packages to take advantage of them.
If the huge middle class you refer to can afford these $200 PCs, enough of them could afford $400 PCs to add a nought to the current 13 million, unless the Indian income bell curve is a very odd shape indeed. The quality of infrastructure makes a difference, as does the pressure to have a PC in the first place. Most technology has to reach a threshold at which you are suddenly out of the loop if you don't have it. I'm old enough to remember this happening in the space of 3 months with fax machines in France. To put it another way, it's no fun having an email address at home if you have no-one to write to. If India has hit that threshold for PCs, the difference between $200 and $400 won't make a massive difference. If it hasn't hit that threshold, the price won't make a massive difference either.
The PC penetration in India is very low, at a measly 13 million, due to the high costs involved."
Doesn't the fact that many of the population are struggling to feed their families or dying of easily-curable diseases, and that many more are illiterate or live in places without a reliable electricity supply, let alone an Internet connection, have something to do with it too? The Indians who are on the receiving end of outsourcing are a small minority. Also, how interesting is a laptop if you rarely leave your village, compared with, say, a cybercafe (costing typically a few cents an hour)?
I'm waiting for SPEWS to start escalating to blocking all machines on the same 32-bit range as a known spammer, just to make really sure they get him. It would make about as much sense as their current policy.
The Telewest customers I know have no other broadband provider to swap to, so pressure on Telewest through this block is zero. All this kind of thing does is penalise innocent users and make a few self-righteous slashdotters feel more self-righteous - unless it's their mail that is being bounced of course.
I was on the receiving end of SPEWS "let's kick a million people for the hell of it" blocks a while back. It was seriously annoying in principle, but, in practice, I think we had 3 mails bounced among the tens of thousands our servers handle, which says something about how many people in the real world use SPEWS. I guess most of them have realised that a random number generator is about as selective.
Why are these cheap entry-level systems always targeted at the "Third World", rather than poor people here in the US?
Because, for many users, this sort of technology just cannot deliver the user experience they want.
I've spent 3.5 years running a cybercafe in France that sounds remarkably like their proposed setup - 10 diskless terminals connected to a fast Linux server. For many things it's fine. But try watching a realplayer video over a remote X session and watch the network saturate. This proposal uses gigabit cards. OK, that will help a bit, but you still hit a wall with many applications. Note that whatever whizzy (read "expensive") switch gear you have, all the packets either start from or go to the same NIC on the same server, and, ultimately, that's your bottleneck.
If you have 200 terminals, as the article suggests, that means 5Mbit/sec uncontended bandwidth per terminal, assuming your gigabit setup will run smoothly at 100% load. 2Mbit/sec sounds more likely to me. Try running X over a flakey 802.11b connection and you may spot the problem.
This will work fine for WP and text-based browsing (the size of OO is a red herring as it lives on the server, not the client). But any kind of large bitmapped image, let alone animation, will kill it.
Yes, yes, I know, this is not X, it's sending pixel images. So it's doing more or less what Citrix does. Try opening an image of random pixels full-screen over a Citrix session and watch the system hang for several seconds.
There's no way around the basic facts. Networks are much slower than hard discs. You can compress a lot of images very efficiently, and you can optimise your compression to handle GUI furniture and so on, but arbitary graphical data doesn't compress well, and the time taken to send it is simply the size of the compressed image divided by the network bandwidth. And the much-touted dumbness of the terminal radically reduces your options for context-sensitive compression.
In other words, it's a low-price solution offering a low-comfort user experience. The assumption tends to be that "poor people" are simple souls who will settle for basic services. Apart from being somewhat patronising, that assumption just doesn't tie up with my experience. Poverty tends to correlate with limited education and limited experience of computer systems. Poor people expect everything to "just work", and are not going to be pleased to learn that they can't use certain sites because of some technical consideration.
Incidentally, my experience suggests that they are also more likely to have trouble using any software other than whatever software they have used before (inevitably Windows), so expect lots of support calls about switching from MSN to GAIM, for example. And, yes, Africans do use Instant Messaging a lot if they can, as it's much much cheaper than international phone calls.
Usually you don't have a terminal service just for the sake of it but for accessing some kind of application or the other.
That's a very un*x-eye view of things. There are plenty of companies connecting large numbers of thin terminals to TS or Citrix servers to run their entire desktop, just because of the intrinsic benefits of thin clients (simpler maintenance, easier to tie down).
Others have mentioned specific benefits of TS, which include persistent sessions and local printers. Another advantage I noticed myself is that X really is a dog over slow Internet connections, whereas TS is useable over dial-up and very useable over ISDN.
Now, sure, there are fixes for all the above problems with X. (And I do have some experience in this area, having used LTSP in a commercial setting for 3.5 years.) But using the Windows solution that works very nicely and out of the box isn't necessarily a dumb move.
If the app also runs on *ix or has appropriate replacements there then run it there and use one of the suggestions above
Even when the application exists for both platforms, my limited experience is that commercial apps for Un*x tend to be a lot more expensive. This was certainly true of Adobe Acrobat last time I looked - for Linux you had to buy the server version. Most Windows desktop apps will work in client-server mode over TS, so, while you may need multiple licences, that may still be cheaper than buying a un*x server pack that is priced for 500 users.
Another example I came across recently was page layout utilities for Canon colour copiers - free PC and Mac drivers, the Linux version costs $1500 for exactly the same functionality. You can get a copy of Server 2003 and a few seat licences for that kind of money, especially if you are a school or an NGO.
There are also a cluster of problems connected with how satellite connections handle character data. The satellite hop is converted from IP packets to something streaming, and then the packets are rebuilt at the other end. That process (the details of which I don't pretend to understand) works very well for streaming data, and very badly for a couple of typed characters a second. The worst case I have come across is ssh, which can lag by 5 seconds on occasions, which I think is either because some of the handshaking takes several exchanges of packets or because the Earth station waits in the hope that a few more packets will come along to fill its "bucket". FTPing a load of small files is pretty bad too, because the handshaking for each file takes several seconds.
I'd always assumed it was the base station in Benelux. I didn't get the impression that there was a whole lot of packet manipulation going on in the satellite, but I may well be wrong.
We're not talking about TV-type satellite broadcasts here. Security is one of the alleged benefits of satellite broadband. It goes straight from the ISP to the customers, and you usually have a lot of dedicated hardware on both ends of the connection that make strong encryption easy enough to implement. I would have thought that just demultiplexing out the signal you want would take a fair bit of work. The weakest link would be between the film studio and the uplink, but my guess is that they'll put a 5m dish in the film studio parking lot.
once the latency goes down a bit
What, when the speed of light increases a bit? I'm typing this on a terrestial satellite connection, and that's about the latency you get.
You can do VoIP over this sort of connection, you just have to get used to the fact that there's about a second's lag, and, if you can hear yourself on the remote speakers, a 2-second echo.
Here's part of my traceroute...
3 firewall (192.168.1.1) 3.067 ms 7.553 ms 12.684 ms4 172.31.254.253 (172.31.254.253) 4.062 ms 5.199 ms 6.160 ms
5 172.16.128.19 (172.16.128.19) 631.013 ms 583.867 ms 1562.693 ms
6 172.16.64.61 (172.16.64.61) 1658.335 ms 1537.707 ms 1099.666 ms
7 172.16.100.1 (172.16.100.1) 1156.839 ms 1302.108 ms 1372.232 ms
8 iamrt01p.idc.aramiska.net (172.16.101.254) 794.488 ms 528.926 ms 569.041 ms
9 213.181.59.142 (213.181.59.142) 602.372 ms 606.007 ms 732.337 ms
10 cbu-pcbru11-pecbru21-pos31.car.belbone.be (213.181.59.198) 707.791 ms 782.617 ms 619.338 ms
11 so-6-1-2-bcr1.bru.cw.net (206.24.147.189) 743.572 ms 759.283 ms 732.355 ms
12 so-5-1-0-dcr2.par.cw.net (195.2.10.34) 906.991 ms 770.496 ms 628.550 ms
13 as0-dcr1.par.cw.net (195.2.10.161) 904.086 ms 658.604 ms 1547.670 ms
14 so-0-0-0-dcr1.was.cw.net (195.2.10.117) 670.472 ms 822.322 ms 698.954 ms
15 dcr1-so-2-0-0.Washington.savvis.net (206.24.238.17) 620.367 ms 654.039 ms 711.492 ms
16 dcr2-loopback.SanFranciscosfo.savvis.net (206.24.210.100) 848.040 ms 795.551 ms 879.444 ms
17 bhr1-pos-0-0.SantaClarasc8.savvis.net (208.172.156.198) 1360.829 ms 994.331 ms 758.175 ms
18 csr1-ve240.SantaClarasc8.savvis.net (66.35.194.34) 1306.824 ms 1061.993 ms 861.418 ms
19 66.35.212.174 (66.35.212.174) 1148.671 ms 1076.630 ms 696.751 ms
20 slashdot.org (66.35.250.150) 983.516 ms 1103.163 ms 778.011 ms
Last time I installed some plugin or other with IE (shockwave I think), Yahoo promptly took over the menu bar, and I've yet to work out how to undo the damage. So now this will happen to Firefox too? And this is something to celebrate?
But for 99.99% of computer users, it's academic whatever the OS. I've been running a Linux cybercafe for 3.5 years, and in that time I reckon we've had maybe 3 people through the door who might have been able to fiddle with our OS without having to reinstall afterwards (if I had let them). Indeed, if we're talking about kernel-level fiddling, what percentage of /.ers routinely take their patches to bits to see how they work?
:) (Much as it pains me to say this, phpBB is a much better bit of code than the perl thing they had before, and support is far better, so we've won on the exchange, but if you want off-the-shelf fixes you're still better off not touching the code yourself).
If I hack a piece of proprietary code, assuming I have the means to do so, I invalidate my warranty and might get sued. If I hack a piece of OSS, I won't get sued, but I probably still invalidate the equivalent of my warranty - it's usually called "forking".
A real example, admittedly a long way from kernels. One of our customers was a long-time user of a proprietary forum package, written in Perl. When we took over hosting his site, we had a look at the code, and found some enormous security problems, which we fixed, which of course meant that we then couldn't run the usual upgrade procedures. Said customer has now moved his site over to phpBB. We found one or two problems with that, fixed them, and, hey presto, we have identical issues as soon as we want to install a "mod" or any of the official patches. The only difference is that we're legally stuffed rather than illegally stuffed
It sounds like a really bizarre argument to me. If the general population wants to browse source code on a Saturday night (and is this true even where MIT staff live?) they don't need their governments' computers to be running it, they just need to download some source code. What might be of interest to the general population is better access to what is stored using the government's OS, but "open data" is not quite the same thing.
Incidentally, I learned to program on an Acorn too, and still use my RiscPC for certain tasks. 10-character filenames weren't that ergonomic though, and of course Acorn and their largest developer could never agree what look and feel to go for anyway.
As for Perl, I'm a philosopher, not a perl programmer, and my knowledge of programming languages stops somewhere around Pascal or BASIC. But I would expect that if you tried to call an undefined variable, it wouldn't "know" what you were telling it, and would be unable to proceed.
By default it will make up a value and carry on. In a lot of situations you force perl not to do this sort of thing, because it's better for the program to crash than to improvise, but Perl does improvisation alarmingly well. And I can think of real-world situations where it would be good if people crashed rather than improvising too :)
As for your foo example, all you seem to be saying is that the office lackey has a different interface. If the lackey has never come across that bit of lexis before, he'll "crash" too. If he doesn't, I wouldn't want him touching anything in my office! If the function is around somewhere in Perl, then, depending how it is set up, there is a fair chance of it finding it. If you give it the wrong number of arguments, it will cope. If you give it the wrong type of arguments, it will cope. If you call the function in a way that throws away the result, it might complain but it will cope.
The original subject was not whether or not AI is a viable project (although your argument sounds a lot like the sort of thing Dreyfuss says about the limitations of AI), it was the extent to which human language skills relate to programming. And, while Perl is clearly not identical to French, I'm saying that the skills involved in dealing with French are more transferable to dealing with Perl than you want to admit. And I'd also say that speaking fluent English doesn't in any way guarantee that the speaker will ever be able to communicate fluently in French, let alone Chinese.
Yeah, but the man running the Paris University network hasn't set up ssh yet, so what is she going to do?