- Losing your $HOME/.ssh or wherever it lives under Windows. - reconfiguring the server. - starting to use ssh for the first time to that particular server.
For example, although my home machine has been in existance for years, and one of my work systems has similarly been around for years, I had somehow never ssh-ed from this particular machine at home to that machine at work. The machines at home don't share homedirs. So I got the warning yesterday, which I was able to copy for the posting above....
If for example you have 10 workstations on one end and 10 servers on the other, you need to type "yes" a total of 100 times. If you've done that 50 times, you're open to a man in the middle attack because you're used to getting that warning every once in a while.
So yes, that's user error. However, those are worth optimizing for, because I think the crypto is safe...
Assuming the crypto is strong enough, and I think it is, the chances of getting hacked amount to "user error".
What kinds of user errors could happen?
The authenticity of host '192.168.1.1 (192.168.1.1)' can't be established.
RSA key fingerprint is 24:ba:36:a4:4b:11:59:e8:ec:dd:75:15:f2:2e:74:dc.
Are you sure you want to continue connecting (yes/no)?
If you occasionally get that, legitimately because you lose your key fingerprint hashes, you might fall for a man-in-the-middle attack. If you leave your sessions running, there will be less chances of intercepting you while you do the "connect", and the chances of you getting used to these warnings and hitting enter too quickly will also reduce.
So I'd say that for normal use, leaving them running is more secure.
(on the other hand, there were attacks that could insert things into your datastream provided the hacker knew (could guess) what you where typing... But that has been fixed as far as I know).
Ah. Someone posted the abstract above. Good. They POSTULATE that no G-spot exists. This means they are guessing. I postulate that the G-spot does exist. Same evidence different guess at the end. So?
What I've gathered from the shabby reporting around this weird research is that their research model is completely flawed.
What they have proven is that the women's perception of the G-spot does NOT depend on genetics. Fine!
So: "all women have a G-spot" is just as likely as "no woman has a G-spot". And most likely: "the activation of the G-spot depends on how it's stimulated".
...' I got a good laugh off of this, but is it really possible to get better quality from a lower bitrate?" That's not a problem. Apparently you don't understand coding. But then don't pretend you do, and laugh without understanding the issues.
Suppose I have an interesting high-res picture, but half of it is sky. Now if I code this in two ways. First I do a high quality jpeg. Very little bits go to the blue background, and lots can be allocated for the interesting parts. Also I save the original as a BMP (uncompressed). The BMP is MUCH larger than the jpeg. Now to reduce the number of bits in the BMP I scale it down by 2x on both sides. So now I might have a 6 megapixel JPEG, and a scaled down 1.5Mpixel BMP. The BMP comes to 4.5 Mbyte. But the 6Mpixel JPEG will look better and it has more details. Still it consumes less bytes (say only 1 or two Mbyte).
This is just an example where the lower bitrate can outperform a higher bitrate. Now I'm pretty sure that the old higher-bitrate encoding wasn't as stupid as being uncompressed as in my example. Still, it can very well be that the old codec was sufficiently outdated that it can be outperformed by a more modern codec at a lower bitrate.
However if you switch codecs people will be used to the old codec and its artefacts. So they will notice the change. Then you'll get complaints valid or not.
Anyway, even IF a modern codec can outperform an older one at a lower bitrate, it remains to be seen if this modern codec at 9.7 Mbps can outperform the older one at 16Mbps.
As I had to do a fresh install anyway, and Karmic was out officially a few days, I decided to try it out.
After a few days it wanted to reboot because of a kernel upgrade and because I changed a display setting, I wanted to log out anyway. So I rebooted.
Well, I got a flickering text screen. It seems GDM was starting X at a rate of about 5 per second, and that it was failing.
I finally got X to work again by (re-)installing the proprietary driver from the display vendor. (ATI/AMD).
On the other hand, I did find out that I can still run my old window manager sawfish, which can do things (which I use!) that newer window managers like compiz and metacity cannot. This means that I'm a happy man.
If one third choses the supposedly inferior codec, then you could say that about 2 thirds simply doesn't know the difference and just choses randomly.
However, If I read things correctly, they tested one codec at 48k against another codec at 160k. This test shows that "the other codec at 160k" is pretty bad: It gets beaten (for a lot of people) by 48k on the other codec. Not that 1/3rd (or 2/3rds) of the people don't know what they are talking about.
Agreed lenticulars are called that way because they are shaped like a lens. These form just like real lenticulars, but are less obivously shaped like a lens.
So the Unix guys like to make jokes by naming programs "Yet Another Compiler Compiler" (I don't know what came before that, even though I'm pretty old by comp sci standards), Because that spells "yacc", the gnu equivalent is called "bison", another animal. Same here.
Two clicks away from the article, I found the name "mammatus lenticularis".
Lenticularis are lens-like clouds that usually hang just above the peak of a mountain. These are caused by a warmer layer of air on top being pushed above the condensation level by the wind having to go over a mountain.
These look like mamatus, but more creepy. Less regular.
So referring to mammatus refers to the way they look. Referring to lenticularis refers to the way they form: In exactly the same way as normal lenticularis does.
In the 1930's someone did careful research on how to study rats in mazes. Some things became clear: the rats could smell their prize from a long way, and would find it using their nose and not by being smart and having learned how to travel the maze. Similarly the rats have good hearing. They hear their feet tapping on the bottom of the maze, and these sound waves travel in a distinct way through the (usually wooden) bottom of the maze, bouncing off hidden prizes and thus revealing their location.
The researchers warned about the many different ways that research on rats running through mazes could go wrong. But they didn't have any actual rats-running-through-mazes results. So their research got mostly ignored. And others continued to be amazed at the smarts of the rats while the rats happily smelled their ways to the prizes.....
"But I didn't mean to take a picture of that" isn't much of a defence.
So if you're walking around paris, shoot the eiffel tower, and happen to shoot a weirdo with his child-porn-magazine in the background. You're not going to say... "but I didn't mean to take a picture of that!" ? Yeah right!
If I name the browser you used to post the above "malicious" and "illegal", you just intentionally committed a crime using malicious and illegal software by posting the above defamatory statements.
The BBC sent a laptop in for a simple repair, to find out what the repair techs would do to the machine. So they installed software that would take a screenshot every few seconds. It took screenshots of their own laptop. Is this illegal? I think not. So is the software illegal? When packaged with an auto-installer is it then illegal?
Do you have a knife in your kitchen? The use of that knife becomes illegal when you stick it into someone.
With the punishments as they are what are you stimulating? That jealous husbands stop using electronic surveillance to try to spy on their wives. Good.
But hospital employees should continue to jeopardize the privacy of the patients by clicking every executable in their inbox. If something bad happens it's always the other guy that did something illegal. Never your fault!
Do you get any spam? I bet you do. Getting spam is a "fact of life". Getting phishing attempts simply is a fact of life as well. It isn't going away because its illegal any time soon. Hospital employees, and more importantly hospitals themselves need to protect themselves against these outside influences.
Before you know it, it is illegal to enter a hospital with a contagious disease. And if a person ends up infecting the whole hospital, that person gets to die in jail, and the hospital gets to continue using unsafe practices that allow the wild dissemination of diseases.
Why don't they fine the guy $100 for trying to spy on his girlfriend, and why don't they fine the woman $50,000 in damages and fire ther for violating hospital security procedures (at least two of them: viewing private Email on work computers, clicking on executable attachments)?
Why don't they fine the hospital $1Million for not properly protecting the privacy of their patients?
Did the guy intend to spy on the medical procedures of those patients? No!
Suppose you're walking around as a tourist somewhere happily shooting pictures of the landmarks with your expensive new 24 megapixel camera with 400mm zoom lens. So you shoot a picture which say captures some trade secret. Now do you get thrown in jail for industrial espionage?
It is completely different if you specifically buy that camera and lens with the intent to take those industrially sensitive pictures, and especially position yourself in a way that you can photograph the competitions board room.
In my example, in the population 99.9% of the people don't have the disease. In those that tested positive, 90% still don't have the disease. That's with a 99% accurate test.
It's very difficult to effectively use a test that has an error rate of on the order of 17%. Or to be more accurate: A false negative rate of about 17%, and a false positive rate of "less than 83""....
But yes, you're right. It only becomes bad when people use the test wrong.
The research I don't remember the reference to however showed that in practise, submitting those 9990 healthy people from my example to further tests was not beneficial to the group as a whole.....
Part of the problem is unsolvable: Doctors aren't statistics-professors. So even though they are well educated they still fall for some of the statistical pitfalls that exist. And with these huge error rates the effects of those pitfalls are big, and the temptation to fall for them large.
Just recentely an article was published about the effects of a "non-perfect" screening test on a large group of people. (They did a case study for a specific test, which gives the same results as the theoretical example below)
If you have a test that is 99% accurate and 1 in 1000 people have the illness, for every million people, you'll find 990 out of a thousand who DO have the disease. And you send home 10 people who do have the disease. Too bad, 990 got the proper treatment. Right? No! Of the 999000 who don't have the disease, you'll find 9990 people with positive results. Are they getting treated? Are there risks to further tests? (for example X-rays create a risk for cancer later on.... )
And finding the 990 people WITH the disease from the 10980 people who tested positive is going to be a hassle. And costly.
Now a test with a 99% accuracy is pretty good compared to this test. As I read the slashdot intro, there is a 17% false negative rate, and up to 82% false positve rate....
which is complicated, and depends on the order you do things. Trying to do this in 2 minutes gives me the creeps that something is wrong, and I messed up the order. 20 years later I might write it (after getting coffee):
swap (object *p) {
object *pp, *q, *qn;
pp=p->prev;
q = p->next;
qn = q->next;// Now we have the items in order pp p q qn,// and we need to swap p and q.
pp->next = q;
q->next = p;
p->next = qn; }
Which also shows that trying to optimize out a few local variables leads to unreadable code.
Sure, I'm still curious why it didn't work the way it was supposed to.
I could have spent several hours to find a typo in my own code or maybe a bug in a library. To this day I continue to find typos in my code but less and less bugs-in-libraries.
Oh yeah. Similar here. In an assignment where we were supposed to learn the advantages of pointers, I had to swap two records in a list. So the idea is that you just swap the pointers, and keep the (large) records in the same spot in memory, to prevent having to move the (large) memory area holding all the data in the record.
When I wanted to finish the assignment it didn't work as expected. After some frustration I ended up just swapping the whole records and then swapping the pointers back. The assignment got OK-ed.
I had already learned to use pointers the proper way. I just had to pass the class goal achieved!
It is because about 50% of the population has a genetic advantage. So we have an "open" category, and a category for the 50% of the population that wants to compete among the others that don't posess this genetic advantage.
Now the question is: How do you prove you don't have this genetic advantage. Apparently there are some cases where you don't have the genetic advantage (commonly called "being male"), but still have some of the exterior symptoms of the genetic disadvantage.
Because the woman has been effectively blind since birth, the results suggest that the brain is able to adapt to new visual stimuli remarkably quickly. We knew that already. People have been equipped with a camera and an actuator that "projects" an image through the tactile nerves on a patch of skin.... Those people similarly report being able to "see" using that hardware.
This leads me to believe that our brains are not hardwired to interpret visual information only on just the optical nerve. During the first few years, we learn what nerves are connected where. And our brains are flexible, and will be able to adapt to changes in what-connects-where.
In an experiment, a long time ago(*), a person was equipped with glasses that flipped the image in his eyes upside-down. In about two weeks he didn't notice the difference anymore. Also, learned skills, like skiing, were instantly possible with the re-wired visual hardware. Adapting to the original situation (no flipping glasses) was quicker than adapting to the flipping glasses.
(*) http://wearcam.org/tetherless/node4.html I intended "a long time ago" to mean something like "in the 19-seventies", turns out the original experiment dates back to 1896!
Why do you want to force the page breaks? It's stupid. HTML is intended to render correctly independent of the resolution, so independent of number-of-characters-that-fit-onto-a-page. Suppose someone gets your academic paper, but he is a bit blind. So he sets character size to 15pt, and prints it (for reading during a hypothetical train ride).
Someone else is concerned with the environment, has good eyes, and prints double sided with a 9 pt font.
Generating documents that handle this well, means you have to take care that you refer to "fig 3" and not to "the figure on page 2", things like that.
Slashdot Mirror
You've got it.
- Losing your $HOME/.ssh or wherever it lives under Windows.
- reconfiguring the server.
- starting to use ssh for the first time to that particular server.
For example, although my home machine has been in existance for years, and one of my work systems has similarly been around for years, I had somehow never ssh-ed from this particular machine at home to that machine at work. The machines at home don't share homedirs. So I got the warning yesterday, which I was able to copy for the posting above....
If for example you have 10 workstations on one end and 10 servers on the other, you need to type "yes" a total of 100 times. If you've done that 50 times, you're open to a man in the middle attack because you're used to getting that warning every once in a while.
So yes, that's user error. However, those are worth optimizing for, because I think the crypto is safe...
Assuming the crypto is strong enough, and I think it is, the chances of getting hacked amount to "user error".
What kinds of user errors could happen?
The authenticity of host '192.168.1.1 (192.168.1.1)' can't be established.
RSA key fingerprint is 24:ba:36:a4:4b:11:59:e8:ec:dd:75:15:f2:2e:74:dc.
Are you sure you want to continue connecting (yes/no)?
If you occasionally get that, legitimately because you lose your key fingerprint hashes, you might fall for a man-in-the-middle attack. If you
leave your sessions running, there will be less chances of intercepting you while you do the "connect", and the chances of you getting used to these warnings and hitting enter too quickly will also reduce.
So I'd say that for normal use, leaving them running is more secure.
(on the other hand, there were attacks that could insert things into your datastream provided the hacker knew (could guess) what you where typing... But that has been fixed as far as I know).
I always put the screen towards my body, since I broke a screen the other way around.... :-)
Ah. Someone posted the abstract above. Good. They POSTULATE that no G-spot exists. This means they are guessing. I postulate that the G-spot does exist. Same evidence different guess at the end. So?
What I've gathered from the shabby reporting around this weird research is that their research model is completely flawed.
What they have proven is that the women's perception of the G-spot does NOT depend on genetics. Fine!
So: "all women have a G-spot" is just as likely as "no woman has a G-spot". And most likely: "the activation of the G-spot depends on how it's stimulated".
Hmm. Isn't the USA the country where that sort of stuff is highly illegal and can land friendly people in jail?
...' I got a good laugh off of this, but is it really possible to get better quality from a lower bitrate?"
That's not a problem. Apparently you don't understand coding. But then don't pretend you do, and laugh without understanding the issues.
Suppose I have an interesting high-res picture, but half of it is sky. Now if I code this in two ways. First I do a high quality jpeg. Very little bits go to the blue background, and lots can be allocated for the interesting parts. Also I save the original as a BMP (uncompressed). The BMP is MUCH larger than the jpeg. Now to reduce the number of bits in the BMP I scale it down by 2x on both sides. So now I might have a 6 megapixel JPEG, and a scaled down 1.5Mpixel BMP. The BMP comes to 4.5 Mbyte. But the 6Mpixel JPEG will look better and it has more details. Still it consumes less bytes (say only 1 or two Mbyte).
This is just an example where the lower bitrate can outperform a higher bitrate. Now I'm pretty sure that the old higher-bitrate encoding wasn't as stupid as being uncompressed as in my example. Still, it can very well be that the old codec was sufficiently outdated that it can be outperformed by a more modern codec at a lower bitrate.
However if you switch codecs people will be used to the old codec and its artefacts. So they will notice the change. Then you'll get complaints valid or not.
Anyway, even IF a modern codec can outperform an older one at a lower bitrate, it remains to be seen if this modern codec at 9.7 Mbps can outperform the older one at 16Mbps.
As I had to do a fresh install anyway, and Karmic was out officially a few days, I decided to try it out.
After a few days it wanted to reboot because of a kernel upgrade and because I changed a display setting, I wanted to log out anyway. So I rebooted.
Well, I got a flickering text screen. It seems GDM was starting X at a rate of about 5 per second, and that it was failing.
I finally got X to work again by (re-)installing the proprietary driver from the display vendor. (ATI/AMD).
On the other hand, I did find out that I can still run my old window manager sawfish, which can do things (which I use!) that newer window managers like compiz and metacity cannot. This means that I'm a happy man.
FYI, in my experience drives that fail rarely have SMART data indicating a problem.
The big google harddrive research paper had similar results.
On the other hand, if you get a hint of trouble from SMART, it's a good idea to retire the drive. It DOES sometimes work . :-)
If one third choses the supposedly inferior codec, then you could say that about 2 thirds simply doesn't know the difference and just choses randomly.
However, If I read things correctly, they tested one codec at 48k against another codec at 160k. This test shows that "the other codec at 160k" is pretty bad: It gets beaten (for a lot of people) by 48k on the other codec. Not that 1/3rd (or 2/3rds) of the people don't know what they are talking about.
Agreed lenticulars are called that way because they are shaped like a lens. These form just like real lenticulars, but are less obivously shaped like a lens.
So the Unix guys like to make jokes by naming programs "Yet Another Compiler Compiler" (I don't know what came before that, even though I'm pretty old by comp sci standards), Because that spells "yacc", the gnu equivalent is called "bison", another animal. Same here.
Two clicks away from the article, I found the name "mammatus lenticularis".
Lenticularis are lens-like clouds that usually hang just above the peak of a mountain. These are caused by a warmer layer of air on top being pushed above the condensation level by the wind having to go over a mountain.
These look like mamatus, but more creepy. Less regular.
So referring to mammatus refers to the way they look. Referring to lenticularis refers to the way they form: In exactly the same way as normal lenticularis does.
In the 1930's someone did careful research on how to study rats in mazes. Some things became clear: the rats could smell their prize from a long way, and would find it using their nose and not by being smart and having learned how to travel the maze. Similarly the rats have good hearing. They hear their feet tapping on the bottom of the maze, and these sound waves travel in a distinct way through the (usually wooden) bottom of the maze, bouncing off hidden prizes and thus revealing their location.
The researchers warned about the many different ways that research on rats running through mazes could go wrong. But they didn't have any actual rats-running-through-mazes results. So their research got mostly ignored. And others continued to be amazed at the smarts of the rats while the rats happily smelled their ways to the prizes.....
"But I didn't mean to take a picture of that" isn't much of a defence.
So if you're walking around paris, shoot the eiffel tower, and happen to shoot a weirdo with his child-porn-magazine in the background. You're not going to say... "but I didn't mean to take a picture of that!" ? Yeah right!
If I name the browser you used to post the above "malicious" and "illegal", you just intentionally committed a crime using malicious and illegal software by posting the above defamatory statements.
The BBC sent a laptop in for a simple repair, to find out what the repair techs would do to the machine. So they installed software that would take a screenshot every few seconds. It took screenshots of their own laptop. Is this illegal? I think not. So is the software illegal? When packaged with an auto-installer is it then illegal?
Do you have a knife in your kitchen? The use of that knife becomes illegal when you stick it into someone.
With the punishments as they are what are you stimulating? That jealous husbands stop using electronic surveillance to try to spy on their wives. Good.
But hospital employees should continue to jeopardize the privacy of the patients by clicking every executable in their inbox. If something bad happens it's always the other guy that did something illegal. Never your fault!
Do you get any spam? I bet you do. Getting spam is a "fact of life". Getting phishing attempts simply is a fact of life as well. It isn't going away because its illegal any time soon. Hospital employees, and more importantly hospitals themselves need to protect themselves against these outside influences.
Before you know it, it is illegal to enter a hospital with a contagious disease. And if a person ends up infecting the whole hospital, that person gets to die in jail, and the hospital gets to continue using unsafe practices that allow the wild dissemination of diseases.
Why don't they fine the guy $100 for trying to spy on his girlfriend, and why don't they fine the woman $50,000 in damages and fire ther for violating hospital security procedures (at least two of them: viewing private Email on work computers, clicking on executable attachments)?
Why don't they fine the hospital $1Million for not properly protecting the privacy of their patients?
Did the guy intend to spy on the medical procedures of those patients? No!
Suppose you're walking around as a tourist somewhere happily shooting pictures of the landmarks with your expensive new 24 megapixel camera with 400mm zoom lens. So you shoot a picture which say captures some trade secret. Now do you get thrown in jail for industrial espionage?
It is completely different if you specifically buy that camera and lens with the intent to take those industrially sensitive pictures, and especially position yourself in a way that you can photograph the competitions board room.
I've always wondered why pigeons fly so loudly. It costs energy to make sound, so it can't be efficient.
It all makes sense now.
In my example, in the population 99.9% of the people don't have the disease. In those that tested positive, 90% still don't have the disease. That's with a 99% accurate test.
It's very difficult to effectively use a test that has an error rate of on the order of 17%. Or to be more accurate: A false negative rate of about 17%, and a false positive rate of "less than 83""....
But yes, you're right. It only becomes bad when people use the test wrong.
The research I don't remember the reference to however showed that in practise, submitting those 9990 healthy people from my example to further tests was not beneficial to the group as a whole.....
Part of the problem is unsolvable: Doctors aren't statistics-professors. So even though they are well educated they still fall for some of the statistical pitfalls that exist. And with these huge error rates the effects of those pitfalls are big, and the temptation to fall for them large.
Just recentely an article was published about the effects of a "non-perfect" screening test on a large group of people. (They did a case study for a specific test, which gives the same results as the theoretical example below)
If you have a test that is 99% accurate and 1 in 1000 people have the illness, for every million people, you'll find 990 out of a thousand who DO have the disease. And you send home 10 people who do have the disease. Too bad, 990 got the proper treatment. Right? No! Of the 999000 who don't have the disease, you'll find 9990 people with positive results. Are they getting treated? Are there risks to further tests? (for example X-rays create a risk for cancer later on.... )
And finding the 990 people WITH the disease from the 10980 people who tested positive is going to be a hassle. And costly.
Now a test with a 99% accuracy is pretty good compared to this test. As I read the slashdot intro, there is a 17% false negative rate, and up to 82% false positve rate....
I was trying to write something like:
swap (object *p)
{
object *tmp;
tmp = p->next;
tmp->next = p;
p->next = p->next->next;
p->prev->next = tmp;
}
which is complicated, and depends on the order you do things. Trying to do this in 2 minutes gives me the creeps that something is wrong, and I messed up the order. 20 years later I might write it (after getting coffee):
swap (object *p) // Now we have the items in order pp p q qn, // and we need to swap p and q.
{
object *pp, *q, *qn;
pp=p->prev;
q = p->next;
qn = q->next;
pp->next = q;
q->next = p;
p->next = qn;
}
Which also shows that trying to optimize out a few local variables leads to unreadable code.
I ended up writing:
#define SWAP(a,b,tmp) tmp=a;a=b;b=tmp
swaprecords (object *p)
{
object tobj, *tp, *q;
q = p->next;
SWAP (*p, *q tobj);
SWAP (p-next, q->next, tp);
}
Sure, I'm still curious why it didn't work the way it was supposed to.
I could have spent several hours to find a typo in my own code or maybe a bug in a library. To this day I continue to find typos in my code but less and less bugs-in-libraries.
If you say you have problems with Windows I believe you. No questions asked.
Oh yeah. Similar here. In an assignment where we were supposed to learn the advantages of pointers, I had to swap two records in a list. So the idea is that you just swap the pointers, and keep the (large) records in the same spot in memory, to prevent having to move the (large) memory area holding all the data in the record.
When I wanted to finish the assignment it didn't work as expected. After some frustration I ended up just swapping the whole records and then swapping the pointers back. The assignment got OK-ed.
I had already learned to use pointers the proper way. I just had to pass the class goal achieved!
Why do we have a "female" category in sports?
It is because about 50% of the population has a genetic advantage. So we have an "open" category, and a category for the 50% of the population that wants to compete among the others that don't posess this genetic advantage.
Now the question is: How do you prove you don't have this genetic advantage. Apparently there are some cases where you don't have the genetic advantage (commonly called "being male"), but still have some of the exterior symptoms of the genetic disadvantage.
This has happened before. ... in Dutch (better article)
Ewa Klobukowska
Foekje Dillema
Stella Walsh
So apparently in '67 they had the "Barr body" test to see if someone is female.
Because the woman has been effectively blind since birth, the results suggest that the brain is able to adapt to new visual stimuli remarkably quickly.
We knew that already. People have been equipped with a camera and an actuator that "projects" an image through the tactile nerves on a patch of skin.... Those people similarly report being able to "see" using that hardware.
This leads me to believe that our brains are not hardwired to interpret visual information only on just the optical nerve. During the first few years, we learn what nerves are connected where. And our brains are flexible, and will be able to adapt to changes in what-connects-where.
In an experiment, a long time ago(*), a person was equipped with glasses that flipped the image in his eyes upside-down. In about two weeks he didn't notice the difference anymore. Also, learned skills, like skiing, were instantly possible with the re-wired visual hardware. Adapting to the original situation (no flipping glasses) was quicker than adapting to the flipping glasses.
(*) http://wearcam.org/tetherless/node4.html I intended "a long time ago" to mean something like "in the 19-seventies", turns out the original experiment dates back to 1896!
Why do you want to force the page breaks? It's stupid. HTML is intended to render correctly independent of the resolution, so independent of number-of-characters-that-fit-onto-a-page. Suppose someone gets your academic paper, but he is a bit blind. So he sets character size to 15pt, and prints it (for reading during a hypothetical train ride).
Someone else is concerned with the environment, has good eyes, and prints double sided with a 9 pt font.
Generating documents that handle this well, means you have to take care that you refer to "fig 3" and not to "the figure on page 2", things like that.