NAT makes a fundamental change to the internet as a whole. End-to=end connectivity is no longer guaranteed. The fixes to existing protocols to get around this are very ugly. FTP is an amazing example of this. To support FTP in active mode, NAT boxes have to edit the PORT command the clients send. The problem being that this changes the length of the TCP packet it's in, and necessitates munging all the sequence numbers from there on out. Very ugly. FTP in passive mode breaks load balancing.
Not that I think FTP is a great thing to keep around, but it's an example of what goes wrong with NAT. Why do you think there's a theoretical problem with including network information in a protocol at a higher layer? A stateful firewall may have to be aware of that sort of information (to pick up on related connections that should be allowed), but it doesn't have to actually change the data. It would be much cleaner to have IPv6 and stateful firewalling.
NAT also creates endless headaches when it becomes common enough that you're trying to connect two computers that are both behind seperate NAT boxes.
Quick note. Hams are subject to a number of rules about content. Encryption is not allowed. Neither is profanity, or any content that could count as business related. A ham using an unmodified 802.11 device uses it under part 15 of the FCC rules and can send whatever they want over it. If you get a ham license and boost the power of your 802.11 device, everything you send over that wireless link is subject to the additional rules that govern the ham bands. In other words, using your ham license to legally boost the power of your 802.11 device isn't as helpful as it sounds. You get more power, but have to follow more rules. It might be interesting for certain applications, but you almost certainly couldn't send general internet traffic over it. You couldn't legally use https or view web pages containing profanity.
Either QST or CQ ran an interesting article about this a bit back. The author concluded that mostly it isn't worth using your ham license for 802.11.
RDF is designed to let computers exchange data over the web with each other. Although this is using a liberal meaning of RDF, which is mostly just a format for writing triples of data. Really RDFS and OWL are designed to help create global vocabularies for computers to use in exchanging data.
The idea is basically that my web page contains a lot of information about me, but most of it is only interpretable by a human. If I have a format for saying "Amy a person", "Amy reads slashdot", and a way that we can describe what I mean by "person" here, and how it relates to what someone else considers a "person" the information becomes something a computer can work with without needing to do natural language processing. (My foaf file is an example of this)
A classic example is being able to search intelligently, understanding things like the difference between "OWL" the language, and "owl", the bird.
Thw w3c produces recommendations. OWL and RDF are recommendations. Yes, eventually users and developers will decide whether they like them as is, want to ignore them, or want to change them. But there are many tasks where people have to discuss the options and suggest one. There are a near infinite number of formats one could use for a web ontology language. Unless someone publically suggests "why don't we agree on this one?" it takes a very long time for a standard to emerge on this kind of thing.
I assume you are asking why rdf uses full uri's to denote items instead of just local ones. (As the reply above me points out rdf doesn't really use namespaces, it's just that writing full uri's is a nuisance, and namespaces are a way of abreviating them.) The idea is to let documents talk about items declared in other files. If in my foaf file I declare "http://sarn.org/foaf.rdf#Amy" is a person, someone else can assert that that same Amy is also a college student. Without these global names you would never get any links between documents.
Sorry if this is somewhat less than coherent, but essentially using full uri's gives you something which should be globally unique to refer to from anywhere.
Personally I find it hard to work with RDF/XML since it can get kind of unreadable. I've found this primer on n3 helpful. n3 is a simpler way to write RDF, which makes the triple structure a little more obvious.
A fun place to start in RDF is making a foaf page. Foaf is the friend of a friend vocabulary. If you search for foaf in google you should find stuff to help you start with it. This lets you track things like degrees of seperation between people.
You can write OWL markup that describes the content of your webpage, but this is somewhat harder to do (there are some graphical tools that would help), and less useful right now. There aren't many tools that make use of/display random OWL markup associated with a web page.
More useful for a small webpage might be including dublin core metadata (should have no problem searching for their homepage either) about the author, title, etc. of each page. The dublin core initiative provides info about how to do this.
It's hard to find this out, but the standard direcway service nat's you. You do not have a publically routable ip address, let alone a static one. Upgrading to a commercial package can get you a static ip. The nat boxes also tend to kill long running connections (i.e. if you leave ssh running logged in somewhere).
The standard satellite modem (for lack of better name) doesn't work well in linux. Hughes has a patent on the LZJH compression algorithm. DirecWay forces you to use IP compression for port 80 connections with the LZJH compression algorithm. So linux drivers have trouble with web traffic.
The latency is awful. During peak hours (afternoon on) ping times get up to 2 seconds. I've never seen them below 700 ms. Some web pages that open a lot of connections to download small items feel slower over the satellite link than over dial up.
Finally Hughes has a fair access policy the details of which they won't share. As far as anyone can tell they're using token bucket qos with a bucket of about 150 megs and and a fill rate of 56kbps. What this means is that if you doenloaded nonstop all day you'd get 56kbps. You're just allowed to save up some of your bandwidth and use it all at once, so the connection feels faster.
Weather has lousy effects on connection quality. Heavy cloud cover can mess it up occaisionally. Mostly though the problem is rain storms. A good thunder storm can knock out the connection completely until it passes. The DirecWay service is much more suceptible to weather related problems than satellite tv is.
Essentially it's ok for large downloads (although be sure to find someone's script to tune the linux ip stack settings, or large downloads will stall partway through.) Most stuff is painful over it. We keep a dial up account with a local isp for ssh , and times the link isn't working. I'm a very unhappy customer. I didn't believe a lot of the complaints I read about DirecWay because the complaints were so negative they didn't sound believable. 4 hours after we had the installation completed I discovered that by and large the complaints are all true.
Mozilla provides a site navigation toolbar. This includes "next", "prev", "up", "first", "last" and a whole bunch of other things (again, based on the contents of link tags in the page header). It also has the advantage of not overriding an existing button, which can be confusing. It's a reasonably thin toolbar, and can be set to only appear when there are link tags in the page header for it to display.
Actually the link tag is still supported in mozilla. They just took it out of netscape 6 because of problems, and have that toolbar off by default. View -> Show/Hide -> Site Navigation Bar -> Show Always will turn it on for you. Very handy little feature. I just wish more sites supported it.
You shouldn't need to serve different pages to do this. CSS2 has a media type for stylesheets. This means you can write a stylesheet for how your page displays on screen, another for how it should print, another for how text only browsers should display it, another for for how braille terminals should interpret it, another for how a screen reader should read it, and another for how a pda should display it. Meaning the only reason for serving different html to different browsers, or different css based on user agent strings is to get around broken browsers.
The problem is that if you work around broken browsers you encourage their authors not to fix them, and they're users to keep using old versions/not demand fixes.
How do you look for it? I've been working on a tool called RET (http://sarn.org/ret/) that pulls rdf from web pages. Unfortunately I've found 11 different methods for embedding it. If you look for dublin core metadata meta tags, creative commons comments with RDF in them, and link tags pointing to dublin core metadata I think you'll find a fair amount. I'm not really sure, I haven't looked for sites using this stuff yet.
One problem with this is if the information needed to recognize an answer for a query lives on multiple sites. Like I want to search for someone with a first name of Jane who works at Acme Inc and Foo Deluxe. I could go through all the results of searching for Janes at Acme Inc and coorelate the resulting names with the list of Janes working at Foo Deluxe, but it would be a pain.
I'm sure there's some way to do this without semantic web stuff, but writing a more effective query isn't likely to solve it.
Discovery of web services for some uses is further along than you think. The point isn't really to let servers do it. It's more things like you walk into a room to give a presentation, and you'd like your laptop to figure out what projectors are available, how it can control them, and how to dim the lights. The lab I work for is playing with some of this stuff right now.
As for authentication that's what signatures are for. For things that need authentication that's perfectly possible. Plenty of things really don't - search engines for example. Yes, it will be possible to screw up search engines, just like it's always been possible to screw up search engines. But everyone knows that the results aren't perfect, and it isn't a huge problem.
These aren't languages designed to display to users. Browsers might do aditional things using them, but I doubt that html is going away anytime soon. One idea is to create more effective search engines - such as one that can search for the web page of markup language named "shoe" and not return a bunch of results about sneakers. Usually the new markup languages are either embedded in html (which is wrong and bad!) or are linked to from web pages using something like the "link" tag in the page head that points to alternate pages.
The new languages are laregly for use by automated agents, not humans.
Basically this isn't overhauling the web as we know it so much as adding a new web in in addition.
I think the problem with having a list of TLD's distributed by the root servers is that I can techincally have a TLD that only my nameserver knows about. I.e. in my house we have a seperate internal dns which believes there is a.int TLD which it is authoritative for. While it would be easy to allow requests only for things on the TLD list and TLD's that the nameserver thinks it's authoratative for it may be a problem if a company has a main nameserver that hosts an internal TLD and machines in the company are running cacheing copies of bind. There may be easy ways to deal with this, I have no idea.
I don't think it's really fair to compare things based solely on percentage of net worth. Look at it this way: if I gave 50% of my net worth it would be meaningful because it would affect my standard of living. But giving 50% of 100 billion dollars is not nearly so meaningful because one person cannot possibly use 50 billion dollars in a lifetime.
It's nice that he's giving away the money, but I don't consider it any kind of sacrifice on his part.
Re:Getting an OpenBSD box to emulate a .Mac serv
on
Fake Your Own .Mac Server
·
· Score: 5, Insightful
This isn't authenticating you to apple's backup server which unlimited trial days, it's authenticating you to your own backup server with unlimited trial days. Hence the rest of the article explaining about setting up DAV and whatnot to do the actual backups. Nothing illegal about that.
Their entire staff is lawyers, and they know nothing about computers. Their job is to keep the campus from getting sued if someone complains about a students computer usage. They frequently get confused over who the victims are and who the bad guys are. My dorm room machine got broken into and they called up threatening me. A friend got dos'd by a poorly configured network. That network's admin called nethics who went after my friend. They are totally unaware of the concepts of spoofing and sniffing.
In other words nethics is definately not a good example.
I've been using vim (the console version) for years, and I'm interested in this because it's a kpart. I don't care about the menus, I'll just ignore them. But being able to embed a vim window in konqueror when I click on a file, rather than opening a seperate konsole with vim, or using kde's native editor, would be great. I've been hoping for a way to do this for several months.
I think the idea is that if the ice breaks up beneath them they can climb back onto solid ice. Otherwise this would all be kind of silly. We've had boats that can navigate icey water for years.
Hmm, how about an adult version for peadophiles on parole? Program in all the local school yards, etc, sample every few minutes and trigger an alarm if he gets too close.
This is actually already being done. I read somewhere about GPS tracking being used to monitor convicted pedophiles. They claimed that they had enough accuracy to catch a guy who slowed down driving past the house of his granddaughter (who he'd previously sexually abused, and was supposed to have no contact with).
Re:Proves strength of Google
on
Google Juice
·
· Score: 1
< shameless plug> There's an idea called the semantic web, that in theory will fix this. The idea is to make pages machine readable so that John Lewis the department store doesn't look the same as John Lewis the person. In theory using the language daml you could search for all John Lewises that are people, and live in whatever place. Or any other useful identifying info like that. </shameless plug>
Note on ADHD. It's not about only paying attention to what you choose to. It's well described as the lack of ability to choose what to pay attention to. In other words kids with ADHD can concentrate. They just can't choose to concentrate. If they concentrate it happens on it's own.
My favorite solution to this is what adobe framemaker and quicken both do. That's provide an interchange format that is well documented and shouldn't change. Then they can do whatever they like with their native format. Transfering files to a different program, or between different versions is done through the interchange format.
Slashdot Mirror
NAT makes a fundamental change to the internet as a whole. End-to=end connectivity is no longer guaranteed. The fixes to existing protocols to get around this are very ugly. FTP is an amazing example of this. To support FTP in active mode, NAT boxes have to edit the PORT command the clients send. The problem being that this changes the length of the TCP packet it's in, and necessitates munging all the sequence numbers from there on out. Very ugly. FTP in passive mode breaks load balancing.
Not that I think FTP is a great thing to keep around, but it's an example of what goes wrong with NAT. Why do you think there's a theoretical problem with including network information in a protocol at a higher layer? A stateful firewall may have to be aware of that sort of information (to pick up on related connections that should be allowed), but it doesn't have to actually change the data. It would be much cleaner to have IPv6 and stateful firewalling.
NAT also creates endless headaches when it becomes common enough that you're trying to connect two computers that are both behind seperate NAT boxes.
Quick note. Hams are subject to a number of rules about content. Encryption is not allowed. Neither is profanity, or any content that could count as business related.
A ham using an unmodified 802.11 device uses it under part 15 of the FCC rules and can send whatever they want over it. If you get a ham license and boost the power of your 802.11 device, everything you send over that wireless link is subject to the additional rules that govern the ham bands.
In other words, using your ham license to legally boost the power of your 802.11 device isn't as helpful as it sounds. You get more power, but have to follow more rules. It might be interesting for certain applications, but you almost certainly couldn't send general internet traffic over it. You couldn't legally use https or view web pages containing profanity.
Either QST or CQ ran an interesting article about this a bit back. The author concluded that mostly it isn't worth using your ham license for 802.11.
RDF is designed to let computers exchange data over the web with each other. Although this is using a liberal meaning of RDF, which is mostly just a format for writing triples of data. Really RDFS and OWL are designed to help create global vocabularies for computers to use in exchanging data.
The idea is basically that my web page contains a lot of information about me, but most of it is only interpretable by a human. If I have a format for saying "Amy a person", "Amy reads slashdot", and a way that we can describe what I mean by "person" here, and how it relates to what someone else considers a "person" the information becomes something a computer can work with without needing to do natural language processing. (My foaf file is an example of this)
A classic example is being able to search intelligently, understanding things like the difference between "OWL" the language, and "owl", the bird.
Thw w3c produces recommendations. OWL and RDF are recommendations. Yes, eventually users and developers will decide whether they like them as is, want to ignore them, or want to change them. But there are many tasks where people have to discuss the options and suggest one. There are a near infinite number of formats one could use for a web ontology language. Unless someone publically suggests "why don't we agree on this one?" it takes a very long time for a standard to emerge on this kind of thing.
I assume you are asking why rdf uses full uri's to denote items instead of just local ones. (As the reply above me points out rdf doesn't really use namespaces, it's just that writing full uri's is a nuisance, and namespaces are a way of abreviating them.) The idea is to let documents talk about items declared in other files. If in my foaf file I declare "http://sarn.org/foaf.rdf#Amy" is a person, someone else can assert that that same Amy is also a college student. Without these global names you would never get any links between documents.
Sorry if this is somewhat less than coherent, but essentially using full uri's gives you something which should be globally unique to refer to from anywhere.
Personally I find it hard to work with RDF/XML since it can get kind of unreadable. I've found this primer on n3 helpful. n3 is a simpler way to write RDF, which makes the triple structure a little more obvious.
A fun place to start in RDF is making a foaf page. Foaf is the friend of a friend vocabulary. If you search for foaf in google you should find stuff to help you start with it. This lets you track things like degrees of seperation between people.
You can write OWL markup that describes the content of your webpage, but this is somewhat harder to do (there are some graphical tools that would help), and less useful right now. There aren't many tools that make use of/display random OWL markup associated with a web page.
More useful for a small webpage might be including dublin core metadata (should have no problem searching for their homepage either) about the author, title, etc. of each page. The dublin core initiative provides info about how to do this.
It's hard to find this out, but the standard direcway service nat's you. You do not have a publically routable ip address, let alone a static one. Upgrading to a commercial package can get you a static ip. The nat boxes also tend to kill long running connections (i.e. if you leave ssh running logged in somewhere).
The standard satellite modem (for lack of better name) doesn't work well in linux. Hughes has a patent on the LZJH compression algorithm. DirecWay forces you to use IP compression for port 80 connections with the LZJH compression algorithm. So linux drivers have trouble with web traffic.
The latency is awful. During peak hours (afternoon on) ping times get up to 2 seconds. I've never seen them below 700 ms. Some web pages that open a lot of connections to download small items feel slower over the satellite link than over dial up.
Finally Hughes has a fair access policy the details of which they won't share. As far as anyone can tell they're using token bucket qos with a bucket of about 150 megs and and a fill rate of 56kbps. What this means is that if you doenloaded nonstop all day you'd get 56kbps. You're just allowed to save up some of your bandwidth and use it all at once, so the connection feels faster.
Weather has lousy effects on connection quality. Heavy cloud cover can mess it up occaisionally. Mostly though the problem is rain storms. A good thunder storm can knock out the connection completely until it passes. The DirecWay service is much more suceptible to weather related problems than satellite tv is.
Essentially it's ok for large downloads (although be sure to find someone's script to tune the linux ip stack settings, or large downloads will stall partway through.) Most stuff is painful over it. We keep a dial up account with a local isp for ssh , and times the link isn't working. I'm a very unhappy customer. I didn't believe a lot of the complaints I read about DirecWay because the complaints were so negative they didn't sound believable. 4 hours after we had the installation completed I discovered that by and large the complaints are all true.
Mozilla provides a site navigation toolbar. This includes "next", "prev", "up", "first", "last" and a whole bunch of other things (again, based on the contents of link tags in the page header). It also has the advantage of not overriding an existing button, which can be confusing. It's a reasonably thin toolbar, and can be set to only appear when there are link tags in the page header for it to display.
Actually the link tag is still supported in mozilla. They just took it out of netscape 6 because of problems, and have that toolbar off by default. View -> Show/Hide -> Site Navigation Bar -> Show Always will turn it on for you. Very handy little feature. I just wish more sites supported it.
You shouldn't need to serve different pages to do this. CSS2 has a media type for stylesheets. This means you can write a stylesheet for how your page displays on screen, another for how it should print, another for how text only browsers should display it, another for for how braille terminals should interpret it, another for how a screen reader should read it, and another for how a pda should display it. Meaning the only reason for serving different html to different browsers, or different css based on user agent strings is to get around broken browsers.
The problem is that if you work around broken browsers you encourage their authors not to fix them, and they're users to keep using old versions/not demand fixes.
How do you look for it? I've been working on a tool called RET (http://sarn.org/ret/) that pulls rdf from web pages. Unfortunately I've found 11 different methods for embedding it. If you look for dublin core metadata meta tags, creative commons comments with RDF in them, and link tags pointing to dublin core metadata I think you'll find a fair amount. I'm not really sure, I haven't looked for sites using this stuff yet.
One problem with this is if the information needed to recognize an answer for a query lives on multiple sites. Like I want to search for someone with a first name of Jane who works at Acme Inc and Foo Deluxe. I could go through all the results of searching for Janes at Acme Inc and coorelate the resulting names with the list of Janes working at Foo Deluxe, but it would be a pain.
I'm sure there's some way to do this without semantic web stuff, but writing a more effective query isn't likely to solve it.
As for authentication that's what signatures are for. For things that need authentication that's perfectly possible. Plenty of things really don't - search engines for example. Yes, it will be possible to screw up search engines, just like it's always been possible to screw up search engines. But everyone knows that the results aren't perfect, and it isn't a huge problem.
These aren't languages designed to display to users. Browsers might do aditional things using them, but I doubt that html is going away anytime soon. One idea is to create more effective search engines - such as one that can search for the web page of markup language named "shoe" and not return a bunch of results about sneakers. Usually the new markup languages are either embedded in html (which is wrong and bad!) or are linked to from web pages using something like the "link" tag in the page head that points to alternate pages.
The new languages are laregly for use by automated agents, not humans.
Basically this isn't overhauling the web as we know it so much as adding a new web in in addition.
I think the problem with having a list of TLD's distributed by the root servers is that I can techincally have a TLD that only my nameserver knows about. I.e. in my house we have a seperate internal dns which believes there is a .int TLD which it is authoritative for. While it would be easy to allow requests only for things on the TLD list and TLD's that the nameserver thinks it's authoratative for it may be a problem if a company has a main nameserver that hosts an internal TLD and machines in the company are running cacheing copies of bind. There may be easy ways to deal with this, I have no idea.
I don't think it's really fair to compare things based solely on percentage of net worth. Look at it this way: if I gave 50% of my net worth it would be meaningful because it would affect my standard of living. But giving 50% of 100 billion dollars is not nearly so meaningful because one person cannot possibly use 50 billion dollars in a lifetime.
It's nice that he's giving away the money, but I don't consider it any kind of sacrifice on his part.
This isn't authenticating you to apple's backup server which unlimited trial days, it's authenticating you to your own backup server with unlimited trial days. Hence the rest of the article explaining about setting up DAV and whatnot to do the actual backups. Nothing illegal about that.
Their entire staff is lawyers, and they know nothing about computers. Their job is to keep the campus from getting sued if someone complains about a students computer usage. They frequently get confused over who the victims are and who the bad guys are. My dorm room machine got broken into and they called up threatening me. A friend got dos'd by a poorly configured network. That network's admin called nethics who went after my friend. They are totally unaware of the concepts of spoofing and sniffing.
In other words nethics is definately not a good example.
I've been using vim (the console version) for years, and I'm interested in this because it's a kpart. I don't care about the menus, I'll just ignore them. But being able to embed a vim window in konqueror when I click on a file, rather than opening a seperate konsole with vim, or using kde's native editor, would be great. I've been hoping for a way to do this for several months.
I think the idea is that if the ice breaks up beneath them they can climb back onto solid ice. Otherwise this would all be kind of silly. We've had boats that can navigate icey water for years.
Hmm, how about an adult version for peadophiles on parole? Program in all the local school yards, etc, sample every few minutes and trigger an alarm if he gets too close.
This is actually already being done. I read somewhere about GPS tracking being used to monitor convicted pedophiles. They claimed that they had enough accuracy to catch a guy who slowed down driving past the house of his granddaughter (who he'd previously sexually abused, and was supposed to have no contact with).
< shameless plug>
There's an idea called the semantic web, that in theory will fix this. The idea is to make pages machine readable so that John Lewis the department store doesn't look the same as John Lewis the person. In theory using the language daml you could search for all John Lewises that are people, and live in whatever place. Or any other useful identifying info like that.
</shameless plug>
Note on ADHD. It's not about only paying attention to what you choose to. It's well described as the lack of ability to choose what to pay attention to. In other words kids with ADHD can concentrate. They just can't choose to concentrate. If they concentrate it happens on it's own.
My favorite solution to this is what adobe framemaker and quicken both do. That's provide an interchange format that is well documented and shouldn't change. Then they can do whatever they like with their native format. Transfering files to a different program, or between different versions is done through the interchange format.
gzip doesn't claim to be able to compress any random data. The story on this compression method appeared to claim it could.