Do you realize just how few of our laws
are obeyed anyway?
So you are saying that if there were actual US privacy laws, big companies like eBay would just ignore them? And leave themselves vulnerable to class action lawsuits?
Point is exactly that biggest companies are in best position to collect personal information, and then sell it. But they are also the easiest ones to control through legislation. So, gee, in this particular case laws might actually be reasonably effective.
There's no good excuse for child porn, and while this law (and therefore method) may not be
the way to go about it, it's a start.
And a lousy start at that. Like n+1 posters have pointed out, the problem is not about shutting down illegal sites, it's about forcing others (in this case, ISPs), to do the work of law enforcement and court system. And neither of those should be duties of entities outside local and federal police and US court system.
To put simply; hand the list of confirmed kiddie porn sites to ISP, demand them to cut the connection, show the court order, and that's it. Do not try to weasel away from your responsibility, mr. Judge.
Well, I'm not really whining too badly, but I seem to remember that some other search engines (of which AV was one) did get to page earlier, after just few days. Perhaps it was just a co-incidence that in my case Google was slower than Alta Vista and whatever the other one was I checked (excite, northern lights, whatever). But like I said even if I was right and Google was slower, it wouldn't matter a lot. And if it's not slower, well, all the better!
As to publicity, I've never really tried to "optimize" ranking of my home pages, so I haven't even read their policies. It's good they give information about the frequency, though.
Well, unfortunately coverage size doesn't make up for poor ranking. Google is bit slow in getting to new pages (took almost a month to index my home pages when I moved to a new ISP), but its ranking accuracy is top-notch. So, as long as coverage is not an order-of-magniture worse with Google than with alternatives, I don't see that as being the determining factor.
Of course, I'm Yet Another of those "used Altavista for years, then switched to Google never looked back" users... so what do I know about AV's current usefulness.:-)
True. Of course, same could be said about most professions in existence, from accountants to CEOs, from janitors to stock brokers... so I don't think programmers are any more or less glamorous than 98% of other work force. That is, "real" work of a professional is, on average, fairly boring especially to untrained eye.
Plus, even some of "sexier" careers (according to many people), are only interesting to people who share the same interests or ideals. To me, work of a skateboarding hero or world's best surfer, doesn't really sound all that interesting. But for others it would be. And I'd venture a guess similar thing applies to many programmers (ie. they might find, say, Linus Torvald's work fairly interesting).
This all reminds me of my ex-boss explaining what war is (or at least used to be, around WW II): "it's days, weeks, months, of doing pretty much nothing, farting around, then there's some hectic action and people die, and then there's lots more of the boring stuff". I guess what I'm trying to say is just that movies are only interesting by making non-realistic but believable entertainment out of otherwise (mostly) dull world.
Whoa, I'm happy you said that, because I would have had to argue same thing. There are way too many "XMLheads" in existence who think that just by starting to use a structured markup system they will automatically be adding information that solves all searching problems.
In fact, if XML (and similar) markup mechanism were being designed to help searching, they certainly could help. But as things are that's not the order -- usually "content specialist" concentrate on classifying things without considering how those classifications are to be used; then try to force content authors to use markup, and only then start asking developers to develop system that make use of their fabulous new markup system.
And then the markup evangelists (like mr. Berners-Lee et al)... Best that they offer are toy examples of figuring out how to find out all Harvard - graduated doctors in 5 mile radius that take new patients (toy in the sense that they are fairly contrived, and that the details are conveniently omitted... claiming implementation is easy enough to be left as an exercise to reader)
I think Google is living proof of what exactly is the biggest known improvement over plain text indexing: inter-document references (links).
Ok, perhaps I misunderstood what you were saying... did you not claim that sender-based blocking is ok (or less evil) than content based one, in regards to free speech issues? I don't think that really is true; I don't think it really matters what the blocking (aka censoring) is based on, but more on who is doing blocking.
One could also argue that discriminating based on origin of communication is "more evil" than one based on content... After all, there are restrictions on free speech, but all are indeed based on content (the tired old "don't yell 'fire' at a crowded theatre" etc), and none restrict based on who is doing talking (except for individual court orders... which are interesting exceptions). I don't think there's huge difference one way or the other, but YMMV.
Of course this has nothing do with effectiveness of various blocking methods, which was your main point.:-)
Content based filtering also is a direct violation of the principles of the US First Amendment right to free speech (although the actual... ).
Infringement of free speech happens when the decision is based on what the content is.
Actually, you are missing the actual active ingredient of infringement. That would be that censoring is done by a third party. Whether that is the case depends on who is using the content based filtering.
Perhaps this is just nitpicking, but if end-user is the one who decided to use such a filter, no infringement can occur. Although people are free to speak their mind, no one is forced to listen to them.
What is not nitpicking, however, is pointing out that in this respect both content and sender-based filtering done by a third party WOULD indeed constitute an infringement (if done by gov't etc. etc.). There is no difference whatsoever.
Re:What is an example that can't run in parallel?
on
Forget Moore's Law?
·
· Score: 1
Others have pointed out obvious examples, about the only one I'd add is pretty much any encryption algorithm that needs feedback (stream ciphers in one of the standard modes). Calculation has to be done sequentially, part of the reason being that way same plaintext does not result in same encrypted text (except if all previous data happens to give same state, which is rare). This is similar to CRC calculations somebody else pointed out.
However, on top of fundamental barriers on system that rely on previous calculations (or have side effects), there is the practical one; communication overhead. Even with many fairly easily parallelizable (?) algorithms, you need very high bandwidth (and low latency) to make things work efficiently. And in many cases, theoretically easy-to-divide algorithm just doesn't scale into real-world systems.
And it is not unfair to dismiss someone's opinions by simply pointing at someone who has
done a better job of the argumentation. It is the UNLISTENING dismissal which you dislike,
not dismissal in general, which may well be reasonable.
Well put. I wasn't arguing it's unfair to dismiss opinions per se at all. But as original author argued, dismissing comments by ACs just because they are anonymous cowards (without considering opinion itself) is intellectually retarded. Plus you are absolutely right about quoting; merely using same (or even similar) phrase quoting does not in itself make.
About quoting in context; I didn't really mean to emphasize correct quoting context (which is important in publishing etc), but rather the general context of any message, not just quotes. Quote example I used since comment I replied to used it.
No. What you need to know is the context; knowing author may help in determining that, or it may as well hinder understanding real context (like prejudices against ACs, for example).
In case of "freedom is slavery", it seems more like quoting instead of (or in wrong) context, than a case of having to know who said it.
Even in case of people whose opinions you hate (say, neo-nazis), it would be preferable to argue their opinions (or facts), instead of just first labeling, then dismissing based on that label.
I believe sarcasm was aimed at "enterprise class OS to the x86 market" comment. Many already consider couple of OSes in this segment (various BSDs, Windows NT and its ilk, Linux) to be "enterprise class", and x86 version of Solaris is generally considered not to be nearly as good technically as its Sparc counterpart.
True, but I dont think enough time has passed since their decision, nor is their significant consensus about it being "the right thing",
to really call it "obviously right".
One could argue that it "seems to have been right thing to do" or "apparently it was right thing to do", but it's hardly fair to say "it was obviously right".
Poor availability. Doh. Of course the backups
would be duplicated to degree required for reasonably statistical probability for having backups. It wouldn't be 4 machines, one 1/4 for each. It could be 100k machines, of which, say, 1000 have some part of backups, of which only small portion (say, 4, 10). Of course there's always a chance some piece is missing, but hey, chance for a total loss is very small as well.
If a peer drops off the network... see above. For any reasonable system there shouldn't be dedicated peers (or if there are, they'd have couple of levels of secondary servers).
Massively prone to poor cryptography. If you use very strong cryptography, the system becomes very slow.
And I never saw any slowdown for my SSH connections? Nowadays I can guarantee it's your network bandwidth that restricts throughtput, not encryption algorithm. Slowdown is a myth, for all normal (strong enough, symmetric) real world encryption.
RAID, lower reliability. Doh. Did you even read the article? The whole idea is that RAID that resides in same physical location as the system, has significant risk of physical damage, AT THE SAME TIME AS primary system gets damaged. That is, when the building burns down, your precious RAID system will be as dead as the system it's supposed to protect. If you need to lower risks, you will of course use both RAID and distributed system... RAID has its advantages, being local, so that restoration for actual system failure is faster and more reliable. But for true catastrophes (a plane hitting big building your data center is in, earthquake, gas explosion) non-distributed backups are useless.
You may want to check
Freenet project for some ideas and discussion.
One way to look at it is that if (due to encryption) person who is holding the data can not (automatically) figure out what is stored, it seems that they can not be held accountable for the material (bit like if somebody buried a corpse in your backyard while you were away or such).
The way Freenet accomplishes this (if I remember correctly) is that the only way to find actual content is to try to search
for documents using its key (which although not required, usually is a filepath/URL-like construct). This id is hashed and used as a key (plus used for encryption/decryption). There are no ways to do wild card or fuzzy searches. And this is also why Freenet isn't ideal for Gnutella style "sharing" of copyrighted data.
So what's the point if only person who created the
document knows its id? There are lists of "well-known" ids that are published on Internet,
that can be used for fetching the data. Thus, "directories" are decoupled from actual data.
In fact, it would seem to be that some of Freenet's ideas would be VERY interesting to use with the distributed backup projects as well?
(for example; on top of Freenet's model, just
encrypt your backups with your private key... if someone tries find your backups, they first have to figure out the id, and then try to crack it open... split the backups into multiple pieces and make tasks even more difficult)
Of course, all this data security may mean trading
the actual backup feature to higher security;
someone may overwrite your backups if they can
figure out the id, even if they can not acces sit.
This is not a viable option because the potential for theft is too great, and
no ammount of encryption will make a difference. Encryption will always be broken.
If you are saying that no encryption method can be proven to be 100% secure, sure enough. But such total proof is unnecessary, just as with all real-life encryption. Are you saying SSL (or SSH for that matter) is useless because there's a slight chance somebody somewhere someday might be able to crack such communication?
It all depends on feasability of acquiring such data from competitor (enemy, whatever), compared to other methods. With reasonable encryption being used combined with fairly simple distribution of parts of incomplete data (say, your backups split into multiple pieces of which certain number are needed to be able to recover), decryption by non-owner requires not only tons of brute force (or a new yet unknown decryption without key), but also lots of co-operation from parties that actually host backed up data.
At which point perhaps it just seems that there are better ways to steal your data than try to obtain encrypted backups and crack it open.
And those who can design large scale systems are usually not so good at the nitty gritty...
I disagree. My definition of a good architect is that that's the person who -- given enough time -- could single-handedly both design and implement the whole system (except for the biggest of systems with very specialized sub-components or domain problems).
And the reason why architects might not end up implementing anything is just a matter of limited supply of people of that caliber, and large size of many modern software projects. Not because they lack implementation skills.
I have seen architects that couldn't program their way out of a wet paper bag, and trust me, none of them has been worth their salary. Such "architects " mostly follow buzzwords, and cause grief to no end to poor souls that need to implement their poor designs.
And I have also seen good ones (luckily, working with one at the moment), and they, surprise surprise, often do implement specific parts, and pretty much always help junior developers in getting implementation right.
but I think you are totally off base with regard to CDATA sections. If anything, they make life easier for the parser, not harder -- at least when I was writing a parser, CDATA made things faster and easier.
Well, CDATA is nothing like entities no matter what, and parsing them is not THAT much of a problem. But there are some subtleties, when
programmatically accessing CDATA blocks (when manually editing there are no problems; it's
just automatic processing that's trickier).
In any case, CDATA is an extra feature that's not really "needed"; normal quoting can be used
instead... it's a convenience feature.
However, doing quote/unquote when parsing/outputting is a breeze as well, and outputting CDATA sections automatically and reliably is tricky. At least if you want to do it 100% foolproof (granted, need to include ]]> token anywhere is a small, but still a possibility). You need to check for ]]> and split contents into two.
When parsing contents the problem (minor I guess) is that whereas text segments are usually normalized (ie. when reading a new doc without mods, you never get 2 text segments in a row), you can get combination of CDATA blocks and text segments; there's no way for parser to combine them on the fly (well, DOM-parsers do have the
normalize method that may do combination? Or perhaps that's not allowed by specs?)
But just like entities, CDATA is meant for manual
quick-quoting of blocks, and makes it easier for humans to quickly understand the contents. For programs it doesn't make big difference.
... oh yeah, I probably spent too much time when writing my own streamable XML-parser; when not doing full read in memory things to get more involved.:-)
(streamability, ie. reading only what is needed currently while still managing some structure unlike SAX, was needed to handle > 100 meg XML export files... implemented both read-only and write-only versions for internal use).
Agreed. However, it is true that most people would consider some feature or another of XML unnecessary and bloated. I would consider various entities (everything but character entities) to be such a feature, for example. I'm not missing C-style #includes in Java, and same goes for XML. If inclusions are needed, another XML-extensions should be used (isn't there something like XInclude?).
Second feature I'd consider removing would be CDATA sections. It is nifty when manually modifying XML, but otherwise it's just a pain (not a huge one for XML-parser, but additional bloat).
For other people list would look different I'm sure.:-)
Chances are many would. However, depending on exactly how easy it is to build a cheap linux box from XBox, I'd think this would be a good alternative to using "that old pentium box" as your print or web server or router. I for one would like the option of having 200-300$ up-to-date hardware box, to act as, say, wireless hub between DSL modem and wireless home network.
In those cases chances are games wouldn't be all that interesting.
Still, a valid point... and it'd be nice if MS considered it too. Trying to make games tempting enough to make mod'ing non-lossy for company; perhaps even profitable all in all.
You are assuming desk top apps are the only kinds of things written in a certain programming language.
Currently Java is one of the major languages on server side (at least for new code being written); for enterprise web apps probably the number one implementation language.
As to desk top apps, we'll see. Sun hasn't really been concentrating on that area too much lately, although there are some new useful things (Java Web Start especially) that would make app writing (or, rather, installation) easier.
FWIW there are lots of new Open Source Java application projects being started, check out SourceForge for examples. Of existing apps JEdit, Limewire and DBVis are ones I use (for text editing, gnutella access and DB debugging) regularly and find pretty useful.
You do realize that Netscape is based on Mozilla 1.x, which is a total rewrite, and shares no code with Netscape 4 series?
FWIW, Gecko (Mozilla's HTML layout engine) is supposedly reasonably lean; Mozilla itself is more bloated due to featuritis (although, many of the features are cool, from JavaScript debugger to the whole UI framework that seamlessly ties C, C++ and JavaScript). However that's not so much a side-effect as a design decision; architecture is ambitious and feature list (too?) sizable.
Slashdot Mirror
I guess "you are under arrest for breaking Murphy's Law" would be kinda useful phrase, come to think of it.
So you are saying that if there were actual US privacy laws, big companies like eBay would just ignore them? And leave themselves vulnerable to class action lawsuits?
Point is exactly that biggest companies are in best position to collect personal information, and then sell it. But they are also the easiest ones to control through legislation. So, gee, in this particular case laws might actually be reasonably effective.
And a lousy start at that. Like n+1 posters have pointed out, the problem is not about shutting down illegal sites, it's about forcing others (in this case, ISPs), to do the work of law enforcement and court system. And neither of those should be duties of entities outside local and federal police and US court system.
To put simply; hand the list of confirmed kiddie porn sites to ISP, demand them to cut the connection, show the court order, and that's it. Do not try to weasel away from your responsibility, mr. Judge.
As to publicity, I've never really tried to "optimize" ranking of my home pages, so I haven't even read their policies. It's good they give information about the frequency, though.
Of course, I'm Yet Another of those "used Altavista for years, then switched to Google never looked back" users... so what do I know about AV's current usefulness. :-)
Plus, even some of "sexier" careers (according to many people), are only interesting to people who share the same interests or ideals. To me, work of a skateboarding hero or world's best surfer, doesn't really sound all that interesting. But for others it would be. And I'd venture a guess similar thing applies to many programmers (ie. they might find, say, Linus Torvald's work fairly interesting).
This all reminds me of my ex-boss explaining what war is (or at least used to be, around WW II): "it's days, weeks, months, of doing pretty much nothing, farting around, then there's some hectic action and people die, and then there's lots more of the boring stuff". I guess what I'm trying to say is just that movies are only interesting by making non-realistic but believable entertainment out of otherwise (mostly) dull world.
Other replies already covered the main points, but as to the last one, AFAIK XEmacs works nicely with UTF-8 (since version 20 or so?).
In fact, if XML (and similar) markup mechanism were being designed to help searching, they certainly could help. But as things are that's not the order -- usually "content specialist" concentrate on classifying things without considering how those classifications are to be used; then try to force content authors to use markup, and only then start asking developers to develop system that make use of their fabulous new markup system.
And then the markup evangelists (like mr. Berners-Lee et al)... Best that they offer are toy examples of figuring out how to find out all Harvard - graduated doctors in 5 mile radius that take new patients (toy in the sense that they are fairly contrived, and that the details are conveniently omitted... claiming implementation is easy enough to be left as an exercise to reader)
I think Google is living proof of what exactly is the biggest known improvement over plain text indexing: inter-document references (links).
Hmmh. Ever heard of UTF-8 encoding? If not, you probably should check it out...
One could also argue that discriminating based on origin of communication is "more evil" than one based on content... After all, there are restrictions on free speech, but all are indeed based on content (the tired old "don't yell 'fire' at a crowded theatre" etc), and none restrict based on who is doing talking (except for individual court orders... which are interesting exceptions). I don't think there's huge difference one way or the other, but YMMV.
Of course this has nothing do with effectiveness of various blocking methods, which was your main point. :-)
Infringement of free speech happens when the decision is based on what the content is.
Actually, you are missing the actual active ingredient of infringement. That would be that censoring is done by a third party. Whether that is the case depends on who is using the content based filtering.
Perhaps this is just nitpicking, but if end-user is the one who decided to use such a filter, no infringement can occur. Although people are free to speak their mind, no one is forced to listen to them.
What is not nitpicking, however, is pointing out that in this respect both content and sender-based filtering done by a third party WOULD indeed constitute an infringement (if done by gov't etc. etc.). There is no difference whatsoever.
However, on top of fundamental barriers on system that rely on previous calculations (or have side effects), there is the practical one; communication overhead. Even with many fairly easily parallelizable (?) algorithms, you need very high bandwidth (and low latency) to make things work efficiently. And in many cases, theoretically easy-to-divide algorithm just doesn't scale into real-world systems.
Well put. I wasn't arguing it's unfair to dismiss opinions per se at all. But as original author argued, dismissing comments by ACs just because they are anonymous cowards (without considering opinion itself) is intellectually retarded. Plus you are absolutely right about quoting; merely using same (or even similar) phrase quoting does not in itself make.
About quoting in context; I didn't really mean to emphasize correct quoting context (which is important in publishing etc), but rather the general context of any message, not just quotes. Quote example I used since comment I replied to used it.
In case of "freedom is slavery", it seems more like quoting instead of (or in wrong) context, than a case of having to know who said it.
Even in case of people whose opinions you hate (say, neo-nazis), it would be preferable to argue their opinions (or facts), instead of just first labeling, then dismissing based on that label.
I believe sarcasm was aimed at "enterprise class OS to the x86 market" comment. Many already consider couple of OSes in this segment (various BSDs, Windows NT and its ilk, Linux) to be "enterprise class", and x86 version of Solaris is generally considered not to be nearly as good technically as its Sparc counterpart.
One could argue that it "seems to have been right thing to do" or "apparently it was right thing to do", but it's hardly fair to say "it was obviously right".
And I never saw any slowdown for my SSH connections? Nowadays I can guarantee it's your network bandwidth that restricts throughtput, not encryption algorithm. Slowdown is a myth, for all normal (strong enough, symmetric) real world encryption.
One way to look at it is that if (due to encryption) person who is holding the data can not (automatically) figure out what is stored, it seems that they can not be held accountable for the material (bit like if somebody buried a corpse in your backyard while you were away or such).
The way Freenet accomplishes this (if I remember correctly) is that the only way to find actual content is to try to search for documents using its key (which although not required, usually is a filepath/URL-like construct). This id is hashed and used as a key (plus used for encryption/decryption). There are no ways to do wild card or fuzzy searches. And this is also why Freenet isn't ideal for Gnutella style "sharing" of copyrighted data.
So what's the point if only person who created the document knows its id? There are lists of "well-known" ids that are published on Internet, that can be used for fetching the data. Thus, "directories" are decoupled from actual data.
In fact, it would seem to be that some of Freenet's ideas would be VERY interesting to use with the distributed backup projects as well? (for example; on top of Freenet's model, just encrypt your backups with your private key... if someone tries find your backups, they first have to figure out the id, and then try to crack it open... split the backups into multiple pieces and make tasks even more difficult)
Of course, all this data security may mean trading the actual backup feature to higher security; someone may overwrite your backups if they can figure out the id, even if they can not acces sit.
If you are saying that no encryption method can be proven to be 100% secure, sure enough. But such total proof is unnecessary, just as with all real-life encryption. Are you saying SSL (or SSH for that matter) is useless because there's a slight chance somebody somewhere someday might be able to crack such communication?
It all depends on feasability of acquiring such data from competitor (enemy, whatever), compared to other methods. With reasonable encryption being used combined with fairly simple distribution of parts of incomplete data (say, your backups split into multiple pieces of which certain number are needed to be able to recover), decryption by non-owner requires not only tons of brute force (or a new yet unknown decryption without key), but also lots of co-operation from parties that actually host backed up data.
At which point perhaps it just seems that there are better ways to steal your data than try to obtain encrypted backups and crack it open.
I disagree. My definition of a good architect is that that's the person who -- given enough time -- could single-handedly both design and implement the whole system (except for the biggest of systems with very specialized sub-components or domain problems).
And the reason why architects might not end up implementing anything is just a matter of limited supply of people of that caliber, and large size of many modern software projects. Not because they lack implementation skills.
I have seen architects that couldn't program their way out of a wet paper bag, and trust me, none of them has been worth their salary. Such "architects " mostly follow buzzwords, and cause grief to no end to poor souls that need to implement their poor designs. And I have also seen good ones (luckily, working with one at the moment), and they, surprise surprise, often do implement specific parts, and pretty much always help junior developers in getting implementation right.
Well, CDATA is nothing like entities no matter what, and parsing them is not THAT much of a problem. But there are some subtleties, when programmatically accessing CDATA blocks (when manually editing there are no problems; it's just automatic processing that's trickier). In any case, CDATA is an extra feature that's not really "needed"; normal quoting can be used instead... it's a convenience feature.
However, doing quote/unquote when parsing/outputting is a breeze as well, and outputting CDATA sections automatically and reliably is tricky. At least if you want to do it 100% foolproof (granted, need to include ]]> token anywhere is a small, but still a possibility). You need to check for ]]> and split contents into two. When parsing contents the problem (minor I guess) is that whereas text segments are usually normalized (ie. when reading a new doc without mods, you never get 2 text segments in a row), you can get combination of CDATA blocks and text segments; there's no way for parser to combine them on the fly (well, DOM-parsers do have the normalize method that may do combination? Or perhaps that's not allowed by specs?)
But just like entities, CDATA is meant for manual quick-quoting of blocks, and makes it easier for humans to quickly understand the contents. For programs it doesn't make big difference.
(streamability, ie. reading only what is needed currently while still managing some structure unlike SAX, was needed to handle > 100 meg XML export files... implemented both read-only and write-only versions for internal use).
Second feature I'd consider removing would be CDATA sections. It is nifty when manually modifying XML, but otherwise it's just a pain (not a huge one for XML-parser, but additional bloat).
For other people list would look different I'm sure. :-)
Still, a valid point... and it'd be nice if MS considered it too. Trying to make games tempting enough to make mod'ing non-lossy for company; perhaps even profitable all in all.
Currently Java is one of the major languages on server side (at least for new code being written); for enterprise web apps probably the number one implementation language.
As to desk top apps, we'll see. Sun hasn't really been concentrating on that area too much lately, although there are some new useful things (Java Web Start especially) that would make app writing (or, rather, installation) easier.
FWIW there are lots of new Open Source Java application projects being started, check out SourceForge for examples. Of existing apps JEdit, Limewire and DBVis are ones I use (for text editing, gnutella access and DB debugging) regularly and find pretty useful.
FWIW, Gecko (Mozilla's HTML layout engine) is supposedly reasonably lean; Mozilla itself is more bloated due to featuritis (although, many of the features are cool, from JavaScript debugger to the whole UI framework that seamlessly ties C, C++ and JavaScript). However that's not so much a side-effect as a design decision; architecture is ambitious and feature list (too?) sizable.