Lots of other comments, as well as the story, have pointed to a number of good tutorials and introductions. I'd like to recommend also the one I wrote for IBM developerWorks. I believe my tutorial is a bit better for real beginners to FP than are most of the others out there.
The magic number 99.84% is one that is sometimes invented as an arbitrary example, meaning roughly "very close to all." It's a sort of joke about false precision. Whether or not Bill Yerazunis is using this number in this fictive sense, it is IMPOSSIBLE for his diff() to actually be exactly 99.84%!
If one message out of 3000 messages differs in classification, that's 0.0333%. Or 99.9666% accurate. Working down, we find that four or five misclassifications are either 99.8666% or 99.8333% respectively. Both are certainly in the same ballpark as the stated accuracy, but neither is correctly rounded to Yerazunis' number. To me, this pretty much proves that the pseudo-exact figure is used in a fictive sense, not as a n actual measurement.
It *IS*, however, true that a person being careful will make occassional errors nonetheless.
The suggestion has come up from time-to-time that it would be a "distributed attack" on SCO to buy their shares. This idea is misguided: both because it props up share price and subsidizes wrong-doing, and also just because most shares are not publicly traded.
But a genuine distributed attack would be for many authors whose copyright SCO is violating to individually sue SCO for statutory damages. There are certainly hundreds of different applications under GPL that SCO continues to distribute, in violation of licensing terms. Hundreds of lawsuits in hundreds of different courts would certainly be fun to follow (and would probably speed up the bankrupcy of SCOX).
Lots of people have commented generally about the fact there is massive prior art on virtual desktops controlled with a pager. I'm wondering what the earliest example people know about is. I remember using one on probably Windows/286, and something like OS/2 2.0. And I think I saw it on some old Sun workstations around the same time, or maybe earlier.
But I would not be too surprised if the implementation went back even farther--say to a utility for the first Macs, or on one of those Xerox/Alto prototypes. What's the first thing you know of? And even better, do you have a screenshot of such an early version?
OK... maybe this note is a little karma whoring by the author of the referred article. But what the heck, I -did- write it, and I'm always pleased when things by or about me make the slashdot headline.
A question a bunch of people have raised in comments is "why bother with Linux if you buy a Mac with OSX on it"... well, read the article. I don't say that switching to or adding Linux is always desirable; but I think I do a good job of describing some scenarios where it is. That said, I certainly -do- like OSX quite a bit (where my favorite installed application is still bash:-)).
Also, contrary to some down-modded poster, IBM can INDEED easily handle the load of slashdotting. In fact, a zillion hits to my article is indirectly good for me (I'm not on commission or anything, but it puts an extra sparkle in my editors' eyes). Also, FWIW, all my articles soon make it to [http://gnosis.cx/publish/] (which reminds me that I need some updating, it's been a couple months)... which is also quite strong enough to survive/.-ing.
... Porn was over 1/3rd of all traffic on the net, last time I checked
Another poster also observed that the parent comment is a bit on the absurd side. Not necessarily that the fact is wrong even; but it's not like my personal Apache log shows "traffic on the net." Unless downix works for the NSA, and is leaking some classified information, there's really no way to verify what everyone on the net does.
BTW, I "5-figure takehome salary" could be $10,000... which I would actually call rather shabby (though if it's for part time work, it might still be good). Claiming something like "high 5-figure" would sound a lot more impressive (i.e. meaning >$50k).
Ah yes, quite right. I tried the command on a much smaller directory of my own files (just to double check the syntax). I forgot about the fact the kernel file list would hit the argument list limits.
BUT, nothing in mathematics is just 'made up'. All mathematics is based on fundamental, logical axioms (truths), and if anything were to violate those axioms, or the completely logical conclusions drawn therefrom, it would not be mathematics.
Yeah, right! Like if you were to try to do Set Theory without the Axiom of Choice it wouldn't be mathematics; or Geometry without the Parallel Postulate, likewise. Because the axioms of mathematics are just built-in to the nature of thought, or the universe, or something... [sarcasm off]
Don't confuse social constructive ideas--or for that matter, mathematical constructivism--with some stupid caricature some right wing ideologue spouted on the Fox Network. Thinkers like Kuhn, Feyerabend, Foucault, Latour, even Lyotard or Baudrillard are far more rigorous and intelligent than their critics are (the critics have no idea what the social constructivists are saying in the first place).
It just so happens that I've both done some real work in social constructive philosophy of science... and also know VASTLY more mathematics than does the parent poster. Don't settle for ignorance!
From what I can see, Graham-Cummings' trial-and-error approach is way too dopey. I use a home-grown filter, that I developed for my article:
http://www-106.ibm.com/developerworks/linux/libr ar y/l-spamf.html
Actually, I've tweaked it a bit since then, but basically the idea is the same (I wrote this before many of the other Bayesian tools were ready for prime time).
The thing is, I don't need to use trial-and-error to find out what words (or trigrams, in my case) are the hammiest. I have a little utility to read the database and spit them out. While I supose I'd need to actually run the calculation to see exactly how many words were needed to meet the ham threshhold, there's absolutely no mystery about which words look nicest to the filter.
But of course, my ham words are not the same as your ham words. For that matter, they won't be the same words once I update my model (I've been remiss in doing it, since it's remained pretty accurate for a couple months... my tool updates by batch, not per every message). So WHO CARES about the fact a few of my personal ham words might get spam by.
Well, first thing, of course, is that Apple CPUs are IBM CPUs. For the G5, the whole design is from IBM; but even the G4s are fabbed by IBM (though designed by Motorola). So big blue likes the PowerPCs all around.
Still, the IBM announcement makes some sense of my IBM developerWorks editor's push for a rush job on an article about Linux on PowerPCs. Despite the rush, I think it turned out well. Take a look at:
Over at EVM2003 and the Open Voting Consortium we are addressing the problems with proprietary and paperless voting systems in the concrete, and in a reasonably short-term time frame. The thing to keep in mind is that the problems are mostly political ones, not technical ones. Cryptographers tend to miss this fact.
From: David Mertz %lt;voting-project_at_gnosis_dot_cx%gt;
Date: Tue Nov 25 2003 - 12:16:05 CST
|I found [the Chaum paper] here in case anyone wants to read it.
|http://www.vreceipt.com/
Thanks Clay, for looking up this paper. It is consistent with what
Mercuri described more briefly, but I found reading the white paper to
contain additional interesting details. Btw. for other readers: the
press release at that URL is fine for a summary, but look at the linked
white paper for real information.
Reading the paper, I see that Chaum's system really is flawed in
practice. The reason it is flawed is precisely because Chaum is TOO
smart--he's great at math, but misses the real world of elections.
One weakness of the system is the one I've raised a couple times.
Voters cannot understand how the system works (in any meaningful
detail). For example, imagine I were a voter who did not have any
graduate-level mathematics training (a large majority of voters, I
think... probably a majority of this list, in fact). Now imagine that I
was not entirely trustful of "the experts", and worry that someone can
puncture the anonymity of my vote by properly analyzing my receipt.
Sure it doesn't contain a visually readable vote, but I know in a
general way that there are barcode scanners, and clever things that
mathematicians and CS people do.
In answer to my concern, all I really get back is the Diebold-style
answer: "Trust us, we're very smart, and we wouldn't let any errors
exist in our voting system." I don't think this answer inspires general
voter confidence. I personally happened to have already known about
Chaum before hearing about this system, and basically trust his motives
and intelligence... but how many voters can say that; how many LIST
MEMBERS can say that, even?
The second weakness is the real world of voting places--typically a
hastily arranged room in a church or a community center, staffed by
well-meaning, but amateur volunteers. Imagine that prior to the
election some guys with brass knuckles stop by my house, and let me know
that they would appreciate a vote for their candidate (or equally, for
example, a coercive or manipulative spouse or relative). As a gesture
of good faith, they suggest, I should keep both layers of the voting
receipt so that it remains clear how I voted.
According to Chaum's system, the poll workers are SUPPOSED TO shred one
layer on my way out. Anyone who has been to a polling place knows that
it would not be of great practical difficulty to "forget" to place a
layer in the shredder prior to leaving the building. Even should I make
such an omission, the electronic vote was already recorded when the
receipts were printed.
+++
Btw. With the EVM2003 system, a related forgetfullness is possible.
Voters might forget to place the printed ballot in the ballot box.
Their electronic ballot is still recorded on the machine, but only a
subset of electronic ballots will be matched by corresponding printed
ballots in the ballot boxes. Hopefully, this will generally be a large
subset (98%+ say), but a certain discrepency rate must be expected.
Placing cryptographic codes on the printed ballots allows us to assure
that every such ballot is -legitimate-, hence preventin
The thing about information is that it's not quite so easy to count as the article suggests. If the question is solely one of how many magtapes to buy, sure the exabyte thing is interesting enough. But in a "human" sense, that's not all that interesting.
For example, the article cites 18 exabytes of what is basically analog data--sound and images--over telephone, radio, TV. It claims that 98% of that is in telephone calls, essentially all, in other words.
First thing is that most telephone calls are not recorded. Well, I dunno, maybe Carnivore and Eschelon are even worse than I think. So mostly this is just a question of how much bandwidth AT&T and MCI need to buy; I'm sure they care about that question, but most people have no reason to. Maybe how many tape drives the NSA needs to buy too.
Just how much information *IS* there in a telephone call though. At a certain level, ten million calls about the same snowstorm aren't really that information rich. But I understand that you want to hear YOUR sister complain about shoveling the snow, not somebody else's sister do so. But just at a technological level, how much is there to a phone call?
If I record the call as CD-Audio WAV format it comes to something like 9 MB a minute. But then, if I compress it to MP3, or Ogg Vorbis, or AAC, I'm down to something more like 1 MB a minute. In fact, if I go for a 56k bandwidth, or something along those lines, I can probably get it down to less than half a MB... and that's not really much different from what I could discern originally on my cell-phone on a noisy street, or over my old wiring in my house. So far, we've reduced the "information content" by 20 times by purely technial means. Then again, it's not clear if this is fair... in those cop shows where they reconstruct background noises to filter the gunshot or car crash in the background, they probably want the full original data... but do *I* care about that when I talk to my sister?
Moreover, audio compression is just the start. There's this old thing called TRANSCRIPTION that compresses quite a bit more. A stenographer (or maybe a computer program, at least at the NSA) can type up our conversation perfectly well. How much information is lost by reducing the "data" to:
Lulu's Sister: We got over 10" of snow, and it took me an hour to shovel it.
Even at the highest audio compression I can find, I need tens of kilobytes to encode this remark... as text we're down to a couple tens of BYTES. Maybe I've lost a little of Sis's inflection, but how much INFORMATION was there really, to start with? Some probably, but is it worth a thousand words? Moreover, I expect some lossy compression to reduce that text by at least another half.
Depending on just what you think is information, perhaps 300,000 times compression is possible. That brings exabytes down to gigabytes. Given some automated transcription technology, maybe I can store the whole last year of family chats on my local harddisk!
...it looks like Amazon will take a book off the list if the author insists, so there really isn't too much of a problem here.
Think about how very fond you are of "opt out" email. The idea that an author could remove their book, after some elaborate procedure, if they are aware their book is indexed in the first place, is less than compelling.
Mind you, even as an author (but one whose writing if available for free, as well as for money), I'm not per se agreeing with the Author's Guild. What I can see on Amazon looks like fair use quotations. But it might well be possible to easily reconstruct more of the text in a book that would qualify under fair use.
One thing to keep in mind is that authors generally get majorly screwed over by publishers. E.g. Random House isn't really a whole lot more interested in "protecting authors" than the RIAA is in "protecting musicians".... so if a publisher has given permission, don't imagine they do it to help authors, nor even in conformance with the contracts they signed with those authors.
In fact, the conceit that closed source software can ever produce fair voting systems is misguided. Unfortunately, even many of the experts who understand much of what is wrong have failed to make the final step--I suppose because of their lack of background in free software (e.g. David Dill, Rebecca Mercuri). They understand the need for paper trails, but don't (yet) quite get why the underlying machines MUST be open source. Narrowly, their threat model is incomplete; but more broadly, they do not understand the realpolitik of getting machines into polling places.
The right solution is one that I am working on (with many other experts in the field, and some committed people). See EVM2003. However, as with many volunteer projects, we've been delayed a bit (developer schedules and so on). I could really use a couple more good volunteers, especially ones who are familiar with wxPython, which is the library/language for the demo touchscreen interface. However, other folks who might want to work on the blind-accessible vocalization interface, documentation, or other parts, would help too.
'cperciva's analysis is wrong. Stipulating the unlikely fact that IBM/SGI DID violate some term of the license, that gives SCO action to enjoin the VIOLATION... pointedly, NOT to revoke the original rights. Under the stipulation, SCO would also merit recovery of resultant damages. But there's just no way it can touch the underlying licensing rights, no matter what.
Indian programmer are, on average, just about as good as US programmers. Some are great, some are terrible. But it's not nationality that does it. Same goes for Russian, Chinese, South African, or any other programmers mentioned in the thread. I've had plenty of problems with US-born programmers too; it's nothing special.
What is a lot more likely is that hoggoth's outsourcing woes have to do with the *outsourcing* part of it. If you dump code on people without adequate specs, documentation, and yes, supervision, you are not going to get something good back. That fact has little to do with how much you pay them, in fact. And it gets even worse trying to move a project from someone who is intimately familiar with it to someone who has to scramble to figure out what it is/does--not that you always have the former to start with, but you always get the latter when you outsource.
I've posted some similar notes to most of the recent articles about problems with commercial voting machines. For this one, I really want to actively recruit some developers to help out. There are parts of EVM2003 that are on track, but other parts need more developers. In particular, we really need some people with experience in blind-accessessibility for that portion of the project (both a system to allow voting, and one to vocalize printed ballots).
The idea of EVM2003 is to create Free Software voting machine, and
to implement machines that also produce voter-verifiable paper trails
(i.e. visually readable printed ballots). We will do a number of
security things right, where the commercial companies have done them
wrong... they have aimed for "security through obscurity" or "just trust us." As well, part of our requirement is to have fully
blind-accessible voting that maintains complete anonymity.
Anyway, I (David
Mertz) have taken over as Developer Lead recently, and am
trying to move the development of the demo along.
Feel free to contact me--the standard ballot system (in the demo version at least) is being done in wxPython; but conceivably we would choose other languages/technologies for bar-code reading, printing, blind-voting, etc. (my preference is to use Python though, for consistency and rapid development).
When I wrote my Ph.D., I used the still quite excellent WordPerfect 5.1, for the most part. However, for better or worse (mostly for worse), I did wind up finishing it in WordPerfect for Windows (I think 6.1). At least the file formats were compatible, but in retrospect I should have stuck with the text screen (unfortunately, I recall that OS/2 actually ran the Windows version a bit better than the DOS one at the time though). FWIW, you can read that at: David Mertz Disseration, WordPerfect version; or since WP compatibility isn't so widespread nowadays: PDF page style and PDF book style. Back in those days, I wrote about PoMo philosophy and the like.
Flash forward a few years. Now I am a writer about computer programming. And mostly because of that transition, I absolutely cannot stand to write anything other than plain text. Well, almost plain text, I have my own little variant called "smart ASCII", which uses just a few of the conventions that email and Usenet often use: *bold*, -itals-, and so on.
Well... once in a while I am compelled to use something awful like MS-Word--or something that exports to it, like AppleWorks or OpenOffice--but I hate doing that. It is tools that convert my smart ASCII into formats like HTML, XML, LaTeX, PDF, and so on. But those tools come at the end of the process. After I put the words down, then is the time to worry about niggly details like fonts, layouts, and so on... all in a way that is far more consistent than a wordprocessor is likely to produce. My book, for example, has been praised as particularly attractive typographically... I did all the preparation myself, by eschewing all the GUI nonsense that gets in the way during writing.
David Mertz
What makes it really tough is when fraudulent credit companies--Sears is a big example--decides periodically "reactivate" your credit card, despite repeated cancellations.
On one occassion, Sears sent me an actual VISA card, at least 10 years after I had cancelled my Sears card (or even been in their store), without any authorization by me. I cut it up, and called to cancel. But on an earlier occassion, In only found out from a credit report that Sears was reporting I had a card with some large available credit line, which is, of course, detrimental to me when I want new credit. Of course, I had no card, did not receive montly statements, and had no outstanding balance... I just allegedly had $30k available credit line. Lots of work writing letters to credit agencies resulted.
I posted a comment on a related thread mentioning the project I am involved in, EVM2003. We had a slightly rocky start, as project do, but things are underway.
The idea of EVM2003 is to create Free Software voting machine, and to implement machines that also produce voter-verifiable paper trails (i.e. visually readable printed ballots). We will do a number of security things right, where the commercial companies have done them wrong... they have aimed for "security through obscurity" or "just trust us." As well, part of our requirement is to have fully blind-accessible voting that maintains complete anonymity.
Anyway, I (David Mertz) have taken over as Developer Lead recently, and am trying to get the development of the demo rolling. Part of that effort is recruiting some more developers, and splitting the project into several only loosely connected parts. Feel free to contact me--the standard ballot system (in the demo version at least) is being done in wxPython; but conceivably we would choose other languages/technologies for bar-code reading, printing, blind-voting, etc. (my preference is to use Python though, for consistency and rapid development).
The whole Segway on the mountain bit was amusing. But it's just plain alarming to see not just/., but also _The Register_ make a crudly common English usage error. The phrase "begs the question" DOES NOT mean "demands to be asked."
Instead...well, this is such a common error it's easy to research. But quoting from http://www.wsu.edu:8080/~brians/errors/begs.html:
An argument that improperly assumes as true the very point the speaker is trying to argue for is said in formal logic to "beg the question."
Slashdot Mirror
DO NOT MISUSE the phrase "begs the question"!
Please, for the love of (insert thing loved), just write "demands to be asked" or "prompts the question" if that's what you mean.
If you use "beg the question", PLEASE mean:
1 : to assume as true or take for granted the thing that is the subject of the argument;
2 : to dodge the issue.
Lots of other comments, as well as the story, have pointed to a number of good tutorials and introductions. I'd like to recommend also the one I wrote for IBM developerWorks. I believe my tutorial is a bit better for real beginners to FP than are most of the others out there.
Anyway, you can find it at IBM dW(free registration required) or at my own Gnosis Software site.
The magic number 99.84% is one that is sometimes invented as an arbitrary example, meaning roughly "very close to all." It's a sort of joke about false precision. Whether or not Bill Yerazunis is using this number in this fictive sense, it is IMPOSSIBLE for his diff() to actually be exactly 99.84%!
If one message out of 3000 messages differs in classification, that's 0.0333%. Or 99.9666% accurate. Working down, we find that four or five misclassifications are either 99.8666% or 99.8333% respectively. Both are certainly in the same ballpark as the stated accuracy, but neither is correctly rounded to Yerazunis' number. To me, this pretty much proves that the pseudo-exact figure is used in a fictive sense, not as a n actual measurement.
It *IS*, however, true that a person being careful will make occassional errors nonetheless.
Isn't there something deeply wrong with any article where that phrase occurs?
The suggestion has come up from time-to-time that it would be a "distributed attack" on SCO to buy their shares. This idea is misguided: both because it props up share price and subsidizes wrong-doing, and also just because most shares are not publicly traded.
But a genuine distributed attack would be for many authors whose copyright SCO is violating to individually sue SCO for statutory damages. There are certainly hundreds of different applications under GPL that SCO continues to distribute, in violation of licensing terms. Hundreds of lawsuits in hundreds of different courts would certainly be fun to follow (and would probably speed up the bankrupcy of SCOX).
Lots of people have commented generally about the fact there is massive prior art on virtual desktops controlled with a pager. I'm wondering what the earliest example people know about is. I remember using one on probably Windows/286, and something like OS/2 2.0. And I think I saw it on some old Sun workstations around the same time, or maybe earlier.
But I would not be too surprised if the implementation went back even farther--say to a utility for the first Macs, or on one of those Xerox/Alto prototypes. What's the first thing you know of? And even better, do you have a screenshot of such an early version?
OK... maybe this note is a little karma whoring by the author of the referred article. But what the heck, I -did- write it, and I'm always pleased when things by or about me make the slashdot headline.
:-)).
/.-ing.
A question a bunch of people have raised in comments is "why bother with Linux if you buy a Mac with OSX on it"... well, read the article. I don't say that switching to or adding Linux is always desirable; but I think I do a good job of describing some scenarios where it is. That said, I certainly -do- like OSX quite a bit (where my favorite installed application is still bash
Also, contrary to some down-modded poster, IBM can INDEED easily handle the load of slashdotting. In fact, a zillion hits to my article is indirectly good for me (I'm not on commission or anything, but it puts an extra sparkle in my editors' eyes). Also, FWIW, all my articles soon make it to [http://gnosis.cx/publish/] (which reminds me that I need some updating, it's been a couple months)... which is also quite strong enough to survive
Yours, David...
... Porn was over 1/3rd of all traffic on the net, last time I checked
Another poster also observed that the parent comment is a bit on the absurd side. Not necessarily that the fact is wrong even; but it's not like my personal Apache log shows "traffic on the net." Unless downix works for the NSA, and is leaking some classified information, there's really no way to verify what everyone on the net does.
BTW, I "5-figure takehome salary" could be $10,000... which I would actually call rather shabby (though if it's for part time work, it might still be good). Claiming something like "high 5-figure" would sound a lot more impressive (i.e. meaning >$50k).
Ah yes, quite right. I tried the command on a much smaller directory of my own files (just to double check the syntax). I forgot about the fact the kernel file list would hit the argument list limits.
I kinda prefer:
$ cat `find kernel-source -type f` | wc -l
It's a hair shorter to use the backtick argument expansion. Either one is a heck of a lot better than that perl monstrosity.
Yeah, right! Like if you were to try to do Set Theory without the Axiom of Choice it wouldn't be mathematics; or Geometry without the Parallel Postulate, likewise. Because the axioms of mathematics are just built-in to the nature of thought, or the universe, or something... [sarcasm off]
Don't confuse social constructive ideas--or for that matter, mathematical constructivism--with some stupid caricature some right wing ideologue spouted on the Fox Network. Thinkers like Kuhn, Feyerabend, Foucault, Latour, even Lyotard or Baudrillard are far more rigorous and intelligent than their critics are (the critics have no idea what the social constructivists are saying in the first place).
It just so happens that I've both done some real work in social constructive philosophy of science... and also know VASTLY more mathematics than does the parent poster. Don't settle for ignorance!
From what I can see, Graham-Cummings' trial-and-error approach is way too dopey. I use a home-grown filter, that I developed for my article:
r ar y/l-spamf.html
http://www-106.ibm.com/developerworks/linux/lib
Actually, I've tweaked it a bit since then, but basically the idea is the same (I wrote this before many of the other Bayesian tools were ready for prime time).
The thing is, I don't need to use trial-and-error to find out what words (or trigrams, in my case) are the hammiest. I have a little utility to read the database and spit them out. While I supose I'd need to actually run the calculation to see exactly how many words were needed to meet the ham threshhold, there's absolutely no mystery about which words look nicest to the filter.
But of course, my ham words are not the same as your ham words. For that matter, they won't be the same words once I update my model (I've been remiss in doing it, since it's remained pretty accurate for a couple months... my tool updates by batch, not per every message). So WHO CARES about the fact a few of my personal ham words might get spam by.
Well, first thing, of course, is that Apple CPUs are IBM CPUs. For the G5, the whole design is from IBM; but even the G4s are fabbed by IBM (though designed by Motorola). So big blue likes the PowerPCs all around.
Still, the IBM announcement makes some sense of my IBM developerWorks editor's push for a rush job on an article about Linux on PowerPCs. Despite the rush, I think it turned out well. Take a look at:
...I suppose from the TV guide.
As it happens I discussed Chaum's system just today on the Voting-Project mailing list. I guess I might as well quote myself:
The thing about information is that it's not quite so easy to count as the article suggests. If the question is solely one of how many magtapes to buy, sure the exabyte thing is interesting enough. But in a "human" sense, that's not all that interesting.
For example, the article cites 18 exabytes of what is basically analog data--sound and images--over telephone, radio, TV. It claims that 98% of that is in telephone calls, essentially all, in other words.
First thing is that most telephone calls are not recorded. Well, I dunno, maybe Carnivore and Eschelon are even worse than I think. So mostly this is just a question of how much bandwidth AT&T and MCI need to buy; I'm sure they care about that question, but most people have no reason to. Maybe how many tape drives the NSA needs to buy too.
Just how much information *IS* there in a telephone call though. At a certain level, ten million calls about the same snowstorm aren't really that information rich. But I understand that you want to hear YOUR sister complain about shoveling the snow, not somebody else's sister do so. But just at a technological level, how much is there to a phone call?
If I record the call as CD-Audio WAV format it comes to something like 9 MB a minute. But then, if I compress it to MP3, or Ogg Vorbis, or AAC, I'm down to something more like 1 MB a minute. In fact, if I go for a 56k bandwidth, or something along those lines, I can probably get it down to less than half a MB... and that's not really much different from what I could discern originally on my cell-phone on a noisy street, or over my old wiring in my house. So far, we've reduced the "information content" by 20 times by purely technial means. Then again, it's not clear if this is fair... in those cop shows where they reconstruct background noises to filter the gunshot or car crash in the background, they probably want the full original data... but do *I* care about that when I talk to my sister?
Moreover, audio compression is just the start. There's this old thing called TRANSCRIPTION that compresses quite a bit more. A stenographer (or maybe a computer program, at least at the NSA) can type up our conversation perfectly well. How much information is lost by reducing the "data" to:
Lulu's Sister: We got over 10" of snow, and it took me an hour to shovel it.
Even at the highest audio compression I can find, I need tens of kilobytes to encode this remark... as text we're down to a couple tens of BYTES. Maybe I've lost a little of Sis's inflection, but how much INFORMATION was there really, to start with? Some probably, but is it worth a thousand words? Moreover, I expect some lossy compression to reduce that text by at least another half.
Depending on just what you think is information, perhaps 300,000 times compression is possible. That brings exabytes down to gigabytes. Given some automated transcription technology, maybe I can store the whole last year of family chats on my local harddisk!
Think about how very fond you are of "opt out" email. The idea that an author could remove their book, after some elaborate procedure, if they are aware their book is indexed in the first place, is less than compelling.
Mind you, even as an author (but one whose writing if available for free, as well as for money), I'm not per se agreeing with the Author's Guild. What I can see on Amazon looks like fair use quotations. But it might well be possible to easily reconstruct more of the text in a book that would qualify under fair use.
One thing to keep in mind is that authors generally get majorly screwed over by publishers. E.g. Random House isn't really a whole lot more interested in "protecting authors" than the RIAA is in "protecting musicians".... so if a publisher has given permission, don't imagine they do it to help authors, nor even in conformance with the contracts they signed with those authors.
The right solution is one that I am working on (with many other experts in the field, and some committed people). See EVM2003. However, as with many volunteer projects, we've been delayed a bit (developer schedules and so on). I could really use a couple more good volunteers, especially ones who are familiar with wxPython, which is the library/language for the demo touchscreen interface. However, other folks who might want to work on the blind-accessible vocalization interface, documentation, or other parts, would help too.
Please email me: David Mertz (mertz@gnosis.cx)
'cperciva's analysis is wrong. Stipulating the unlikely fact that IBM/SGI DID violate some term of the license, that gives SCO action to enjoin the VIOLATION... pointedly, NOT to revoke the original rights. Under the stipulation, SCO would also merit recovery of resultant damages. But there's just no way it can touch the underlying licensing rights, no matter what.
Indian programmer are, on average, just about as good as US programmers. Some are great, some are terrible. But it's not nationality that does it. Same goes for Russian, Chinese, South African, or any other programmers mentioned in the thread. I've had plenty of problems with US-born programmers too; it's nothing special.
What is a lot more likely is that hoggoth's outsourcing woes have to do with the *outsourcing* part of it. If you dump code on people without adequate specs, documentation, and yes, supervision, you are not going to get something good back. That fact has little to do with how much you pay them, in fact. And it gets even worse trying to move a project from someone who is intimately familiar with it to someone who has to scramble to figure out what it is/does--not that you always have the former to start with, but you always get the latter when you outsource.
Yours, Lulu...
The idea of EVM2003 is to create Free Software voting machine, and to implement machines that also produce voter-verifiable paper trails (i.e. visually readable printed ballots). We will do a number of security things right, where the commercial companies have done them wrong... they have aimed for "security through obscurity" or "just trust us." As well, part of our requirement is to have fully blind-accessible voting that maintains complete anonymity.
Anyway, I (David Mertz) have taken over as Developer Lead recently, and am trying to move the development of the demo along.
Feel free to contact me--the standard ballot system (in the demo version at least) is being done in wxPython; but conceivably we would choose other languages/technologies for bar-code reading, printing, blind-voting, etc. (my preference is to use Python though, for consistency and rapid development).
Flash forward a few years. Now I am a writer about computer programming. And mostly because of that transition, I absolutely cannot stand to write anything other than plain text. Well, almost plain text, I have my own little variant called "smart ASCII", which uses just a few of the conventions that email and Usenet often use: *bold*, -itals-, and so on.
In fact, I have written hundreds of articles, tutorials, and the like about programming (for well-known publishers like IBM, Intel, O'Reilly, etc.), all in plain text. My book Text Processing in Python is written the same way.
Well... once in a while I am compelled to use something awful like MS-Word--or something that exports to it, like AppleWorks or OpenOffice--but I hate doing that. It is tools that convert my smart ASCII into formats like HTML, XML, LaTeX, PDF, and so on. But those tools come at the end of the process. After I put the words down, then is the time to worry about niggly details like fonts, layouts, and so on... all in a way that is far more consistent than a wordprocessor is likely to produce. My book, for example, has been praised as particularly attractive typographically... I did all the preparation myself, by eschewing all the GUI nonsense that gets in the way during writing. David Mertz
What makes it really tough is when fraudulent credit companies--Sears is a big example--decides periodically "reactivate" your credit card, despite repeated cancellations.
On one occassion, Sears sent me an actual VISA card, at least 10 years after I had cancelled my Sears card (or even been in their store), without any authorization by me. I cut it up, and called to cancel. But on an earlier occassion, In only found out from a credit report that Sears was reporting I had a card with some large available credit line, which is, of course, detrimental to me when I want new credit. Of course, I had no card, did not receive montly statements, and had no outstanding balance... I just allegedly had $30k available credit line. Lots of work writing letters to credit agencies resulted.
The idea of EVM2003 is to create Free Software voting machine, and to implement machines that also produce voter-verifiable paper trails (i.e. visually readable printed ballots). We will do a number of security things right, where the commercial companies have done them wrong... they have aimed for "security through obscurity" or "just trust us." As well, part of our requirement is to have fully blind-accessible voting that maintains complete anonymity.
Anyway, I (David Mertz) have taken over as Developer Lead recently, and am trying to get the development of the demo rolling. Part of that effort is recruiting some more developers, and splitting the project into several only loosely connected parts. Feel free to contact me--the standard ballot system (in the demo version at least) is being done in wxPython; but conceivably we would choose other languages/technologies for bar-code reading, printing, blind-voting, etc. (my preference is to use Python though, for consistency and rapid development).