Actually, the first thing that came to my mind was:
Finally! A real explanation for ZeoSync!
(For those who haven't been keeping up, ZeoSync is the company that claims to have broken / bypassed Shannon's laws of information entropy to create some sort of encoding or compression that can compress random data. Except that they don't call it compression - they call it "Information Crystals" or something equally stupid.)
I asked for and got ZeoSync's Investors Package, and it truly has some strange stuff in it... they are suing some previous employees, have some financial stuff that looks weird even to me (I know very little about corporate finance), and those computer scientists that are so prominently featured on their web site are not actually associated with the company - ZeoSync just paid them (an unspecified amount) for some sort of unspecified consulting. Basically meaningless.
They admit in the fine print that the alleged "technology" has never been demoed to anyone outside the company...
Really, I would not be surprised if ZeoSync was an elaborate ruse to teach gullible investors a lesson.
... ignoring the fact that most thieves would wipe linux off the computer and attempt to install a pirated copy of Windows ME - especially since they all come up at runlevel 3 with an unfriendly looking text-mode login prompt... Oh well...
All my machines already run SSHD all the time. So, I just need to find out where they are if they are stolen.
Solution: Write a little script that runs whenever any network interface comes up. The script would:
- check to see if an internet connection is available, if not, bail out.
- Otherwise determine the machine's DHCP-assigned IP address
- If PPP was used, find the phone numbers
- dumps the resulting info into a small text file and then uses "wget" to do an HTTP PUT and upload it to a CGI or PHP running on my website...
Whenever the machine connects to the internet, it will figure out where it is on the net and send that information to a log file that I can read....
Then I can secure shell in to the machine and wreak my terrible revenge, yea, even unto the third generation they shall speak of it in hushed voices with cautious glances over their shoulders...
Here's what I sent, feel free to copy anything that strikes a chord with you:
- - - - -
I'm writing to express my concern and disappointment with the proposed judgement in the Microsoft Antitrust case. It seems that the Department of Justice has won the case, only to concede defeat at the end.
As a software engineer, I've watched with disappointment for years as Microsoft has leveraged their desktop operating system dominance to crush one competitor after another.
I believe that few informed purchasers choose Microsoft products on their merits alone - rather, people buy Microsoft mainly because they need to exchange Microsoft Word and Excel documents.
One way to restore competition to the market would be to require Microsoft to completely specify and fully document their Microsoft Office file formats. No changes should be allowed without several months notice and complete documentation in advance.
This would allow competitors to create viable alternatives with the ability to interoperate with Microsoft, and would restore competition to an industry that badly needs it.
Please, consider a judgement which forces Microsoft to allow competitors to interoperate with their products.
Ahhh... A Fields medal winner (!) you know, I thought I'd heard his name before. This makes me even more suspicious that he doesn't really know what's going on with Zeosync.
I suspect he's just really, really busy, and was invited to be an advisor to the company... they probably told him they were doing compression research, and he may not know about the claims they are making...
Betcha that as soon as he finds out, he'll get himself disassociated from them ASAP.
Digging around I found some more interesting stuff. First of all, if you get the PDF org chart, the information on that essentially contradicts the Flash propaganda - the technical staff seems to be divided into two teams, one "Advanced Compression Technologies Team" and one "Singular Bit Varience [sic] Encoder Team".
The org chart also mentions Wavelets, Fractals, and Sub-band compression... So much for the website that claims their technology isn't actually compression...
Maybe, just maybe, the scientists actually do have some sort of interesting compression technology, but the marketing / business people have spun and hyped it up, totally out of control and totally out of touch with reality. But I don't think so - marketing people alone wouldn't be able to come up with the pseudo-scientific drivel on that website.
Moving on, you see that Dr. Burko Fuhrt and Dr. Piotr Blass are from Florida Atlantic University. Sure enough, doing a search of the university website turns up a few computer science classes... but that's interesting... All of Fuhrt's classes for Fall 2001 were cancelled, and they don't seem to be teaching anything in Spring 2002... Blass is an Instructor, apparently not a tenured professor, while Furht is a professor. They don't seem to have home pages so it's hard to know much about them.
Dr. Steve Smale of Berkely, on the other hand, looks like the real thing - a serious mathemetician. Someone should contact him and find out if he knows he's on the Zeosync org chart, and if so, what he thinks of their web site... I'd hate to see a genuine researcher inadvertently associated with something phony.
The Zeosync website claims that John Post of the University of Arkansas is on the Zeosync team, but a search of that University's directory turns up no hits for that name, and he doesn't have a home page there.
You are correct, they should look up the Pigeonhole principle.
The "Layman Process" explaination of their technology is worth a read just for the amusement value. They claim that it is not a compression technology, but that it works by "sending more data across less bandwidth while saving time", and that it "stores massive amounts of data compared to standard binary compression". Well, that sounds like compression to me. You might think that maybe they're referring to a different encoding method, but no, they also say that the data is able to "move rapidly on a fixed set of binary carriers through existing digital transmission devices".
So: They take binary data, do something magical to it, and then it can go across a digital, binary network, faster than any standard binary compression. OK, so what kind of magic is this?
Moving on to the "Technical Process", they have some astonishingly blatant smokescreen to their impossible claims.
First of all, they talk about the "solution to the Pidgeonhole Principle". Well, that's not something you solve, it just is. That's like saying you've learned to fly by "solving" gravity.
And then they "define" the pidgeonhole principle:
Given a number of pidgeons within a sealed room that has a single hole, and which allows only one pigeon at a time to escape the room, how many unique markers are required to individuall mark all of the pigeons as each escapes, one pigeon at a time? After some time a person will reasonably conclude that: "One unique marker is required for each pigeon that flies throught the hole, if there are one hundred pigeons in the group then the answer is one hundred markers".
Well, that's not what the pigeonhole principle is. The pigeonhole principle is simply: If you have more pigeons than holes, then there must be more than one pidgeon in at least one hole. Conversely, if you have less pidgeons than holes, then there must be at least one hole with no pidgeon.
Getting this basic theory wrong proves that they are either hopelessly ignorant or total frauds.
Furthermore, the reason the pidgeonhole principle disproves ridiculous compression claims is there are exponentially more long bitstrings than short bitstrings. So if you claim to be able to represent every long bitstring (the pidgeons) as a short bitstring, (the holes) then there must be at least two long bitstreams represented by one short bitstream. (Two pidgeons in one hole). But that means you can't tell which long bitstream was represented by the short bitstream, and you don't have a real compression algorithm - at least not a lossless one.
And then, Zeosync's alleged "technical explanation" veers off into the most amazing bullsh*t I've read in a long time:
In higher, multi-dimensional projective theory, it is possible to create string nodes that describe significant components of simultaneously identically yet different mathematical entities."
But wait! Reading further, I see that they use the word "lossy". The surrounding context simply doesn't make sense, but if you're talking about lossy compression, the pidgeonhole principle is irrelevant, because it's OK to not know exactly which long bitstream a shorter bitstream encodes - that's the whole point of lossy compression - you loose some detail. But then why discuss the pidgeonhole principle at all?
I hope someone sues these hucksters into a smoking crater. I hate it when people lie about fundamental mathematics.
I agree.. except that not even games require the fastest system available anymore.
I've finished Return to Castle Wolfenstein on my 800 Mhz Pentium III with GeForce2 card - it ran without any problem at 1024x768, I even turned up the detail level on the graphics.
And that's a three-year old motherboard with just PC100 SDRAM, and a 100Mhz system bus. It's still just an ordinary hard drive, no UltraDMA, the video card and CPU are both a year old... heck, I couldn't sell the whole thing second hand for more than a couple hundred bucks.
And it runs Wolfenstein great at 1024x768!
I don't plan to upgrade until the next ID game engine comes out. I don't know what Intel and AMD are going to do in a year when everyone's upgraded... where is the software that needs more power than last years system?
Unless you're doing video editing, software development, or something else unusual, there hasn't been any need to update for over a year now.
Yeah, I've heard that it's about three bits per word on average, based on the contextual redundancy of english. If you took an average paragraph of english and replaced 20% of the words with "###" most english readers would still be able to read it, guessing the missing words with a high degree of accuracy.
Taking an example from your post: "that is the text of Moby XXXX, or was XXXX by somebody who"
Even with less than a sentence of context, it's really easy to guess the missing word after "Moby" is "Dick". And the second word has to be a verb referring to text, so the likely choices are "read, written, studied". Unlikely choices are maybe "translated, remembered, typed...". I think it would be hard to come up with more than a dozen plausible words there. And three bits lets you pick exactly from a list of eight.
Using a variable length encoding (like Huffman encoding) where the most likely choices are encoded with the shortest bit sequences, the most likely choices could be encoded in two or three bits most of the time, with the unusual choices taking the longer bit strings...
It would be kind of fun to write a word-by-word compression algorithm for text along these lines...
Getting back on topic, this also makes it easy to see why compression algorithms are so useless when given nearly random input - there's no context, no redundancy, no "likely choice".
It seems like some company pops up every five years or so, claiming some astonishing compression breakthrough... They try to get investment money to "commercialize" or "fine-tune" it, and then they just disappear. It's a pity, stuff like that makes investors wary of getting involved in real research.
.
Well, you are correct it's a hoax, but they didn't actually say "random". It was something like "essentially random" and they probably left a few other weasel words in there.
It is easy to prove that it is impossible for any lossless compression scheme to compress every file, or to compress random data more than half the time. This is why their claims to get around entropy and supercede Shannon's work are so ridiculous.
The informal version of the proof is something like:
Digital files are essentially numbers.
Bigger files have more bits.
There are 2^n possible files n bits long.
Compression is essentially using shorter numbers to index longer numbers.
There are many more longer numbers than shorter numbers.
So, no matter what indexing scheme you choose, there won't be enough short numbers to go around.
So, you can only index some of the long numbers.
So, you can only compress some files.
And that's it... For any encryption method, there are files it can't compress. Period. In fact, for any encrytion method, there are so many more big numbers than small numbers that you won't be able to compress most files.
The reason you don't notice this, and gzip and Zip and the rest actually work really well in practice is that humans tend to be interested in a very tiny fraction of all the possible numbers/files out there.
We don't usually have large random files. Our files tend to have lots of structure, lots of repetition... this is "entropy", and Shannon's proofs about information entropy are very deep and very brilliant.
Personally, I think it's more likely that someone will find a way to bypass Einstein's theory of relativity than to bypass Shannon's theory of information entropy.
Why do we need to *copy* anything anyway? Or at least, why isn't there a project to bring a new type of UI that is not WIMP, not Mac, not Win-like in any regards, that would be unique and well suited for maximum efficiency for power users?
Dude, we already have it. It's been done for years, it's more powerful than you can imagine.
Here's a screenshot:
[thoffman@arnor thoffman]$ ll
total 40
drwx------ 10 thoffman thoffman 416 Dec 24 08:40 Art/
drwx------ 2 thoffman thoffman 120 Dec 28 14:14 bin/
-rw-r--r-- 1 thoffman thoffman 18096 Jan 6 14:54 blackbox-menu
drwxr-xr-x 3 thoffman thoffman 72 Dec 21 15:26 depot/
drwx------ 3 thoffman thoffman 624 Dec 23 13:14 Desktop/
drwxr-xr-x 5 thoffman thoffman 136 Oct 6 11:58 Development/
drwx------ 13 thoffman thoffman 424 Oct 11 13:04 Documents/
drwx------ 6 thoffman thoffman 424 Jan 9 09:19 evolution/
drwx------ 3 thoffman thoffman 72 Sep 27 20:37 GNUstep/
drwx------ 3 thoffman thoffman 80 Jan 3 10:35 mp3/
drwx------ 24 thoffman thoffman 816 Jan 3 10:34 music/
drwx------ 3 thoffman thoffman 192 Sep 16 14:11 Nautilus/
drwx------ 3 thoffman thoffman 72 Sep 3 14:23 nsmail/
-rw-r--r-- 1 thoffman thoffman 2109 Dec 20 17:16 patch.linus
drwx------ 15 thoffman thoffman 576 Oct 11 18:04 Pictures/
drwx------ 2 thoffman thoffman 288 Jan 6 14:59 playlists/
drwx------ 2 thoffman thoffman 104 Oct 20 15:20 src/
drwx------ 2 thoffman thoffman 80 Nov 16 20:47 tmp/
-rw-r--r-- 1 thoffman thoffman 1928 Oct 31 20:34 usb.txt
drwxr-xr-x 3 thoffman thoffman 168 Dec 14 15:26 Website/
[thoffman@arnor thoffman]$
Just kidding. Sort of.
Seriously, I do agree somewhat, but don't forget the training issue when you give people something totally new. When you look at something like KDE, it's both good and bad that it's similar to Windows. Good because it's easy for people to make the switch - Linux seems less scary and unfamiliar. Bad because a lot of things that suck about Windows-like GUI's suck in KDE/Gnome too.
But hey - at least on Linux you can do something about it. Or... there's always the command line...:-)
Many historians have praised the historical studies that Mr. Frayn undertook before writing the play. Still, in contrast to the complex Heisenberg of the play, the physicist in reality may have been easier to understand, Dr. Bernstein said.
Hmmm. So... historians are uncertain of Heisenberg's principles.
Back when usenet was where the action was, (before http), all the future politicians were in law school. And the law school students were way off on the other side of the campus, and thought the compsci/engr students were dorks.
And now, the only people that still post on Usenet are...
Personally, I gave up on Usenet in the early 90's, after following the Clipper Chip debate on comp.org.eff.talk all summer.
I don't think it would actually be found unconstitutional. After all, compare it to vehicle licenses:
You don't need a license to operate a bicycle. You do need a license to operate a car. You need a different, harder-to-get license to operate an 18-wheeler or a bus.
You don't need a license or background checks to buy, own, or use a slingshot or BB gun. But you generally need to get a background check to buy a real firearm, and you need to get permits for concealed carry.
There's lots of other examples.
So.... my prediction is you won't need a license for Windows or Apple, but you will need a license for any operating system where you can change the source code.
With all of these examples, it's all about how much power the user has, and how "dangerous" that power can make you to society / government / big business. You need a license for the "powerful" tool - be it a car, a gun, or maybe soon, a source-code-available operating system.
Therefore, as much as I hate it, I think a law requiring licensing for free software would be constitutional.
.
Oh, I totally agree that Microsoft would love to do that be the only "legal" OS. And it may happen, but it won't have anything to do with this patent.
You see, now that the DMCA is law, they don't have to bother with this anymore. They don't have to have real secure hardware, or secure software. They just have to implement some half-assed, weak, pathetic attempt at security, and then sue the hell out of anyone who points out how pathetic and weak it is... Much more convenient then doing real security!
Here's my prediction of what will really happen with all this crap.
The government will extend the DMCA in a direction similar to that proposed by the SSSCA, but since that was clearly insane and would have made Linux and BSD illegal, they will "compromise".
The "compromise" will be that people can either (a) run "Digital Rights Management Compliant" operating systems from Microsoft, Apple, and maybe a few others, or (b) Get a license to run a "non-Certified" operating system. Getting the license will put you in a big database. Your IP address will be tracked. The government will get away with this because they will point out that only a small percentage of computer users will need to get licensed, and most of those will actually be ISP's running Linux servers.
Besides the ISP's and other companies, the only individuals needing licenses will be a few thousand software developers, and a small number of computer "hobbyists".
Microsoft will love this because it will be a huge obstacle to Linux on the desktop, counterbalancing the cost of Microsoft. People will think:
Well, I can pay $200 for Windows and be up and running tonight, or I can fill out a big, scary form, send it in to the government with a $50 dollar fee, and be licensed to run Linux in two weeks..."
So what would happen?
A bunch of Linux users would leave the US. A lot of them would get licensed. A lot of them would give up Linux and go back to MS or Apple. And Microsoft would win.
No, a PCI card would not work. USB would be even worse - all you have to do is hook up a bunch of logic probes and you can read the "secret" numbers right off the wires. This isn't really difficult to do - although I suppose under the DMCA this would make logic analyzers "illegal circumvention devices". Hah.
The private key would have to be either burned into the CPU itself (just like the multiplier settings are) or possibly embedded into a separate chip sitting close to the CPU, and the whole thing covered with epoxy.
So far, none of the posters here have actually read the details of the patent. So everyone chill out for a second and read this critical little quote from the patent text:
The CPU manufacturer equips the CPU 140 with a pair of public and private keys 164 that is unique to the CPU [...] Other physical implementations may include storing the key on an external device to which the main CPU has privileged access (where the stored secrets are inaccessible to arbitrary application or operating systems code). The private key is never revealed and is used only for the specific purpose of signing stylized statements, such as when responding to challenges from a content provider, as is discussed below.
And, if you take the trouble to read the description of how the whole thing works, it comes down to the fact that the CPU can authenticate itself over the network at runtime by using this private key that ONLY the CPU can access.
Now, I don't know about you, but I haven't heard anything about Intel or AMD building public key / private key pairs into their CPUs. In fact, the whole Intel processor ID fiasco has probably scared them away from this area. Don't forget that this patent was filed in 1998, and was probably designed long before the PIII was released.
I think the most interesting thing about this is that it shows where Microsoft wanted to go in 1998 - they probably were working with Intel on the processor ID thing, and the next step would have been public / private keys to enable the design shown in this patent.
But it won't be happening anytime real soon. Unless maybe all those Pentium 4's out there actually have this as an unannounced feature. Unlikely, but possible - the P4 hyperthreading stuff was like that...
The cheapest backup media now is more hard drives. Really!
So I got 4 x 60 GB Maxtors (cheapest MB/$ ratio when I purchased), two Promise Ultra TX2-100 controllers, and set up a 180 GB software RAID-5 under Linux. I'm running ReiserFS on it, so I don't have to fsck 180 GB if I crash the system running an experimental kernel, and I keep the whole thing on a big UPS. Total cost for hardware was about $500.
I mirror data onto the RAID using rsync from my other computers, (or just drag-and-drop to the Samba server from the Win2K box). I think this is cheaper and more effective than a tape backup system for 180 GB of data.
The Linux software RAID gives me the reliability - I've inadvertently tested it when a power connector popped loose from one of the hard drives - I didn't even notice until I read the kernel log (for a different reason). After powerering down, I plugged the drive back in, restarted, did the "raidhotadd" command, and away I went again. No data loss, no hassle.
For stuff that I really, really want backed up beyond the reach of thieves and fire, I use CDR's and a safety deposit box. Luckily there is not too much stuff that falls into this category - source code and documents for projects I've worked on, my email, stuff like that - it still all fits on one CDR.
If you don't use Linux, I think Win2K can do software RAID too. Never checked though.
As a reader of the kernel development mailing list, I think I can answer this question, although I certainly don't claim to speak for Marcelo or anyone else...
You have to understand that most of the "development" that goes on in the stable series is actually bug fixes, and "new features" are mostly new drivers that don't affect the core code. New features that affect the core code are much less likely to be accepted, Marcelo will likely tell the person to send it to Linus for the 2.5 series.
The kernel maintainer, (Alan for 2.2, Marcelo for 2.4, Linus for 2.5) has to look at new code submissions with the question "What will it be like to maintain this code?"
In some ways, this is more important than if it actually works or not - if the design and code is very clean and straightforward, and has minimal dependencies, then future bug fixes and maintenance will be easy. If the code has a fragile design, or uses a coding style completely different from the rest of the kernel, then other developers will have a hard time reading it, fixing it, and updating it.
If code like that gets into the kernel, it would become an unmaintainable heap of crap. Even if each individual new feature "worked".
This is one of the things that Linus is pretty good at. (At least for the core code, Alan Cox apparently thinks Linus lets too much crappy code get into the drivers... but now I'm spreading stories from the mailing list without the benefit of the context, so take that with a grain of salt.)
Or read the mailing list yourself. But be warned, I had 582 messages in my inbox this morning, and that was just the kernel mailing traffic from the weekend.
I'd rather have Microsoft be forced to completely specify the Microsoft file formats for Office applications.
That way, Star Office, KOffice, Gnumeric, and the rest can get the import filters 100% correct.
That's really where Microsoft's monopoly is - many businesses would happily switch to Linux if they could be 100% sure that they could still reliably read and edit the thousands of documents they have already created.
You are correct, user education is critical, and that was the main point of the original "In defense of Microsoft..." message. And yes, Linux distributions tend to install too many services. Recently this has improved, though, and modern distributions are very easy to keep up to date with point-and-click tools.
But the point I am trying to make is: Linux will never have the kinds of email worms and viruses that plague the Microsoft world. It just ain't gonna happen.
Why? Several reasons. Basically, Linux has facilities for security that Windows doesn't, and this really limits the damage that a Linux "virus" can do.
Imagine two average home users. One is running Windows 98, the other is some modern distro of Linux, Red Hat for example. They get an "email virus". First of all, it's a lot easier to get infected from Outlook than from any Linux email client, and most viruses are for Windows... but ignore all that, suppose they both get an email with a malicious attachment.
The Windows 98 user double-clicks the file. They are infected. The mail to the Linux user arrives with instructions to save it to disk, chmod +x it, and run it. Suppose they are smart enough to do that, but ignorant enough to not wonder if it is a good idea. So, it's difficult, but they get infected too.
What is the worst that can happen to these two users?
The Windows 98 user can lose everything. Maybe even hardware damage - there are viruses that trash the CMOS. The Linux user can only lose their own document files. Yes, that's pretty bad. But the hardware and system software is safe.
Assume both users have backups of their document files on a CDR, and they have paid for support contracts from Microsoft / Red Hat respectively. So they call for help and explain what happened. Who will be up and running first, with the least amount of pain?
The Red Hat tech support can find out what happened, walk the user through logging in as root, cleaning up, and restoring documents from backup. This can really be as simple as typing in a half-dozen short commands. It might take half an hour, including the time to update a few RPMs for good measure.
Microsoft tech support will say: "Reinstall Windows. Reinstall all your drivers. Reinstall all your applications. Set up your system settings all over again. Don't click on email attachments. "
If the user is lucky (remember they have probably never installed Windows before, it is not really a simple process) they may be up and running after three or four stressful, painful hours.
[...](sketchy explanation of how to set up a throwaway test account deleted)[...]
I think it's about time the zealots pull their heads out of their asses before they go and flame someone on a topic they know nothing about.
Sorry, you lose. Here's why:
1. That doesn't work on Windows 95, 98, or ME. Those systems just don't have security. Period.
2. It doesn't work if you aren't using NTFS. A LOT of NT, 2K, and XP systems don't.
3. You don't have a short, simple description of how to "Set up NTFS ACLs properly". But I don't blame you - a short, simple explanation of that subject is impossible.
Compare that to Linux. The instructions I gave for setting up a throwaway test account are very simple, can be executed in seconds, and will work on any Linux distribution from the last five years at least.
That's impossible on Windows, and your post basically proved the point. Thanks!
What also pisses me off is the people that say "oh I run Linux so I'm fine"... well buddy, I could send you
#!/bin/sh
rm -rf/*
Gee, I just tried that, and all it did was print a million "Permission denied" messages. Oh, and it messed up my test account, but I fixed that with "su, deluser test, rm -rf/home/test, adduser test", and everything's back to normal.
Anyone else out there got some email viruses they want me to try out on my Linux box? They probably won't work either.
Warning to Linux non-experts: if you want to try this yourself, note that running rm -rf/* will delete any file owned by the person who runs the command.
Before you run anything off the network, you should switch your user (using the su command) to a "test" user that doesn't own any important files. You can set up a test user account by doing an "su root", "adduser test", and then "passwd test" to set the test user's password.
Carry on mocking Windows at your leisure... Or maybe the Microsoft apologists could write a little explanation of how to set up a safe testing account on Windows? Oh, that's right you can't, too bad about that.
Yeah, your point (d) is a problem. I was looking at portable MP3 players last weekend, and then I realized that all my favorite new stuff is oggs and I would have to re-rip it... (sigh) but actually I would probably need to re-rip it at a lower bitrate anyway to fit it on those little flash memory based players. All the hard drive based ones are still too big or expensive anyway.
Slashdot Mirror
Actually, the first thing that came to my mind was:
Finally! A real explanation for ZeoSync!
(For those who haven't been keeping up, ZeoSync is the company that claims to have broken / bypassed Shannon's laws of information entropy to create some sort of encoding or compression that can compress random data. Except that they don't call it compression - they call it "Information Crystals" or something equally stupid.)
I asked for and got ZeoSync's Investors Package, and it truly has some strange stuff in it... they are suing some previous employees, have some financial stuff that looks weird even to me (I know very little about corporate finance), and those computer scientists that are so prominently featured on their web site are not actually associated with the company - ZeoSync just paid them (an unspecified amount) for some sort of unspecified consulting. Basically meaningless.
They admit in the fine print that the alleged "technology" has never been demoed to anyone outside the company...
Really, I would not be surprised if ZeoSync was an elaborate ruse to teach gullible investors a lesson.
What would be really interesting is if they start putting it in the next generation P4, but leaving it disabled...
but then somebody figures out how to enable it without permission or support from Intel, and releases a patch for Linux and/or a VXD for Windows...
and all of a sudden Intel won't be able to sell Itanium no matter how hard they try.
What would they do then?
Hmmmm, how to do this with my Linux machines...
... ignoring the fact that most thieves would wipe linux off the computer and attempt to install a pirated copy of Windows ME - especially since they all come up at runlevel 3 with an unfriendly looking text-mode login prompt... Oh well...
All my machines already run SSHD all the time. So, I just need to find out where they are if they are stolen.
Solution: Write a little script that runs whenever any network interface comes up. The script would:
- check to see if an internet connection is available, if not, bail out.
- Otherwise determine the machine's DHCP-assigned IP address
- If PPP was used, find the phone numbers
- dumps the resulting info into a small text file and then uses "wget" to do an HTTP PUT and upload it to a CGI or PHP running on my website...
Whenever the machine connects to the internet, it will figure out where it is on the net and send that information to a log file that I can read....
Then I can secure shell in to the machine and wreak my terrible revenge, yea, even unto the third generation they shall speak of it in hushed voices with cautious glances over their shoulders...
Here's what I sent, feel free to copy anything that strikes a chord with you:
- - - - -
I'm writing to express my concern and disappointment with the proposed judgement in the Microsoft Antitrust case. It seems that the Department of Justice has won the case, only to concede defeat at the end.
As a software engineer, I've watched with disappointment for years as Microsoft has leveraged their desktop operating system dominance to crush one competitor after another.
I believe that few informed purchasers choose Microsoft products on their merits alone - rather, people buy Microsoft mainly because they need to exchange Microsoft Word and Excel documents.
One way to restore competition to the market would be to require Microsoft to completely specify and fully document their Microsoft Office file formats. No changes should be allowed without several months notice and complete documentation in advance.
This would allow competitors to create viable alternatives with the ability to interoperate with Microsoft, and would restore competition to an industry that badly needs it.
Please, consider a judgement which forces Microsoft to allow competitors to interoperate with their products.
Thank you for your consideration.
Torrey Hoffman
- - - - -
Ahhh... A Fields medal winner (!) you know, I thought I'd heard his name before. This makes me even more suspicious that he doesn't really know what's going on with Zeosync.
I suspect he's just really, really busy, and was invited to be an advisor to the company... they probably told him they were doing compression research, and he may not know about the claims they are making...
Betcha that as soon as he finds out, he'll get himself disassociated from them ASAP.
Digging around I found some more interesting stuff. First of all, if you get the PDF org chart, the information on that essentially contradicts the Flash propaganda - the technical staff seems to be divided into two teams, one "Advanced Compression Technologies Team" and one "Singular Bit Varience [sic] Encoder Team".
The org chart also mentions Wavelets, Fractals, and Sub-band compression... So much for the website that claims their technology isn't actually compression...
Maybe, just maybe, the scientists actually do have some sort of interesting compression technology, but the marketing / business people have spun and hyped it up, totally out of control and totally out of touch with reality. But I don't think so - marketing people alone wouldn't be able to come up with the pseudo-scientific drivel on that website.
Moving on, you see that Dr. Burko Fuhrt and Dr. Piotr Blass are from Florida Atlantic University. Sure enough, doing a search of the university website turns up a few computer science classes... but that's interesting... All of Fuhrt's classes for Fall 2001 were cancelled, and they don't seem to be teaching anything in Spring 2002... Blass is an Instructor, apparently not a tenured professor, while Furht is a professor. They don't seem to have home pages so it's hard to know much about them.
Dr. Steve Smale of Berkely, on the other hand, looks like the real thing - a serious mathemetician. Someone should contact him and find out if he knows he's on the Zeosync org chart, and if so, what he thinks of their web site... I'd hate to see a genuine researcher inadvertently associated with something phony.
The Zeosync website claims that John Post of the University of Arkansas is on the Zeosync team, but a search of that University's directory turns up no hits for that name, and he doesn't have a home page there.
Very strange indeed.
The "Layman Process" explaination of their technology is worth a read just for the amusement value. They claim that it is not a compression technology, but that it works by "sending more data across less bandwidth while saving time", and that it "stores massive amounts of data compared to standard binary compression". Well, that sounds like compression to me. You might think that maybe they're referring to a different encoding method, but no, they also say that the data is able to "move rapidly on a fixed set of binary carriers through existing digital transmission devices".
So: They take binary data, do something magical to it, and then it can go across a digital, binary network, faster than any standard binary compression. OK, so what kind of magic is this?
Moving on to the "Technical Process", they have some astonishingly blatant smokescreen to their impossible claims.
First of all, they talk about the "solution to the Pidgeonhole Principle". Well, that's not something you solve, it just is. That's like saying you've learned to fly by "solving" gravity.
And then they "define" the pidgeonhole principle: Well, that's not what the pigeonhole principle is. The pigeonhole principle is simply: If you have more pigeons than holes, then there must be more than one pidgeon in at least one hole. Conversely, if you have less pidgeons than holes, then there must be at least one hole with no pidgeon.
Getting this basic theory wrong proves that they are either hopelessly ignorant or total frauds.
Furthermore, the reason the pidgeonhole principle disproves ridiculous compression claims is there are exponentially more long bitstrings than short bitstrings. So if you claim to be able to represent every long bitstring (the pidgeons) as a short bitstring, (the holes) then there must be at least two long bitstreams represented by one short bitstream. (Two pidgeons in one hole). But that means you can't tell which long bitstream was represented by the short bitstream, and you don't have a real compression algorithm - at least not a lossless one.
And then, Zeosync's alleged "technical explanation" veers off into the most amazing bullsh*t I've read in a long time: Simultaneously identical yet different. Sure. Uh-huh. Giggle.
But wait! Reading further, I see that they use the word "lossy". The surrounding context simply doesn't make sense, but if you're talking about lossy compression, the pidgeonhole principle is irrelevant, because it's OK to not know exactly which long bitstream a shorter bitstream encodes - that's the whole point of lossy compression - you loose some detail. But then why discuss the pidgeonhole principle at all?
I hope someone sues these hucksters into a smoking crater. I hate it when people lie about fundamental mathematics.
I agree.. except that not even games require the fastest system available anymore.
I've finished Return to Castle Wolfenstein on my 800 Mhz Pentium III with GeForce2 card - it ran without any problem at 1024x768, I even turned up the detail level on the graphics.
And that's a three-year old motherboard with just PC100 SDRAM, and a 100Mhz system bus. It's still just an ordinary hard drive, no UltraDMA, the video card and CPU are both a year old... heck, I couldn't sell the whole thing second hand for more than a couple hundred bucks.
And it runs Wolfenstein great at 1024x768!
I don't plan to upgrade until the next ID game engine comes out. I don't know what Intel and AMD are going to do in a year when everyone's upgraded... where is the software that needs more power than last years system?
Unless you're doing video editing, software development, or something else unusual, there hasn't been any need to update for over a year now.
Yeah, I've heard that it's about three bits per word on average, based on the contextual redundancy of english. If you took an average paragraph of english and replaced 20% of the words with "###" most english readers would still be able to read it, guessing the missing words with a high degree of accuracy.
Taking an example from your post: "that is the text of Moby XXXX, or was XXXX by somebody who"
Even with less than a sentence of context, it's really easy to guess the missing word after "Moby" is "Dick". And the second word has to be a verb referring to text, so the likely choices are "read, written, studied". Unlikely choices are maybe "translated, remembered, typed...". I think it would be hard to come up with more than a dozen plausible words there. And three bits lets you pick exactly from a list of eight.
Using a variable length encoding (like Huffman encoding) where the most likely choices are encoded with the shortest bit sequences, the most likely choices could be encoded in two or three bits most of the time, with the unusual choices taking the longer bit strings...
It would be kind of fun to write a word-by-word compression algorithm for text along these lines...
Getting back on topic, this also makes it easy to see why compression algorithms are so useless when given nearly random input - there's no context, no redundancy, no "likely choice".
It seems like some company pops up every five years or so, claiming some astonishing compression breakthrough... They try to get investment money to "commercialize" or "fine-tune" it, and then they just disappear. It's a pity, stuff like that makes investors wary of getting involved in real research.
.
Well, you are correct it's a hoax, but they didn't actually say "random". It was something like "essentially random" and they probably left a few other weasel words in there.
It is easy to prove that it is impossible for any lossless compression scheme to compress every file, or to compress random data more than half the time. This is why their claims to get around entropy and supercede Shannon's work are so ridiculous.
The informal version of the proof is something like:
Digital files are essentially numbers.
Bigger files have more bits.
There are 2^n possible files n bits long.
Compression is essentially using shorter numbers to index longer numbers.
There are many more longer numbers than shorter numbers.
So, no matter what indexing scheme you choose, there won't be enough short numbers to go around.
So, you can only index some of the long numbers.
So, you can only compress some files.
And that's it... For any encryption method, there are files it can't compress. Period. In fact, for any encrytion method, there are so many more big numbers than small numbers that you won't be able to compress most files.
The reason you don't notice this, and gzip and Zip and the rest actually work really well in practice is that humans tend to be interested in a very tiny fraction of all the possible numbers/files out there.
We don't usually have large random files. Our files tend to have lots of structure, lots of repetition... this is "entropy", and Shannon's proofs about information entropy are very deep and very brilliant.
Personally, I think it's more likely that someone will find a way to bypass Einstein's theory of relativity than to bypass Shannon's theory of information entropy.
.
Here's a screenshot:
[thoffman@arnor thoffman]$ ll
total 40
drwx------ 10 thoffman thoffman 416 Dec 24 08:40 Art/
drwx------ 2 thoffman thoffman 120 Dec 28 14:14 bin/
-rw-r--r-- 1 thoffman thoffman 18096 Jan 6 14:54 blackbox-menu
drwxr-xr-x 3 thoffman thoffman 72 Dec 21 15:26 depot/
drwx------ 3 thoffman thoffman 624 Dec 23 13:14 Desktop/
drwxr-xr-x 5 thoffman thoffman 136 Oct 6 11:58 Development/
drwx------ 13 thoffman thoffman 424 Oct 11 13:04 Documents/
drwx------ 6 thoffman thoffman 424 Jan 9 09:19 evolution/
drwx------ 3 thoffman thoffman 72 Sep 27 20:37 GNUstep/
drwx------ 3 thoffman thoffman 80 Jan 3 10:35 mp3/
drwx------ 24 thoffman thoffman 816 Jan 3 10:34 music/
drwx------ 3 thoffman thoffman 192 Sep 16 14:11 Nautilus/
drwx------ 3 thoffman thoffman 72 Sep 3 14:23 nsmail/
-rw-r--r-- 1 thoffman thoffman 2109 Dec 20 17:16 patch.linus
drwx------ 15 thoffman thoffman 576 Oct 11 18:04 Pictures/
drwx------ 2 thoffman thoffman 288 Jan 6 14:59 playlists/
drwx------ 2 thoffman thoffman 104 Oct 20 15:20 src/
drwx------ 2 thoffman thoffman 80 Nov 16 20:47 tmp/
-rw-r--r-- 1 thoffman thoffman 1928 Oct 31 20:34 usb.txt
drwxr-xr-x 3 thoffman thoffman 168 Dec 14 15:26 Website/
[thoffman@arnor thoffman]$
Just kidding. Sort of.
Seriously, I do agree somewhat, but don't forget the training issue when you give people something totally new. When you look at something like KDE, it's both good and bad that it's similar to Windows. Good because it's easy for people to make the switch - Linux seems less scary and unfamiliar. Bad because a lot of things that suck about Windows-like GUI's suck in KDE/Gnome too.
But hey - at least on Linux you can do something about it. Or... there's always the command line...
heh heh heh.
My prediction: Never.
/engr students were dorks.
Back when usenet was where the action was, (before http), all the future politicians were in law school. And the law school students were way off on the other side of the campus, and thought the compsci
And now, the only people that still post on Usenet are...
Personally, I gave up on Usenet in the early 90's, after following the Clipper Chip debate on comp.org.eff.talk all summer.
I don't think it would actually be found unconstitutional. After all, compare it to vehicle licenses:
You don't need a license to operate a bicycle. You do need a license to operate a car. You need a different, harder-to-get license to operate an 18-wheeler or a bus.
You don't need a license or background checks to buy, own, or use a slingshot or BB gun. But you generally need to get a background check to buy a real firearm, and you need to get permits for concealed carry.
There's lots of other examples.
So.... my prediction is you won't need a license for Windows or Apple, but you will need a license for any operating system where you can change the source code.
With all of these examples, it's all about how much power the user has, and how "dangerous" that power can make you to society / government / big business. You need a license for the "powerful" tool - be it a car, a gun, or maybe soon, a source-code-available operating system.
Therefore, as much as I hate it, I think a law requiring licensing for free software would be constitutional.
.
You see, now that the DMCA is law, they don't have to bother with this anymore. They don't have to have real secure hardware, or secure software. They just have to implement some half-assed, weak, pathetic attempt at security, and then sue the hell out of anyone who points out how pathetic and weak it is... Much more convenient then doing real security!
Here's my prediction of what will really happen with all this crap.
The government will extend the DMCA in a direction similar to that proposed by the SSSCA, but since that was clearly insane and would have made Linux and BSD illegal, they will "compromise".
The "compromise" will be that people can either (a) run "Digital Rights Management Compliant" operating systems from Microsoft, Apple, and maybe a few others, or (b) Get a license to run a "non-Certified" operating system. Getting the license will put you in a big database. Your IP address will be tracked. The government will get away with this because they will point out that only a small percentage of computer users will need to get licensed, and most of those will actually be ISP's running Linux servers.
Besides the ISP's and other companies, the only individuals needing licenses will be a few thousand software developers, and a small number of computer "hobbyists".
Microsoft will love this because it will be a huge obstacle to Linux on the desktop, counterbalancing the cost of Microsoft. People will think: So what would happen?
A bunch of Linux users would leave the US. A lot of them would get licensed. A lot of them would give up Linux and go back to MS or Apple. And Microsoft would win.
That's my nightmare scenario, anyway.
.
No, a PCI card would not work. USB would be even worse - all you have to do is hook up a bunch of logic probes and you can read the "secret" numbers right off the wires. This isn't really difficult to do - although I suppose under the DMCA this would make logic analyzers "illegal circumvention devices". Hah.
The private key would have to be either burned into the CPU itself (just like the multiplier settings are) or possibly embedded into a separate chip sitting close to the CPU, and the whole thing covered with epoxy.
And, if you take the trouble to read the description of how the whole thing works, it comes down to the fact that the CPU can authenticate itself over the network at runtime by using this private key that ONLY the CPU can access.
Now, I don't know about you, but I haven't heard anything about Intel or AMD building public key / private key pairs into their CPUs. In fact, the whole Intel processor ID fiasco has probably scared them away from this area. Don't forget that this patent was filed in 1998, and was probably designed long before the PIII was released.
I think the most interesting thing about this is that it shows where Microsoft wanted to go in 1998 - they probably were working with Intel on the processor ID thing, and the next step would have been public / private keys to enable the design shown in this patent.
But it won't be happening anytime real soon. Unless maybe all those Pentium 4's out there actually have this as an unannounced feature. Unlikely, but possible - the P4 hyperthreading stuff was like that...
The cheapest backup media now is more hard drives. Really!
So I got 4 x 60 GB Maxtors (cheapest MB/$ ratio when I purchased), two Promise Ultra TX2-100 controllers, and set up a 180 GB software RAID-5 under Linux. I'm running ReiserFS on it, so I don't have to fsck 180 GB if I crash the system running an experimental kernel, and I keep the whole thing on a big UPS. Total cost for hardware was about $500.
I mirror data onto the RAID using rsync from my other computers, (or just drag-and-drop to the Samba server from the Win2K box). I think this is cheaper and more effective than a tape backup system for 180 GB of data.
The Linux software RAID gives me the reliability - I've inadvertently tested it when a power connector popped loose from one of the hard drives - I didn't even notice until I read the kernel log (for a different reason). After powerering down, I plugged the drive back in, restarted, did the "raidhotadd" command, and away I went again. No data loss, no hassle.
For stuff that I really, really want backed up beyond the reach of thieves and fire, I use CDR's and a safety deposit box. Luckily there is not too much stuff that falls into this category - source code and documents for projects I've worked on, my email, stuff like that - it still all fits on one CDR.
If you don't use Linux, I think Win2K can do software RAID too. Never checked though.
As a reader of the kernel development mailing list, I think I can answer this question, although I certainly don't claim to speak for Marcelo or anyone else...
You have to understand that most of the "development" that goes on in the stable series is actually bug fixes, and "new features" are mostly new drivers that don't affect the core code. New features that affect the core code are much less likely to be accepted, Marcelo will likely tell the person to send it to Linus for the 2.5 series.
The kernel maintainer, (Alan for 2.2, Marcelo for 2.4, Linus for 2.5) has to look at new code submissions with the question "What will it be like to maintain this code?"
In some ways, this is more important than if it actually works or not - if the design and code is very clean and straightforward, and has minimal dependencies, then future bug fixes and maintenance will be easy. If the code has a fragile design, or uses a coding style completely different from the rest of the kernel, then other developers will have a hard time reading it, fixing it, and updating it.
If code like that gets into the kernel, it would become an unmaintainable heap of crap. Even if each individual new feature "worked".
This is one of the things that Linus is pretty good at. (At least for the core code, Alan Cox apparently thinks Linus lets too much crappy code get into the drivers... but now I'm spreading stories from the mailing list without the benefit of the context, so take that with a grain of salt.)
Or read the mailing list yourself. But be warned, I had 582 messages in my inbox this morning, and that was just the kernel mailing traffic from the weekend.
I'd rather have Microsoft be forced to completely specify the Microsoft file formats for Office applications.
That way, Star Office, KOffice, Gnumeric, and the rest can get the import filters 100% correct.
That's really where Microsoft's monopoly is - many businesses would happily switch to Linux if they could be 100% sure that they could still reliably read and edit the thousands of documents they have already created.
You are correct, user education is critical, and that was the main point of the original "In defense of Microsoft..." message. And yes, Linux distributions tend to install too many services. Recently this has improved, though, and modern distributions are very easy to keep up to date with point-and-click tools.
But the point I am trying to make is: Linux will never have the kinds of email worms and viruses that plague the Microsoft world. It just ain't gonna happen.
Why? Several reasons. Basically, Linux has facilities for security that Windows doesn't, and this really limits the damage that a Linux "virus" can do.
Imagine two average home users. One is running Windows 98, the other is some modern distro of Linux, Red Hat for example. They get an "email virus". First of all, it's a lot easier to get infected from Outlook than from any Linux email client, and most viruses are for Windows... but ignore all that, suppose they both get an email with a malicious attachment.
The Windows 98 user double-clicks the file. They are infected. The mail to the Linux user arrives with instructions to save it to disk, chmod +x it, and run it. Suppose they are smart enough to do that, but ignorant enough to not wonder if it is a good idea. So, it's difficult, but they get infected too.
What is the worst that can happen to these two users?
The Windows 98 user can lose everything. Maybe even hardware damage - there are viruses that trash the CMOS. The Linux user can only lose their own document files. Yes, that's pretty bad. But the hardware and system software is safe.
Assume both users have backups of their document files on a CDR, and they have paid for support contracts from Microsoft / Red Hat respectively. So they call for help and explain what happened. Who will be up and running first, with the least amount of pain?
The Red Hat tech support can find out what happened, walk the user through logging in as root, cleaning up, and restoring documents from backup. This can really be as simple as typing in a half-dozen short commands. It might take half an hour, including the time to update a few RPMs for good measure.
Microsoft tech support will say: "Reinstall Windows. Reinstall all your drivers. Reinstall all your applications. Set up your system settings all over again. Don't click on email attachments. "
If the user is lucky (remember they have probably never installed Windows before, it is not really a simple process) they may be up and running after three or four stressful, painful hours.
1. That doesn't work on Windows 95, 98, or ME. Those systems just don't have security. Period.
2. It doesn't work if you aren't using NTFS. A LOT of NT, 2K, and XP systems don't.
3. You don't have a short, simple description of how to "Set up NTFS ACLs properly". But I don't blame you - a short, simple explanation of that subject is impossible.
Compare that to Linux. The instructions I gave for setting up a throwaway test account are very simple, can be executed in seconds, and will work on any Linux distribution from the last five years at least.
That's impossible on Windows, and your post basically proved the point. Thanks!
oops, correction: rm -rf /* will delete any file in a directory writable by the person who runs the command, with a few exceptions.
On most Linux systems, that's almost exactly the same thing as what I said, but I thought I'd correct myself before someone else does.
Anyone else out there got some email viruses they want me to try out on my Linux box? They probably won't work either.
Warning to Linux non-experts: if you want to try this yourself, note that running rm -rf
Before you run anything off the network, you should switch your user (using the su command) to a "test" user that doesn't own any important files. You can set up a test user account by doing an "su root", "adduser test", and then "passwd test" to set the test user's password.
Carry on mocking Windows at your leisure... Or maybe the Microsoft apologists could write a little explanation of how to set up a safe testing account on Windows? Oh, that's right you can't, too bad about that.
(snicker)
Yeah, your point (d) is a problem. I was looking at portable MP3 players last weekend, and then I realized that all my favorite new stuff is oggs and I would have to re-rip it... (sigh) but actually I would probably need to re-rip it at a lower bitrate anyway to fit it on those little flash memory based players. All the hard drive based ones are still too big or expensive anyway.
Torrey