I'll second this. When I was young and poor I had my parents get me his Amiga C compiler as a christmas gift. I learned a lot from reading the code for his C library, editor (DME), and networking layer (dlink). And I learned a lot playing with these tools.
I spend most of my time with Linux now, but I was happy to learn a few years ago that he was still at it, hacking on BSD.
Tiger, county assessor, postgis, and map server
on
Open Maps?
·
· Score: 2, Informative
Postgis, an add-on to postgres is a nice way to store map data. It does R-tree indexing, can store polygons, lines, and points, and can do coordinate system converison.
Tiger works quite well for me. I read the docs and wrote a simple perl script that took a sorted list of the road segments and intermediate points file, and inserted polylines into postgis. Tiger is off in a few places, which can be seen by overlaying it with more accurate data.
Another good source of data is the county assesors office. e.g. Clark County, Nevada, which builds a lot of new roads, has data available for free download in ARC/Info shape file format. (There exist converters to Postgis.)
Search for GIS, shape files, county assessor (+ your county name), etc.
The minnesota map server is a nice way to build maps images from shape files or postgis databases.
And GRASS, available in Debian, is a more complex database system for manipulating GIS data. It handles import, export, transform, mapping image files, and so forth.
FWIW, the commentary on TTT says that the falling scene was inspired by some artwork. (I forget the author's name.)
I will admit that I like EP1 and EP2, for mindless entertainment, special effects, and backstory. I guess EP4 probably best captures what Lucas' stated goal: make a movie that captures the spirit of cheesy 50's sci fi shows. But, then, Flash Gordon does a better job of that.:)
You're obviously not a SysAdmin, or someone else who runs mail servers. Otherwise, you'd be cheering very loudly (and a lot less sarcastically) in response to this (as I am!)
On the other hand, I have a friend who was cut off when somebody complained to Cox claiming he was infected. The "infected" box in question was a Linux machine. (This was on a business - not residential class line.)
So, it's a nice idea until you get your own systems cut off because the cable company can't tell an infected box from an uninfected one.
Also, Tolkien created more than one language. Quenya being the most complete, followed by Sindarin, and a few bits and pieces in other languages. And he created an etymology for the languages based on a root language (in the same way that proto-indoeuropean is a common root for most indo-european languages).
Detailed information can be found in book four of Christopher Tolkien's "papers I found lying around my dad's house" series.
"Computer Science" is a very wide ranging discipline, including people who do software/hardware engineering (as you describe), people who do mathematics (type theory, process algebras,...), and people who do science (propose hypothesis, make software/hardware as experiments, measure results, analyse, repeat).
In each branch of comp sci, you'll find people who will tell you that the others "are not computer science", but IMHO they all are real and useful aspects of the discipline.
Recent versions of Python have some MOP functionality in addition to reflection. (I can't remember if CLOS does reflection or not). It has metaclasses, and, IIRC, there are some hooks into dispatch outside of the metaclass stuff.
I happen to like the syntax of Lisp, and it's a lot faster than Python, but Python definitely wins in the library department.
I think the only example I've personally seen of MOP in use is UncommonSQL (an object/relational layer done by cleaning extending the class declaration syntax to cover the SQL information), which I like, but haven't used much.
And remember how quickly RSA-129 (a 426-bit or so key) was factored by distributed.
You could break DH-128 on a PC if you had to,...
Note this isn't 128-digit DH, rath 128 bits. That works out to about
about 39 decimal digits. I can't find exact numbers for modern hardware, but in 1991 LaMacchia and Odlyzko published Computation of Discrete Logarithms in Prime Fields, describing how they broke a larger, 192-bit key using about 1200 machine hours of a 25MHz R3000 chip, 40 hours on an unspecified machine (probably the SGI), and a few hours on a VAX. They conclude (over a decade ago) that using less that 200 bits is "very insecure".
The author states that the implementation could be improved upon, the 128-bit problem is smaller, and modern machines are much faster, so I was guessing a couple of days on modern desktop hardware.
Trillian does appear to change the prime, so the computation could only be used for one intercepted conversation.
Yeah, I got the 128-bit blowfish part from their web page. This would be fairly secure encyption, but their protocol weakens it by using 128-bit diffie-hellman (DH). Currently, the discrete log problem is about the same complexity as the factoring problem, for which conservative people recommend 1024 bits or better. (And never less than 512 bits.)
I determined that they used 128-bit DH from packet dumps. The DH negotiation is done in hex characters in the first few messages between the users. (Later, the actual encrypted data is sent in binary form.) It's also interesting to note that they appear to use the openssl code.
I believe this could be broken in about a day or two on modern machines, but I don't have exact numbers. So you can decode monitored trillian traffic without too much effort.
A second issue with their security is that the DH exchange is susceptible to man-in-the-middle attacks. So AOL wouldn't have to break anything, just intercept the communications. Without a certificate, you can't tell if you're exchanging keys with your AIM Buddy or AOL.
Finally, as Bruce Schneier frequently points out, it is unwise to use any security protocol that hasn't been publicly disclosed.
When I last checked Trillian negotiated its 128-bit blowfish encryption key via 128-bit DH key exchange, which is not very secure.
(It's about as secure as using a 128-bit RSA key.)
Mutters something about Z's place being in/dev/null
At the very least there is the "generate thesis topics" use.:)
valgrind just wraps the executable (you use it like strace). The way it works is quite impressive - it's an x86->x86 dynamic translater, which instruments the code.
It also has "skins" that allow you to plug in different checks. In addition to the memcheck, there is a cachegrind skin which does a cache miss profile of your application (marks up code with # of each type of cache miss). And you can make your own. You write a function that takes a basic block of risc code and returns your augmented version.
The downside is that it's x86 Linux only. And porting to another processor would be major work.
Java has JIT (including method inlining and other optimizations), which seems to give a performance advantage.
Java has garbage collection, and more recently, decent garbage collection. This is important for long-running and/or large applications. (When I last checked, Python didn't have garbage collection. It used reference counting, which usually is slower, and leaks data.)
I'm not sure if Python has any object/relational layers. (There are a few choices on Java.) It should be easily implemented though, given Python's dynamicism and reflection capabilities.
So, I'd use Java over Python for projects where these factors are important. Language-wise I like
Python a lot. I prefer Common Lisp for speed, features, and syntax, but Python has the libraries.
IMHO, dynamic languages seem to be the way to go for a lot of tasks. But staticly typed languages are usable when type inference is added, as in ML family languages. (c.f. OCaml for a fast, usable example.)
My class project was specifying an X11 window manager in Z.
Z has it's place (and Lotos, which the class also covered), but I don't think it applies to general purpose programming.
Assuming that you have the tech to check if your program matches the spec, and the spec is somewhat consistent, you still don't know if your spec matches what you want - so you're back to the original problem but in a slightly different language.
I have, however, been quite impressed with what the compiler group at stanford has done. The system is not 100% perfect (e.g. false positives), but it's detected a lot of real errors in real software (linux and bsd kernels) without too much noise.
As for run-time checking, I'm particularly impressed with valgrind. Whenever I suspect that a C/C++ program is breaking because I did something stupid, I use valgrind to find the fault.
That said, I think using safer programming languages, with GC and either dynamic or static typed (per your religion) would greatly improve the quality of software.
DirectTivo already does this.
on
Linus on DRM
·
· Score: 2, Informative
The kernel and initrd on the DirectTivo are signed, and the boot ROM will only load a signed kernel. (The initrd checks the root partition for modifications.)
They do this because you can get DirectTV without paying by tweaking the software. (They currently do not do this in their standalone units.)
I've done this too with XEmacs. Two authors working on a paper at the same time. The only issue is that you have to avoid commands that cause a prompt in the minibuffer (it will show up on both displays).
For PDA use, you should consider porting "dillo". It doesn't do frames or javascript, but it can render most other sites and the executable sizes is about 230k.
It amazes me that people still bitch about only one mouse button. If you want more buttons, go buy a friggin' 2- or 3-button mouse and plug it in! Mac OS X supports it natively.
Yes, I realize this, and I've done it in the past. But we're talking about a laptop here, which becomes less useful when you have to use an external mouse with it. (Now if I could replace the onboard one, I'd be happy.)
InkLink, $85 at buy.com, appears to be a pen that does this without special paper. (IBM also made a product with a tethered pen and clipboard, but I don't know if it exists anymore.)
Also, it seems to be common practice now to put these lookup/casts in static methods of a utility class, (XDoclet will write the class for you) both to keep them in one place and to keep the code clean.
Slashdot Mirror
I'll second this. When I was young and poor I had my parents get me his Amiga C compiler as a christmas gift. I learned a lot from reading the code for his C library, editor (DME), and networking layer (dlink). And I learned a lot playing with these tools.
I spend most of my time with Linux now, but I was happy to learn a few years ago that he was still at it, hacking on BSD.
Postgis, an add-on to postgres is a nice way to store map data. It does R-tree indexing, can store polygons, lines, and points, and can do coordinate system converison.
Tiger works quite well for me. I read the docs and wrote a simple perl script that took a sorted list of the road segments and intermediate points file, and inserted polylines into postgis. Tiger is off in a few places, which can be seen by overlaying it with more accurate data.
Another good source of data is the county assesors office. e.g. Clark County, Nevada, which builds a lot of new roads, has data available for free download in ARC/Info shape file format. (There exist converters to Postgis.)
Search for GIS, shape files, county assessor (+ your county name), etc.
The minnesota map server is a nice way to build maps images from shape files or postgis databases.
And GRASS, available in Debian, is a more complex database system for manipulating GIS data. It handles import, export, transform, mapping image files, and so forth.
I will admit that I like EP1 and EP2, for mindless entertainment, special effects, and backstory. I guess EP4 probably best captures what Lucas' stated goal: make a movie that captures the spirit of cheesy 50's sci fi shows. But, then, Flash Gordon does a better job of that. :)
On the other hand, I have a friend who was cut off when somebody complained to Cox claiming he was infected. The "infected" box in question was a Linux machine. (This was on a business - not residential class line.)
So, it's a nice idea until you get your own systems cut off because the cable company can't tell an infected box from an uninfected one.
Detailed information can be found in book four of Christopher Tolkien's "papers I found lying around my dad's house" series.
As of 2.5.73, the synaptics driver had issues on my dell laptop. (Neither it nor the APM BIOS reset it to absolute mode after a suspend/resume cycle.)
In each branch of comp sci, you'll find people who will tell you that the others "are not computer science", but IMHO they all are real and useful aspects of the discipline.
Recent versions of Python have some MOP functionality in addition to reflection. (I can't remember if CLOS does reflection or not). It has metaclasses, and, IIRC, there are some hooks into dispatch outside of the metaclass stuff.
I happen to like the syntax of Lisp, and it's a lot faster than Python, but Python definitely wins in the library department.
I think the only example I've personally seen of MOP in use is UncommonSQL (an object/relational layer done by cleaning extending the class declaration syntax to cover the SQL information), which I like, but haven't used much.
It was a seperate package in "experimental" but it's now been moved to "sid". I expect Takuo will upload 1.4 soon.
You could break DH-128 on a PC if you had to, ...
Note this isn't 128-digit DH, rath 128 bits. That works out to about about 39 decimal digits. I can't find exact numbers for modern hardware, but in 1991 LaMacchia and Odlyzko published Computation of Discrete Logarithms in Prime Fields , describing how they broke a larger, 192-bit key using about 1200 machine hours of a 25MHz R3000 chip, 40 hours on an unspecified machine (probably the SGI), and a few hours on a VAX. They conclude (over a decade ago) that using less that 200 bits is "very insecure".
The author states that the implementation could be improved upon, the 128-bit problem is smaller, and modern machines are much faster, so I was guessing a couple of days on modern desktop hardware.
Trillian does appear to change the prime, so the computation could only be used for one intercepted conversation.
I determined that they used 128-bit DH from packet dumps. The DH negotiation is done in hex characters in the first few messages between the users. (Later, the actual encrypted data is sent in binary form.) It's also interesting to note that they appear to use the openssl code.
I believe this could be broken in about a day or two on modern machines, but I don't have exact numbers. So you can decode monitored trillian traffic without too much effort.
A second issue with their security is that the DH exchange is susceptible to man-in-the-middle attacks. So AOL wouldn't have to break anything, just intercept the communications. Without a certificate, you can't tell if you're exchanging keys with your AIM Buddy or AOL.
Finally, as Bruce Schneier frequently points out, it is unwise to use any security protocol that hasn't been publicly disclosed.
When I last checked Trillian negotiated its 128-bit blowfish encryption key via 128-bit DH key exchange, which is not very secure. (It's about as secure as using a 128-bit RSA key.)
At the very least there is the "generate thesis topics" use. :)
valgrind just wraps the executable (you use it like strace). The way it works is quite impressive - it's an x86->x86 dynamic translater, which instruments the code.
It also has "skins" that allow you to plug in different checks. In addition to the memcheck, there is a cachegrind skin which does a cache miss profile of your application (marks up code with # of each type of cache miss). And you can make your own. You write a function that takes a basic block of risc code and returns your augmented version.
The downside is that it's x86 Linux only. And porting to another processor would be major work.
- Java has JIT (including method inlining and other optimizations), which seems to give a performance advantage.
- Java has garbage collection, and more recently, decent garbage collection. This is important for long-running and/or large applications. (When I last checked, Python didn't have garbage collection. It used reference counting, which usually is slower, and leaks data.)
- I'm not sure if Python has any object/relational layers. (There are a few choices on Java.) It should be easily implemented though, given Python's dynamicism and reflection capabilities.
So, I'd use Java over Python for projects where these factors are important. Language-wise I like Python a lot. I prefer Common Lisp for speed, features, and syntax, but Python has the libraries.IMHO, dynamic languages seem to be the way to go for a lot of tasks. But staticly typed languages are usable when type inference is added, as in ML family languages. (c.f. OCaml for a fast, usable example.)
My class project was specifying an X11 window manager in Z.
Z has it's place (and Lotos, which the class also covered), but I don't think it applies to general purpose programming.
Assuming that you have the tech to check if your program matches the spec, and the spec is somewhat consistent, you still don't know if your spec matches what you want - so you're back to the original problem but in a slightly different language.
I have, however, been quite impressed with what the compiler group at stanford has done. The system is not 100% perfect (e.g. false positives), but it's detected a lot of real errors in real software (linux and bsd kernels) without too much noise.
As for run-time checking, I'm particularly impressed with valgrind. Whenever I suspect that a C/C++ program is breaking because I did something stupid, I use valgrind to find the fault.
That said, I think using safer programming languages, with GC and either dynamic or static typed (per your religion) would greatly improve the quality of software.
The kernel and initrd on the DirectTivo are signed, and the boot ROM will only load a signed kernel. (The initrd checks the root partition for modifications.)
They do this because you can get DirectTV without paying by tweaking the software. (They currently do not do this in their standalone units.)
I've done this too with XEmacs. Two authors working on a paper at the same time. The only issue is that you have to avoid commands that cause a prompt in the minibuffer (it will show up on both displays).
Dog slow in 16bit mode and simply prints "Error" in 24bit mode.
My best guess is that I need NVidia's drivers, but that probably won't be an option anytime soon on my 2.5 kernel.
I think it's safe to say he smoked, albeit not cigarettes.
For PDA use, you should consider porting "dillo". It doesn't do frames or javascript, but it can render most other sites and the executable sizes is about 230k.
It amazes me that people still bitch about only one mouse button. If you want more buttons, go buy a friggin' 2- or 3-button mouse and plug it in! Mac OS X supports it natively.
Yes, I realize this, and I've done it in the past. But we're talking about a laptop here, which becomes less useful when you have to use an external mouse with it. (Now if I could replace the onboard one, I'd be happy.)
Mouse buttons.
It's a cute machine, but I need at least a 2 button mouse to comfortably use the software I need.
InkLink, $85 at buy.com, appears to be a pen that does this without special paper. (IBM also made a product with a tethered pen and clipboard, but I don't know if it exists anymore.)
2 &loc=101
http://www.buy.com/retail/product.asp?sku=1032022
I've always been impressed by clock reactions. This time of year, if you can get an orange/black one going for halloween it would be perfect:
p us_a/woodlandm/Demo1.html
http://www.mun.ca/educ/ed4361/virtual_academy/cam
will also break when using a non-RMI messaging layer (like CORBA). I've done the same myself in projects but to be fully compatible, you need:
Also, it seems to be common practice now to put these lookup/casts in static methods of a utility class, (XDoclet will write the class for you) both to keep them in one place and to keep the code clean.