Slashdot Mirror
Classic Computer Science Papers
Dean Chouinard writes "For inspiration I occasionally peruse the web pages of Dennis Ritchie, Brian Kernighan and Ken Thompson.
I recently found a pointer to the
following classic paper by Ken Thompson.
Other interesting papers are listed under the
classics directory
as well."
Hey, something I have ALWAYS wondered is how they made the first compiler? Since obviously they had nothing to compile it with. And compilers for new languages, are they made in different languages? Or what? And... if the c-compiler was written in c, how did Ken Thompson compile it the first time?
-Confused
My recitation instructor when I took 6.004 at MIT was Steve Ward. (This was 3 or so years ago.) Ward told us this same story. It seems that Ward and Ken Thompson went out to a bar one time, and Thompson was describing how to plant this trojan in a compiler. After a few more drinks, Thompson admitted to actually creating such a compiler, and compiling the login program with it. It worked. Ward kept feeding Thompson drinks, and pressing for more information. But Thompson would never quite say if he actually did anything with it, like if he merged it in on an install tape.
;)
I expect Thompson didn't do anything with it; I expect he was just holding back to keep an air of mystery around it all. But still... interesting...
(That's one of my best memories of that class. All us youngins hanging on every word of Ward's story about the "real" hacking done by our heros. Or at least mine.
Once you have any computer with the ability to run code, you can simply write code on that system to build the first compiler on another system, then copy that compiler over to the new system and go from there. Once you have a working code environment on any system, you'd be crazy not to use it to develop the tools for the next.
- jon
Ganymede, a GPL'ed metadirectory for UNIX
Well, of course it does. I didn't literally mean
'compiler' in the sense of gcc, I meant all the bits that is required to get a new computer system able to reasonably support its own code development.
- jon
Ganymede, a GPL'ed metadirectory for UNIX
I'm a little beyond the 'hello world' stage of programming, but not much. Can anyone tell me how far along developers have/have not come in terms of detecting the type of 'trojan' code discussed in this paper?
This sounds like the kind of 'warning' that could be leveled at the open source process, and I'd like to be able to refute it as accurately as possible.
**>>BELCH
Supposedly that's one of the big tests of the viability of a language: whether you can write a compiler for the language in the language. But you've always gotta write the first compiler in some other language.
After taking my compiler design course, I've never looked at programming the same way. Same with assembly programming. IMNSHO, every programmer should be exposed to both of these topics.
I actually work with Denis, Ken, and Brian and they're still cranking out some really cool stuff. None of it is mainstream due to Lucent's (Bell Labs) marketing of it and the huge installed base of current poop (MS Win) and software (*NIX)..
True that GOTOs are still running amuck in their code sometimes. Some of their newer stuff, papers, and such should be public like the theories behind OSs like Plan9 and Inferno, two OSs based completely on a file stucture for everything.
If you check out http://plan9.bell-labs.com/cm/cs/ you can find them and some off their recent work. Check it out.