Hmmm, I looked at your silver bullet link there and while I agree with some of the general ideas it has some serious problems.
First of all the discussion about Turing machines is very wrong. Turing machines are universial because they are basically as powerful as you can get in a computability sense. They can simulate multitape multihead machines just fine. Look at Siper's theory of computation.
Next, actor based computing is not new. It has been done in artifical intelligence as agents (see AIMA Norvig), the OS community as reactive computing (see tinyos) and it is becoming a hot topic for sensor networks. And in a sense it has been looked at via MPI and in some work on composeable software (oddly enough mainly in composeable network stacks). But no one knows how to do it and gets all sorts of wierd race conditions and odd resonant behavior. Baiscally it is all much HARDER to debug than simple sequential algorithms. So, it is interesting but without something more it certainly is NOT the silver bullet for program reliability.
Which brings us to the hardware comparision. CPU's are finite state they can be checked using symbolic execution or model checking. Both of these (currently) fail to scale up to handle even moderately complex approximations of software. (see Bebop, ACL2, or the work on predicate abstraction) So, yes finite state things are easy to verify, unbounded things are not.
Finally, it is not my area of specialty (and I don't like them) but the connection appraoch and COSA work look a lot like Petri Nets. You should really compare your work and work done in this field.
Anyway, I am doing a Ph.D. on software reliabiity and analysis. I am also interested in a message based (reactive) approach to some of the problems. I like the general idea the site presents (there are some serious issues with the details though). There is a lot of work out there that the author of the site has not looked at that would help them a lot.
On the grounds that there are not substantial non-infringing uses.
One can claim that there are good reasons to encrypt data when transfering it. One can claim that there are reasonable and non-infringing uses for file sharing networks. But when an application is designed to connect to a file sharing network that has a non-trivial amount of infringing material, search for files and then transfer them as secretly as possible it becomes a much harder claim to make.
The courts are not stupid. I don't like the current state of copyright law or the MPAA. But engaging in this hide and seek battle does not seem to be the way to change the laws.
Slashdot Mirror
Hmmm, I looked at your silver bullet link there and while I agree with some of the general ideas it has some serious problems.
First of all the discussion about Turing machines is very wrong. Turing machines are universial because they are basically as powerful as you can get in a computability sense. They can simulate multitape multihead machines just fine. Look at Siper's theory of computation.
Next, actor based computing is not new. It has been done in artifical intelligence as agents (see AIMA Norvig), the OS community as reactive computing (see tinyos) and it is becoming a hot topic for sensor networks. And in a sense it has been looked at via MPI and in some work on composeable software (oddly enough mainly in composeable network stacks). But no one knows how to do it and gets all sorts of wierd race conditions and odd resonant behavior. Baiscally it is all much HARDER to debug than simple sequential algorithms. So, it is interesting but without something more it certainly is NOT the silver bullet for program reliability.
Which brings us to the hardware comparision. CPU's are finite state they can be checked using symbolic execution or model checking. Both of these (currently) fail to scale up to handle even moderately complex approximations of software. (see Bebop, ACL2, or the work on predicate abstraction) So, yes finite state things are easy to verify, unbounded things are not.
Finally, it is not my area of specialty (and I don't like them) but the connection appraoch and COSA work look a lot like Petri Nets. You should really compare your work and work done in this field.
Anyway, I am doing a Ph.D. on software reliabiity and analysis. I am also interested in a message based (reactive) approach to some of the problems. I like the general idea the site presents (there are some serious issues with the details though). There is a lot of work out there that the author of the site has not looked at that would help them a lot.
Mark
On the grounds that there are not substantial non-infringing uses.
One can claim that there are good reasons to encrypt data when transfering it. One can claim that there are reasonable and non-infringing uses for file sharing networks. But when an application is designed to connect to a file sharing network that has a non-trivial amount of infringing material, search for files and then transfer them as secretly as possible it becomes a much harder claim to make.
The courts are not stupid. I don't like the current state of copyright law or the MPAA. But engaging in this hide and seek battle does not seem to be the way to change the laws.
Mark