http://longnow.org/essays/richard-feynman-connection-machine...
123 karma · joined June 29, 2008
http://longnow.org/essays/richard-feynman-connection-machine...
I think it'd be a hard sell to use ML in safety-critical systems due to the non-deterministic runtime of the GC and potential for memory exhaustion.
As a counter-example, C doesn't prevent formal methods - FWIW VxWorks puts out Arinc 653 and MILS products that are verified* using formal methods.
* - This is carries the following caveat, I believe the core OS is formally verified but the particular BSP/configuration needs be tested/certified by the customer or at the customer's expense.
I don't think a lawsuit was ever filed. At the time there was speculations as to whether it was considered a public performance and whether it was considered US jurisdiction for copyright.
Concrete Example: 7 years ago I couldn't (using available libraries) reliably parse in lisp a DB dump in XML (~8GB), select some records based on a set of calculations, and convert it to CSV. Most libraries insisted on needing the whole document in memory, some libraries leaked like a sieve and caused heap exhaustion, others couldn't handle UTF8. In the end, I could do it python (couple hours to develop and took 45-60 minutes to run), I could do it in java (several hours to develop and ran in ~8min), but I couldn't do it in LISP without writing a custom XML parser or delving into FFI. Maybe it was an abnormal case, maybe not.
I miss macros and the expressiveness; I miss the community; I don't miss the insanity.
The giant caveat is that the documentation online does not seem to indicate whether or not a transaction log is stored in the trusted hardware.
This gets you a number of benefits: separation of the front-end logic and the back-end logic, better scalability - there may be a bunch of workers on distributed among different machines, and security - the C++ programs aren't as worried about unvalidated input since their input comes from the front-end.
That being said - retail seems to be our generation's victim. In the last decade, we've replaced media consumables with digital equivalents. This alone has mostly killed music, book, and movie rental stores. It's not just retail stores but retail services - tax prep, insurance, and financial services. Some of this took place in the 90's, but it continued in the 00's and it will continue in the 10's.
http://www.scsh.net/docu/html/man.html
The story goes: while Olin was working on his doctorate he had the idea to create an anti-acknowledgements page within his thesis. While the typical acknowledgements page covers those who gave lover and support during the author's effort. This page would be dedicated to everyone he got his degree despite of their efforts/encouragements. He, eventually, was talked out of this. Later, this idea re-emerged while at MIT's AI lab and ended up in the SCSH manual. It's definitely worth a read if you want a laugh.
Brings a whole new level to the saying "Always code as if the person who ends up maintaining your code is a violent psychopath who knows where you live."
As for the overall mission mandate may directly or indirectly be related to prompt global strike, a kinetic kill system (either for satellites or for ground targets), or electronic warfare.
I imagine this release is more contentious since it undoubtedly contains sources, methods, and likely personal opinions of the diplomatic corps and the clandestine services operating under the guise of the diplomatic corp.
Most of the VM tools seems to miss an ability to branch a VM (i.e. take a snapshot and break it off as a seperate instance). This would be hugely useful for installer testing.
The OMAP-L138 Processor only supports usb 1.1 and usb 2.0. In general, a fair number of low power processors in the PDA/netbook realm only support usb 1.1. It may be possible to use an external usb 3.0 controller but that would dramatically complicate your board design.
Basically, I scraped the site, removed formatting/spacing/dead-words, stemmed (using a modified porter), constructed a matrix of word-frequencies per post. After which I did several various analytical techniques (statical, geometric, etc). The net result from using a blended method was able to identify several most aliases of board posters However, the issue google would no doubt have is sampling size. In a small enough sampling size quirks work as identifying characteristics, in a larger dataset you would no doubt see clusters of people who have similar backgrounds (e.g. education).