Jwz reports:

For decades, hackers have strived to retire ‘Le Grand Sexpr’ — the platinum and iridium symbolic expression that for 126 years has defined the cons cell from a high-security vault outside Paris. Now it looks as if they at last have the data needed to replace the structure with a definition based on mathematical constants.

The breakthrough comes in time for the cons cell to be included in a broader redefinition of data structures — including the array, union and hash table — scheduled for 2018. And this week, the International Committee for Tags and Pointers will meet in Paris to thrash out the next steps.

“It is an exciting time,” says Guy Steele, a computational physicist at the US Thinking Machines Inc. “It is the culmination of intense, prolonged efforts worldwide.”

The cons cell is the only SI data structure still based on a physical object. Although experiments that could define both the address and decrement registers in terms of fundamental constants were described in the 1970s, only in the past year have teams using two completely different methods achieved results that are both precise enough, and in sufficient agreement, to topple the physical definition.

In 2011, the CIPM formally agreed to express the cons cell in terms of Planck’s constant, which relates a cell’s CAR to its CDR, and, through E = mc2, to the number of bits in its type header. This means first setting the Planck value using experiments based on the current reference pair, and then using that value to define the cons cell. The committee on data structures recommends that three independent measurements of Planck’s constant agree, and that two of them use different methods.

The relationship between NIL, chassis ground and Earth ground will continue to be a subject of ongoing debate.

https://www.jwz.org/blog/2015/10/cons-conflict-resolved-at-last/

Originally shared by Liz Quilty

https://en.wikipedia.org/wiki/Margaret_Hamilton_(scientist)

Originally shared by Ramin Honary

Cache is the new RAM

I’m not sure how I missed this awesome article from last year, presenting a brief history of the various trends and fads in big-data and enterprise computing since the year 2000. All of the engineering problems brought on by new technology are really just Hollywood-movie-reboots of problems that are as old as computing itself. The only difference now is that things happen faster and at larger scales.

But the mathematics of Turing machines have never changed since inception in the 1940s. One result of Turing’s thesis was that adding more Turing machines in parallel doesn’t make a more powerful machine that can compute stuff we couldn’t compute with just one. Adding more Turing machines just makes a larger, more complicated Turing machine that may produce your result in less time at the expense of more energy. So none of our problems are really that new no matter how much we scale-up or parallelize our workloads.

(Tagging John Hardy not a Turnbull fan on this, if he hasn’t read it already. You and your clique may like this article.)

http://blog.memsql.com/cache-is-the-new-ram/