Re: A Simple Notation

"paul c" <toledobythesea_at_oohay.ac> wrote in message news:%sfji.90316$NV3.47632_at_pd7urf2no...
> David Cressey wrote:
>> ...
>> Lisp is my favorite language that I'm not proficient in. But I
>> understand
>> the data structures of Lisp better than those of many other languages.
>> That's only because there is so little to understand and so much value in
>> understanding it!
>
> I have heard other people wax poetic about that aspect of Lisp and wonder
> what would have happened if Codd had been born ten years earlier or
> McCarthy ten years later. Years ago I worked at a place where the
> combinations to the doors were regularly changed. But the company wasn't
> worried about theft from their own premises, they were paranoid about even
> the most mundane of IBM manuals being brought in by patent troublemakers.
> Contrary to policy, we did have one or two manuals which concerned the IBM
> System/38, a machine that IBM deliberately crippled wrt speed and memory
> because it was thought to be a threat to their mainframe cash cows. It
> was later to become AS400 but before it did I had high but naive hopes
> that the far-sighted developers of the S/38 would put some kind of RA in
> the hardware.
>
> (In those days, RISC was a big future promise from the h/w mfrs, ie.,
> selling what they had rather than what the future demanded, the perpetual
> ironic and phoney symbiosis of the h/w and s/w so-called deep thinkers. I
> was appalled by RISC, it was the diametrically wrong direction to go,
> whereas I thought EMCCISC (Even more complicated CISC), cpu clocks be
> damned!) There's much to cricitize Intel for on that score, while they
> are right to put multi-media opcodes in their processors because part of
> the market would be helped by it, with all their money, I can't understand
> why they don't put a relation algebra into a few processors, patent it and
> force AMD to pay. Or AMD could do it first. Either way, I'd be pleased
> and wouldn't criticize them so much!)
>

I really don't understand why it should be so compelling to have relational algebra built into the processor. What difference does it make whether the algorithms that implement the relational operators are hard-wired into the processor or loaded into processor cache: the algorithms are still of at least linear complexity, and the input streams can be composed of thousands if not millions or even billions of tuples to process a single query.

> I never knew much about Britton-Lee and the other company, sorry, forget
> the name, that made "database machines", but I vaguely recall their boxes
> were of something less than universal usefulness.
>
> p
Received on Fri Jul 06 2007 - 07:45:10 CEST