Son of Devil's Advocate, February 2004

Click here to make UNIX performance tuning easier!
advertisement

Previous SODA Index of SODA Columns Next SODA

Son of Devil's Advocate

Stan Kelly-Bootle


Picture of Stan Kelly-Bootle

Fair's Fair

Persi Diaconis, the Stanford University Statistician, seems to have proved that coin tossing is inherently biased. Following studies with Susan Holmes of Stanford and Richard Montgomery of UC Santa Cruz, Diaconis's verdict is "I don't care how vigorously you throw it, you can't toss it fairly." There is a small but measurable inclination for a coin to land on the same face it started from. Will this discovery affect real life as we know it?

Many vital decisions are made on the assumption that heads and tails have equally-likely outcomes. We tend to relate such decisions to such trivial pursuits as sports [ref 1]. But, there are earlier examples of vital import stretching back to Biblical times and earlier. The "casting of lots" is reported in Proverbs 16:13 to determine the "Will of God," and in Acts 1:26 to pick a disciple to suceed Judas. Devices for these divinations are referred to as Urim and Thummim (Sam. 14:41-2) although scholars are uncertain exactly how these worked. The Akkadian word Purim ("lot") survives in the eponymous Jewish Feast, celebrating the frustration of Haman's planned genocide (Esther 3:7). Tossing the Purim to pick the fatal day of Shoa is clearly a more dramatic application of probability than seeing which cricket team gets to bat (or bowl) first. The Purim may have been more like our die, the singular of dice, as thrown in crap-shoots [ref 2] the world over for fun or profit. A die, of course, has more than two outcomes (usually six), but the ideas of fairness and probability apply mutatis mutandis.

In the typical "toss-up" for a sports event, a "neutral" person makes the spin while one of the rivals makes the call. Which rival makes the call? Well, you can always toss for that, and so on ad nauseam. Before you ask, it has been known for a coin to avoid landing as head or tail. In 1965 a match between Liverpool FC and FC Koln (played in Amsterdam) had to be decided by a toss-up following a draw after extra-time. The coin landed on its edge in the muddy turf. After a pause for prayers, the coin was re-thrown giving Liverpool a richly deserved victory. With typical Scouse bravado, we called this win "The Massacre of Koln."

The Law of Large Numbers posits that, given a "fair" coin and a sequence of independent "fair" tosses, the ratio of heads-to- tails converges "in the long run" to the limit 0.5. But the circularities of this Law continues to puzzle us. What, indeed, is a "fair" coin, how does one toss it "fairly" each time (and independently of all previous spins) and for "how long" before the Law can be said to be reasonably validated? A sequence of, say, 500 heads might well indicate to the cautious, lay observer a lack of "fairness" in either coin, tossing method, or both. "Let me see that coin" or "Let me have a throw" would be natural reactions. You may rule out the obvious "double-headed" scam and rule out conjuring palming-tricks, but your 501st spin is, paradoxically, given a really fair "fair" coin and fair "fair" toss, equally likely to be yet another damned head as not. By definition, if the throws are truly independent, each toss knows nothing of the previous history. Indeed, the Law can be equally tested by tossing a large number of "identical" fair coins in parallel.

The Diaconis team studied the fairness of a coin toss in terms of whether

  1. the coin actually flips, spinning on a diametrical axis or
  2. merely wobbles, staying flat but appearing to spin.
Case (1) turns out to be "more fair," as you might guess, whereas case (2) can be exploited by cheats since the outcome is more-or-less predictable. Whence, of course, the aforementioned need for an independent coin thrower.

An anecdote from Richard XXXX underscores this situation, as well as reminding non-Brits that "tosser" is a mild pejorative:

"... Big Ron Atkinson's favourite tale concerns an international tournament in the 1990s. One team manager insisted that he did not have time for individual interviews with the two competing British TV companies, so it was agreed that BBC and ITV would pool resources, and spin a coin to see which reporter would put questions for joint use. The unfortunate Mr Newbon was absent when this agreement was reached, and when he discovered that Garth Crooks (BBC) was to undertake the interviewing on his behalf, went ballistic, saying "I'm the tosser for ITV."

Intel Science Talent Search

Last month, I pondered the "melting pot" effect on the top students reaching the finals of the 2004 Intel Science Talent Search. Of the forty finalists, only two had what you might call WASPish names. Anne Butzen emailed me wondering if the finalists' mothers' name might reveal a less biased "ethnic" distribution. I promise to investigate. Meanwhile, here are the winners.

Top prize: Herbert Mason Hedberg, aged 17, who receives a $100,000 scholarship. Lest you are thinking of your old high- school science projects, Hedberg's citation reads: "A novel approach to screening compounds that inhibit tolomerase, an enzyme that permits uncontrolled cell division in cancer."

Second prize: Boris Alexeev, also aged 17, for research on mathematical patterns and automata.

Third prize: Ryna Karnik, also aged 17, for microchip design and manufacturing advances.

Miles Kington Update

Serendipity strikes again. Having lauded Miles Kington, the English humorist, in my February SODA, I encountered a pleasant nod in his direction from Bill Bryson. Bill was discussing the ludic side of language and, surprise, Miles is a pioneer of the Holorime, a poetic play on homophonicity best explained by examples.

First, a poem, titled "A Lowlands Holiday Ends in Enjoyable Inactivity" offers two lines to be read aloud with due care:

"In Ayrshire hill areas, a cruise, eh, lass?"
"Inertia, hilarious accrues, he'las!"

Less meaningful but exploiting the melodious quirks of French:

"Par le bois du Djinn, ou s'entasse de l'effroi."
"Parle! Bois du gin, ou cent tasses de lait froid!"

Literally, to stretch the adverb:

"When going through Djinn's woods, surrounded by so much fear."
"Speak! Drink gin or a hundred cups of cold milk."

Picking Cherries

Cherry-picking is a noun-phrase with a newly emerging lease of semantic drift. From gathering fruit in Chekhov's Orchard to deflowering virgins it has now moved into the daily ritual of polemic altercation. Given any large, rich enough corpus such as the Bible, Voltaire's Collected Works, or the Hutton Report [ref 3], you can extract just those statements that support your particular hypothesis and agenda while ignoring the rest. Your opponents then accuse you of cherry-picking, and counter with their own irrefutable selections. I met an unexpected awareness of such bias in Lenin's Imperialism -- The Highest Stage of Capitalism (International Publishers, New York, 1939). Ah! I thought that would grab your attention!

Writing in 1917 Vladimir Ilyitch warns the comrades:

"In order to depict [an] objective position one must not take examples or isolated data (in view of the extreme complexity of social life, it is always quite easy to select any number of examples or separate data to prove any point one desires), but the whole of the data concerning the basis of economic life in all the beligerent countries and the whole world." (The italics are Lenin's.)

I suppose there's a Computer-Scientific angle to all this. Consider, for instance, the vast Natural Language documentation purporting to establish an ISO Computer Language Standard. It is by no means certain that its self-consistency can be formally established, as the history of the Algols confirms. The "theologians" can engage in cherry-picking to either praise or condemn the language while the compiler-writers do the best they can to "work around" any contradictions. The pragmatic conclusion is that the compiler itself determines the well-formed legal constructions.

Age Shall Not Wither

Picture of Maria, Stan and Andrew I am freshly paroled from my Petaluma nursing home, aka Chateau Despair, and my new Point Richmond environment, bubbling with younger friends, Andrew and Maria is confounding the dismal Third Law of Thermodynamics. "Entropy, be not proud, though some have openly called thee monotonic increasing, for thou art not so..." (John Donne, undone)

The occasional "Senior Moments," according to my daughter Michele Coxon, are more risibly known as CRAFT momemts (Can't Remember a F***in' Thing).

There's a comforting and timely entry for April 1st in "A Cherokee Feast of Days -- Daily Meditations" (Council Oak Books, Tulsa, Oklahoma, 1992) by Joyce Sequiche Hifler, also known in Cherokee as Di Ka No SGi-Di Go We Li SGi (She Who Writes Her Philosophy On Paper). In Cherokee, April is Tsi law'ni, the Flower Month (not at all cruel, pace T. S. Eliot_)

"In ages past, our old ones were the story-tellers. This was the way things were passed along to the generations that followed. For this reason the aged people made it a point to remember every detail so they could relate it at a later time. They were the word and picture carriers making history and spiritual values alive and important. We spoof their stories and make them feel foolish. The truth is that we are ignorant of what is precious and how to a da li he li tse di [ref 4]. Rigidity can creep in and set even the young mind if there are no soft memories, no laughter, no times too deep for tears. Age is grace -- a time too valuable to waste. We can get over being poor, but it takes longer to get over being ignorant."

C++ Self-Syntax Redux

An ancient quibble still turns up: should Bjarne Stroustrup have named his "improved" C as ++C (C pre-incremented) or C++ (C post-incremented)? In fact, it was Rick Mascitti who dubbed Bjarne's "C with Classes" as C++. The designation caught on so rapidly (as did the language) that the subsequent pedantic quibbling became pointless. I offer the following exegesis from Christopher Hecker at nyu.edu:

"If you are programming (figuratively) in C, C++ does indeed mean (figuratively) 'enhance the C language, then use the unenhanced language.' However, it could mean something completely different (figuratively) in C++. In fact, we aren't quite sure what it means (figuratively) unless we know the (figurative) type of C (not ANSI or K&R, mind you) because the postfix operator++ could be overloaded for that type. Of course, to a pre-2.1 compiler, ++C and C++ were identical if C was a user defined type, so either the original (pre-2.1) operator++ returned an enhanced C, just in case, or else Rick Mascitti assumed that implementors would be lazy in the post-2.1 days and define the following (assuming C is of type language (a big assumption, I know)):

// pre-2.1 code that we are reusing

language& language::operator++( void ) { BetterSelf(); // in this case, make a 'better C' return(*this); }

// post-2.1 addition for post-increment operator++

inline language& language::operator++( int Worthless ) { return(++(*this)); }

The syntax for defining a (post-2.1) post-increment operator++ can be a little confusing, hence my self documenting variable name. Borland C++ 3.0 did not give its usual warning:

Parameter 'Worthless' is never used in function = language::operator++(int)

so I assume that it knew that normal people would never use the explicit function call mechanism for the post-increment operator++. Hmm, there's a thought for a new name, C.operator++(0).

Well, it's still better than 'C with Classes.'"

If Christopher reads this, perhaps he could update his comments, and perhaps give us his views on C# (pronounced D flat?)


ref 1: I risk the ire of my fellow soccer fanatics. Bill Shankly, deified coach of Liverpool FC, is widely quoted: "Football isn't a matter of life or death. It's far more important than that."

ref 2: "Crap" in the British sense of "shit," is unconvincingly related by the Compact OED to the Dutch krappe ("chaff"). Seeking in vain for a decent etymology for "crap/craps" the game, I was intrigued by a verse dated 1770 in my 1985 DARE (Dictionary of American Regional Usage):

"My new crap's pitch'd, from which I hope to share "At least two thousand, all good notes, next year." Alas, this "crap" comes from a common Southern dialectal spelling of "crop.'

ref 3: Lord Hutton's much-debated 600-page report was commissioned by UK Prime Minister Tony Blair to clarify who said what to whom prior to the Iraqi invasion. Totalitarian states eschew such dirty-linen washing in public, to their eventual downfall. As Churchill famously declared, "Democracy is the worst possible system, apart from all the others."

ref 4: "Appreciate age/tradition." It really deserves the noble Sequoyah syllabary


This column 15 fleeting years ago

Devil's Advocate, UNIX Review, April 1989 -- Stan Kelly-Bootle

Open Column Seeks Support

Open Column repesents a long overdue attempt to end the jungle of conflicting columnar standards. The dream of inter-magazine reader-portability will remain a dream unless firm, courageous action is taken at once. The situation demands leadership, a rare commodity indeed until I was persuaded to shed my own petty self interests in order to champion the petty self interests of my dear readers.

Ah, sweet, precious, and dare I hope, regularly renewing reader -- I suffer and bleed in your ranks. I too receive those absolutely-no-obligation free charter issue reservation cards. I too fall for that "cancel at any time, owe nothing, and keep the ballpoint pen" trick. Like you, I am tempted into childish, Calvinistic-Hobbesian regression by sticking those little golden YES/NO/MAYBE decals in the spaces marked...O damn! Mummy, I've gone and put the NO sticker in the YES box. The bugger won't budge. Kick. Scream. Rage. Now I'll never get my free introductory MIPS, The Magazine of Intelligent Personal Systems. There, there, ma hinney, dinna fash yersel, Mum's got a brand new Dumb Terminal Coloring Beuk for you.

MIPS is the magazine that asks "Are you ready for the 80486?" at a time when most of us are wondering if Microsoft and IBM will ever exploit the hidden powers lurking inside the 286 (not to mention the 386). Note, in passing, how the overloading of acronyms continues (see Multi Use Mnemonics, this column Vol 7 #2, passim). To avoid the growing MIPS confusion, I propose that "millions of instructions per second" be henceforth known as MOIPS, which is probably how they say it in Brooklyn anyway. The MIPS in the RISC chip from Stanford stands for "Microprocessor without Interlocked Pipeline Stages," and surely that should have been called MWIPS from the start if only to preserve anagrammatic parity with Sun's SPARC (an explanation of this sentence will be sent to adults only upon receipt of a plain SAE).

In addition to the low standards in magazine launching technology (I particularly loathe the cut-out YES circle that never quite fits the target slots) we are witnessing a rapid decline in respect for even the most elementary notions of typographical uniformity and quality columnar contents. I urge the world-wide fraternity of Publishers, Editors, Writers, Typesetters, and Font Designers to join forces under the aegis of Open Column. The screams of a billion agonized readers cannot be ignored!

Open Column Charter -- Here's the Deal

Open Column Founder-Membership shall be available without restriction or fee, and regardless of age, sex or competence, to all my relatives (except for Uncle Jim, of course, unless he apologizes for...well, he knows what for). The Founder members shall constitute the Standards Committee, and I am pleased to announce that they have already unanimously voted that I should serve as paid-permanent Chairman. Auntie Ethel has graciously agreed to serve tea and take the Minutes for a modest honorarium.

Non-Founder membership shall be offerred at three basic levels: conforming, non-conforming, and read-only.

Conforming members shall pay only $1 million p.a. and shall receive regular, advanced warnings of changes to Open Column Standards. Such members shall submit sample columns upon request for conformance validation. The Committee's verdict is final. Conforming members may optionally contribute an additional $500,000 p.a. to achieve the status of Sustaining Conforming members. My brother Sean has designed a nice Sustainer's certificate suitable for framing.

Non-conforming members shall pay $5 million p.a. in return for which Open Column undertakes not to pester them with the Daily Standards Update Review, the Weekly Provisional Standards Digest, or the Monthly Consolidated Standards Sneak Preview. They are also excused attendence at our incredibly boring so-called Plenary Sessions held round the clock. Samples must be submitted weekly to allow the Commitee to verify non-conformance. These ruling are appealable upon payment of a suitable fee.

Read-only members, paying $25 p.a., shall have their names entered in a big book which shall be open to inspection at times to be announced. Additional fees may be paid to stem the flow of our Standards mailing.

Disclaimer

Open Column has no connection with or love for Open Look, Open Desktop, X Open, OSF, Open City, Open Season, Open Sesame, Open Warfare, Open Weekdays Only, Open Drain, Open Sore, Open Toed Shoes, Open Other End, or Open In the Name of the Law. However, we fully support the ANSI C function [b]int fopen(char *filename, char *rwmode)[b].

We are disturbed by reports that a new group calling itself Open Open has been formed to coordinate and standardize the activities of all Open organizations. Open Column has nothing to hide from Open Open, and we will certainly apply for either Dubious- Observer or Passive-Indifferent membership, but we would reject any attempts to interfere with our creative fiscal structure.

Our letters requesting affiliation with ISO, ANSI and ECZEMA have not yet received the courtesy of a reply, but Cousin Jack tells me that the mail has been something rotten, what with the holidays and all. He says, give them another week, then sod them, like.

RISC vs. CISC Commotion

I urge you to read Nick Tredennick's five page Viewpoint called "It's Not RISC vs. CISC -- It's New vs. Old" in the February issue of Microprocessor Report, Vol 3 Number 2 (550 California Avenue, Suite 320, Palo Alto, CA 94306 -- 415-494-3718). Editor Michael Slater is to be congratulated for allowing some hilarious skepticism to appear alongside the technical articles extolling the virtues of RISC.

I'm not sufficiently wafer-oriented to judge all the CISC vs. RISC issues, but Tredennick's piece has the startling, iconoclastic ring of truth that purges doubt: "The new architectures will repeat the history of the previous set of new architectures by adding instructions, address modes, features, and options -- in short, creating their own architectural baggage." Tredennick's track record as a key member of the Motorola M68000, IBM 370-on-a-chip, and Nexgen 386-compatibles design teams certainly gives him the clear right and authority, not to mention a strong motivation, to knock many of the RISC fads. Look out for a violent counter-attack from MIPSists, SPARCtans, and eighty-eight thousand of Nick's former colleagues.

When Tredennick calls RISC a fad, he is careful to add that "even fads can have their merits." They are, he asserts, part and parcel of all scientific and engineering fields, although "we find it hard to admit that engineering shares some rather unscientific characteristics with the fashion and entertainment industries."

Fad is one of those fascinating etym. unknown words, probably related to "fiddle-faddle". The OED defines fad as "a crotchety rule of action; a peculiar notion as to the right way of doing something...a hobby, 'craze'". A language has been defined as a "dialect with its own army and navy." There is a similar feeling that all innovations start as someone's pet fad. The fads that emerge correctly patented and fully armed with market-share soon shed their opprobrious origins to become respectable MBA case-studies.

The earliest OED citation for "fad" is 1834, which surely demands improvement. Last month, I mentioned the unending struggle of citationeers to push back source dates to and beyond the historical limits of human literacy. A remarkable example has just upset the natural calm of the lexicographical world: Claude Boisson's paper "Earlier quotations for Amerindian loanwords in English," (International Journal of Lexicography Vol 1 #4) lists seventy-six words with citations considerably younger than those found in the OED. The star in the list is totora, a South American plant that will be more familiar to you as the Typha domingensis. The OED's 1936 citation for totora gets rolled back all the way to 1604! These French scholars have eyes comme un rat des chiottes, n'est-ce pas?

Mae Hen Wlad Fy Nhadau (The Land of my Fathers)

The Welsh, too, clearly have good golygon (sight, eyes). Glyn Thomas Gowering, of Functions by Design, wrote to me in a delightful mix of Saesneg (English) and Cymraeg (Welsh) regarding my recipe for Welsh tea biscuits (Vol 7, #2), which he found ardderchog (excellent). (If you like to read my column aloud, the Welsh "dd" is pronounced like the "th" in "this," while "ch" resembles the Scottish guttural in "loch.") However, Glyn queried the phrase "metal love spoon" since traditional Welsh love spoons (Lliwiau Cariad -- no, I can't explain the "ll" sound in writing, so don't ask) are carved from wood by the non- paradigmatic swains who would a-courting go. Well, Glyn bach, to quote an old Welsh folksong: "Cosher Bailey only jokin'!" The original recipe indeed demands a metal spoon for some mundane cullinary reason, but I could not resist adding the cariad: my daughters Carol and Michele live in North Wales, two of my grandsons have won prizes for their Welsh poetry, and they all send me the genuine hand-carved wooden love spoons as Christmas presents. As the Fortran singers have it: "O bydded i'r hen iaith barhau!" ("O, may the ancient language live!")

I Can't Get No Instantation!

I have been delving into another "fad" but one that has all the signs of moving quickly and permanently into the mainstream of software orthopraxy. I refer to Object-Oriented Programming Systems, a subject that now sports several dedicated journals, conferences, and embattled standards' committees (in fact, all the signs of a mature computer linguistic discipline!). Of course, the origins go back to Simula in 1967, but fads have peculiar growth curves, and can lie dormant for long periods awaiting the right mix of climate, fertilizer, and manure- spreading methodologies. (Cosher Bailey only jokin'!)

My initial readings in C++ indicated that it had been designed to solve certain obscure problems in vehicular classification: did the class of Fords inherit engine objects and servicing methods from the class of Cars? Have you inherited a Ford lately? Did my particular instance of a Ford have an instance of an engine? How many objects can you fit in the back seat of a VW without overloading? I was about to give up when I discovered the true generality of object-orienteering: the car examples could be extended to aircraft without recompiling.

The resulting acronym, OOPS, provides yet another example of semantic clashery, this time with the restrained, archaic cry of horror one might emit when dropping one's cucumber sandwich on the vicar's carpet.

Gregory Haerr, President of Century Software, Salt Lake City, Utah, makers of the TERM communications package, has cleverly exploited this coincidence. He suggests in his adverts, tongue in cheek, that "...Oops!" is what you might say if a lengthy, long-distance file-transfer is inadvertently interrupted with loss of data (I presume the vicar must be close by!). The TERM software offers fault-free data transmission and an end to imprecations, mild or otherwise. Hence, at the recent Uniforum, Gregory was passing out lapel badges showing the word OOPS struck through with the familiar interdictory red diagonal, the universal subiconic modifier for prohibition and anathema. I mention this in case you encounter the NO-Oops badge and assume as I did initially that there was a campaign afoot to ban Smalltalk, Actor, C++, Eiffel, and the other OOPS now emerging.

Oops, my campaign to stamp out the paradigm (UR vol 6 #10 and Vol 7 #2) has been worse than a total failure. Discussing the matter with Bjarne Stroustrup at SD'89 inclines me to accept the fact that the word, like my love for Ella Fitzgerald, is "here to stay." It was still a shock, though, to encounter the paper "Multiparadigmatic Programming in Modcap" by Mark B. Wells (New Mexico State University) in the Journal of Object Oriented Programming (Jan/Feb 1989) which refers to Smalltalk and Prolog as uniparadigmatic languages. Although he doesn't say so, I reckon that Modcap is tetraparadigmatic (functional, imperative, object-oriented, and logical). The quadri- prefix would not be appropriate. (Why not? "Latin me that yer trinity scholard, out of eure Sanskrit into our eiryan!" -- Finnegans Wake, J. Joyce.)

All is forgiven, however, in view of Mark's rare and dangerous honesty: "Because of initial implementation annoyances and inefficiencies, and a lack of adequate user documentation, Modcap is not yet a commercially viable language." This kind of talk could ruin the compiler industry!

Luck of the Irish Footnote

Met a kindred soul during the UNIX Review Uniforum Regatta: publisher Tim O'Reilly who produces the Nutshell UNIX and X Windows Handbooks. Now dere's a lucky fella, I'm thinking, working for a company called O'Reilly & Associates. Most of us are forced to remember two names: our own and the firm's. I mean the probabilities are beyond estimation. Suppose, like, that your name is So-and-So, and you saunter into the ould Labour Exchange. The odds right there are a 100 to 1. "Any vacancies for a fine swaying man such as meself?" you ask with a hint of reluctance. Flip, flip. Tap, tap. "Well now," sez he at the grill, "So-and-So Limited, purveyors of such-and-such to the you- know-who are needing a willing hand so I'm told." This must be 10,000 to 1, and we still have the interview and all. I reckon the total odds including landing the job must be...[I t'ink you've made your point...Mr. U. Review, Editor.]

Quote of the Month

"This [Open Desktop] means a user could go from a UNIX operating system to a PC operating system without any training."

Tim Bajarin, Creative Strategies Inc. (quoted in SF Chronicle, 2/27/89)


Stan's bio:

Liverpool-born Stan Kelly-Bootle has been exposed to computing, on and off and vice-versa, since 1953 when, after graduating in Pure Mathematics at Cambridge University, he switched to impure post-grad work on the wondrous EDSAC I. After some trenching with IBM and Univac in the 1960s and 70s, Stan opted for self-employment as a consultant, writer, folk-song revivalist, after-dinner entertainer, and cunning linguist.

His monthly DA ("Devil's Advocate") column ran and ran in UNIX Review (aka Performance Computing) from 1984 until January 2000 (a date that will live in infamy) but lives on as SODA ("Son of DA") via www.sarcheck.com the homepage devoted to UNIX performance.

The latest of his umpteen books are "The Computer Contradictionary" (MIT Press) and "UNIX Complete" (Sybex). More on his biblio- and disco-graphy can be found on http://www.feniks.com/skb/ soon due for its millennial update.

Stan welcomes reader reaction: skb@atdial.net

To the SarCheck home page

The URL of this page may change in the future. Please bookmark the home page instead of this one.

Portions © copyright Stan Kelly-Bootle 2004.
Portions © copyright Aptitune Corporation 1997-2004, all rights reserved.
This column is sponsored by Aptitune Corporation, makers of SarCheck. The opinions of Stan Kelly-Bootle are his own and do not necessarily reflect the opinions of Aptitune Corporation. Other products and companies referred to herein are trademarks or registered trademarks of their respective companies or mark holders.