| From: Stewart C. Russell via talk <talk@gtalug.org> | Easier said than done. Remember that the entire HS maths curriculum in the US | is effectively owned by TI calculators, and their lock-in allows them to sell | a 1980s-tech 'approved' calculator for ~$100.
Aren't "moats" great (Warren Buffet's term, I think)! | Compare this to the $1 scientific calculators you can get in dollar stores | (and supermarkets near "back to school" time). These are perfectly adequate, | but not "approved". A retired academic friend, ex CalTech, introduced me to | these super cheap calculators. He's done a whole suite of accuracy benchmarks | on a number of models, and they come out as well as the market leaders. Old geezer mode: I remember in the 1970's attending a couple of talks at U of T by Velvel Kahan about floating point accuracy. Always amazing and hair raising. One was about calculator accuracy. <https://en.wikipedia.org/wiki/William_Kahan>, a very notable numerical analyst. - a key architect of IEEE floating point - moved to UC Berkeley before I got to U of T. But he still had colleagues there. Notably T. E. Hull who I studied with. - showed that IBM SYSTEM/360 floating point had disastrous loss of accuracy due to base-16 quantization. At his suggestion, IBM fixed this (mostly) by adding a "guard digit". This entailed physically upgrading each machine already in the field! It also halved the performance of some machines because their FP ALUs weren't wide enough for the new requirement. - developed the program "paranoia" to test for bugs in IEEE FP implementations In that talk, he showed that all calculators made bozo errors, many unique to a calculator. As a consultant to Victor, he got their errors fixed. I don't remember whether HP and TI listened to him. This makes me very wary of random-brand calculators. (I've taken at least three Numerical Analysis course. My main take-awy is that getting FP right is really hard. Naive calculations are often wrong with no hints of problems.) ==== I loved calculators but I actually rarely need them. I've stopped buying them. But not before I bought too many. The first calculator I bought was a used Sinclair. You, Stewart, will know of those. Amazing but very cheaply built. It didn't last and I no longer have it. <http://www.vintagecalculators.com/html/sinclair_scientific.html> The next was a Commodore (later known for the PET, the Commodore-64 and the Amiga). Something like this: <https://www.flickr.com/photos/retrocomputers/30764477974/in/pool-26244609@N00> The weird mushy keys showed up again in the PET keyboard. The most recent (years ago): Dollarama sold nice Sharp D.A.L calculators for $2.99. While I was buying and not using cheap calculators, by cubicle-mate Henry Spencer was buying and open-carrying fancy HP ones. I got a couple of cast-offs and they were wonderful, but still not useful. I got a Sharp calculator out for tax time this year, but it just wasn't what I needed. I used bc(1) and LibeOffice Calc instead. Neither was quite what I wanted but they were good enough. Maybe a Mathematica notebook or something like that wuld be better (I've never used that). --- Post to this mailing list talk@gtalug.org Unsubscribe from this mailing list https://gtalug.org/mailman/listinfo/talk
