In an expanding market a given item can have a declining market share and still be experiencing growth. Apple sells more iPhones than ever even though its share of the smartphone market has declined. The same logic can apply to programming languages as well as commodities.

Thanks for the response. Your experience pretty much mirrors mine, except I went for Ruby instead of Python.

A friend was editing a perl program when his cat walked across the keyboard. It took him a while to figure out which parts were written by the cat. That's when I decided to avoid perl.

It always struck me as one of the poster children of "write-only" languages, right up there with APL.

What can perl do that newer languages such as python and ruby can't do, and do more readably/maintainably?

I understand about path dependence and sunk costs, which is why we still have COBOL, I'm asking about language features that are unique to perl.

So another poster clicked their submit button a minute or two before I did. Your comment seems rather gratuitous given that we were both trying to clarify things for you.

Actually, the mathematically correct answer is a convex combination of the results for N=365 and N=366 with weightings 3/4 and 1/4, respectively. If you're going to pick a single number as representative of the answer, guess which one is closer? Besides which, all of this is premised on an assumption that birthdays are uniformly distributed throughout the year. That's close enough to get insight, but not actually true - there is a slight seasonality component in human birthrates.

You definitely lose the uniformity assumption by adding year - a larger proportion of the population are in their first couple of decades than in their 70's and 80's. As I mentioned earlier, non-uniformity actually increases the likelihood of duplicates so 484 should be an upper bound on the 99th percentile.

If you have a pool of N possible outcomes, and outcomes are pretty uniformly distributed across the pool, the probability of a duplicate existing is around 0.99 by the time you sample 3*sqrt(N). For the birthday problem, this means 3*sqrt(365) or about 57 people in the room means 99 times out of 100 there's at least one duplicate birthday. For a hash table with a million storage locations, you can be pretty sure you'll see key duplication by the time you have added a mere 3000 elements to the table. That's why collision resolution schemes are necessary even in relatively sparse hash tables.

If outcomes aren't uniformly distributed, they bunch more in certain locations and the probability of getting duplicates goes up.

Never mind, I had just hit "submit" when I figured it out. Just pull a "Weird Science" and wear them on your head.

The problem with +7 Knickers of Seduction is what happens when you remove them?

Thanks for the explanation, I am now enlightened.

How do you get it off your shoes?

Actually, no, I don't know where you're going. I think the issues I mentioned are important and real. I also think that there are noteworthy policy differences between the two parties on several of the issues you raised. You may disagree with me about how big the differences are or which way many of those things should go, but to claim that there are no differences is ludicrous.

I beg to differ, it matters a great deal. The two major candidates have a lot of differences about how they'd handle the economy. Even more importantly, whichever side you come down on for issues such as "money == speech", "corporations are people", or Roe v. Wade, the long-term balance of the Supreme Court will very likely be decided by who gets elected.

True, but when people throw out the word "correlation" that's almost always what they're talking about.

No, you have it exactly backwards. Causation usually implies correlation(*). But there are lots of correlations that are not in any way causally related, such as a) the decline in piracy and increase in global temperatures in the past two centuries; b) the rabbit population in Australia and the performance of the London stock exchange for the past century; or c) the monthly per capita consumption rate of ice cream and the monthly per capita rate of drowning deaths for seaside locations north of 40 degrees latitude. The first two are because both sets of observations have changed in consistent fashions over time. The third is because people seek/avoid both swimming and ice cream based on seasonal variations of temperature, i.e., there's what's called a "lurking covariate". In none of these cases does one of the sets of observations cause the other. You won't change global warming by encouraging piracy, swing the stock market by raising rabbits in Oz, or change drowning statistics by banning ice cream.

(*) Causation is not always associated with correlation, because correlation measures tendency towards a linear fit. If there's a non-linear relationship, you can have perfect causality and zero correlation. Example - Let X be uniformly distributed between -1 and 1, and let Y = X^2. If I tell you a particular X value you can predict Y perfectly, but if you work the math of correlation you'll find Corr(X,Y) = 0.

Nope. It's a great story, but it didn't happen.