Why, for the love of God, are we still designing pages with fixed widths? The dreaded "This page best viewed at 800x600" was bad practice in the 1990's; haven't we learned anything about we design in the last 20 years?

My web browser is not 1024 pixels across. I don't want my web browser to be 1024 pixels across. You see, CSS has this wonderful thing were you can say width="15%" and the browser will decide how big the column is based on the current size of the window. Change the window, and the columns grow or shrink to match. It's lovely, it's portable, it works on mobile devices, web browsers; big screens, small screens. But it doesn't do a flaming bit of good if the code monkeys doing your web page design insist on saying width="1024 px"!

I'm sorry, but having to scroll left-right for every single line of text is a royal pain in the ass! If your window is smaller than the designed size, the new layout is completely unusable, and quite frankly, looks like shit because I can't see half of what's on the page.

If MATLAB is optimized for 32-bit integer arithmetic, then maybe it's time to change that?

Have you ever even used Matlab? Integer data types are there for convenience. The natural data type is double, because any engineer or physicist worth their salary spends most of their life calculating real-valued variables.

The there is profit. The pharmaceutical firms are doing research, but then what happens when they try to pay for the research?

Not even big pharma anymore. Instead of basic research, big pharma works hand-in-hand with the venture folks.

The process is something like this:

1. Researcher A at university B makes a potentially marketable discovery in his/her laboratory.
2. University B patents the discovery and, in return, will receive a portion of any future licensing revenue (typically a 50-50 split with the researcher; this is all part of the researcher's contract).
3. Researcher A, together with some ex-graduate students, forms a start-up with venture funding and does the initial animal studies (or sometimes just licenses the rights to someone else's start-up).
4. If the initial research looks promising enough, then big pharma swoops in and buys the entire company, with profits all around.
5. Big pharma does the human studies (expensive, but low-risk, since they already know it works in animals), gets FDA approval, markets and sells the drug to make massive profits.

The only research big pharma does anymore, then, is the human clinical studies needed to satisfy the FDA — basically it's product development disguised as research.

Also, if you look at their balance sheets, you'll discover that Pharma spends significantly more on marketing than R&D. Viagra is $10/pill has as much to do with paying for the TV ads as it does with recouping any research costs.

-JS

Disposable plastics in medicine are critical in stopping infections.

Autoclaving for sterilizing medical tools is old tech. Disposable plastics are ubiquitous because that's how the device manufacturers make money (I used to do work related to medical devices). If you don't have either have a disposable bit or a per-unit cost of over $10M, your business plan will never be funded — the return on investment is too small for the venture folks to even bother reading your proposal.

-JS

Basically journals get academics to edit and review for free, to write for free, they force you to sign over copyright, and they charge you to access your own paper. [...] Most of the research is probably government and publicly funded anyways. Anyone see anything wrong with this??

No, I don't (and I say this having both published and reviewed academic articles myself).

The point most people here seem to not understand (or find inconvenient) is that most of these journals are published by non-profit organizations. The only significant exception is Elsevier, and I don't publish in their journals.

We researchers submit and review for free because otherwise the journals would stop publishing. Physical Review, for example, publishes something like 150000 pages of articles a year — and that costs money. Yes, they charge libraries a lot, but financially, they're luck to break even each year.

As for charging you for access to your own papers, the policy varies from journal to journal, but here's the APS policy from their author copyright FAQ:

Similary policies are spelled out for Wikipedia articles, re-use of figures in other articles, on-line reprints, and the like. Frankly, I've never heard of a copyright transfer getting in the way of getting work done...

-JS

The point of working in academia is to seek knowledge and share it with others. Copyright prevents or severely limits that. If knowledge isn't shared, we're all more ignorant because of it.

This is a silly argument. Copyright is granted when material is published. If it's published, then it has been made public for anyone to read. In what possible world is this not the definition of sharing knowledge?

-JS

We always try to keep in mind that correlation does not equal causation, but if that is so, what does the "55% of scientists are Democrats" statistic mean?

It's probably irrelevant.

It's been well documented for decades that people with advanced degrees (Masters and, in particular, PhD's) are, statistically speaking, more likely to vote Democrat than Republican. There's no reason I can see why scientists shouldn't mirror the general population in this respect.

-JS

I suppose that is the burden of being the oldest computer language in use today: lots of people evaluate it on the basis of what the language was way back when they learned it.

Aww c'mon, this is Slashdot. How many posters here actually ever learned Fortran in the first place?

On Slashdot, we evaluate it on the basis of what our CS professors and told us about Fortran. That and what we read on the Internet...

-JS

For the record, yes, I both know and have written significant programs in Fortran, but on the other hand I'm a physicsist and not a CS major (and when I learned Fortran, F77 was the most recent spec available!).

Also, good communication is the key to defusing people's annoyance.Also, good communication is the key to defusing people's annoyance.

Second that! If you want to be treated professionally, you have to begin by treating your co-workers as professionals too.

As a programmer [well, scientist-programmer], losing the connection to the outside world is bad enough — it gets in the way of doing my job — but it's even worse when you don't know what's going on and have no way to find out. If I know

1. That the problem has already been reported
2. That the problem is being worked on [or its relative place in the queue]
3. That the current best-guess time to resolution is X

then I can try to work around the disruption. Furthermore, this level of communication isn't too hard to implement: all of this could be done, say, by making the open trouble tickets readable on the in-house webpage.

Network problems are annoying. Being left in the dark scratching my ass wondering what's going on is worse.

-JS

One of the unsung virtues of LaTeX (and TeX) is the durability of archived documents.

I have documents going back to 1995 on my laptop (I lost everything older to a hard drive crash). Since they're just text, these documents are perfectly legible in any operating system. Moreover, the TeX API has been stable for decades — I can still turn every one of these documents into a PDF (or DVI, or PS, or ...) from the original TeX source.

How many of you can still open Word files you wrote using Windows 3.1?

-JS

So, you have millions of phones in 212, thousands in 979. The result: saved effort in dialing.

Nice idea, but you give the phone company too much credit. In the old days telephone switches still used physical relays (this is well before transistors were invented). This significantly limited the number of connections in progress each switch could handle. Since switches are expensive, you naturally wanted to pass on the call as fast as possible so you could free up the switch for the next caller. A number like '212' wasn't just easy to dial, it was fast — remember this is the era of pulse dialing as well, so a '9' took literally 9 times longer to dial than a '1'. Assigning fast numbers like '212' to New York saved money for the phone company because Ma Bell could buy fewer switches. Any benefit to the customer was purely accidental.

-JS

Benson's law crops in things like tax fraud where people are making up numbers instead of using actual costs (for example, increasing the reported purchase price of some shares of stock so as to decrease your capital gains). Humans usually tend to pick too few 1's and too many 9', and this is something a statistical analysis can pick up. It's not proof of fraud, of course, but it's enough to flag a return for further human inspection.

I would expect any stock market fraud that's based on fraudulent accounting (think Enron) might be flagged by a comparison against Benson's law as well.

-JS

Benford's "law" is not a law at all... any exponential distribution will exhibit this behavior.

A law, as the word is commonly used in math and physics, is a mathematical expression of a universal relationship. As you say, Benson's law is a property of any exponential distribution, so we agree it's universal. Why then can't we call it a law? Just because it's obvious after you understand it doesn't make it any less a law.

-JS

When happens with the primes are represented in base-9 or base-11?

Benson's law comes out of the distribution of logarithms. Change the base and you change the probability that the leading digit is a "1," but the law itself works for any base because logarithms are a well-behaved function.

-JS

#include <stdio.h>

int fib (int n)
{
if (n == 1 || n == 0) return n;
else return (fib (n - 1) + fib (n - 2));
}

...

If you're going to time it, at least write it properly!

#include <stdio.h>

int main(int argc, char *argv[])
{
int fib0 = 0, fib1 = 1, fib2 = -1, i;

printf("n=%i %i\n", 0, fib0);
printf("n=%i %i\n", 1, fib1);

for (i = 2; i < 35; i++) {
fib2 = fib0 + fib1;

printf("n=%i %i\n", i, fib2);

fib0 = fib1;
fib1 = fib2;
fib2 = -1; /* Paranoia */
}

return 0;
}

Going from your O(N^2) to my O(N) reduced the local runtime from 0.4 seconds to something smaller than the resolution of /bin/time (shows as 0.000 seconds).

-JS