++ If you want perpetual support, you need to be prepared to spend huge sums of money to get it. You have to be prepared to pay a team of developers, testers, and support staff to support your outdated platform, and that does not come cheap, either in the initial purchase or with ongoing support contracts. My former employer offered perpetual support, but licensing fees ran in the millions of dollars and support contracts started at hundreds of thousands per year for even the smallest clients.

Depends on the university. At some schools, failing a course for cheating results in an F that stays on your transcript (and is included in your GPA) even if you repeat the course. But yeah, if that option is off the table, punishing people for cheating probably isn't worth the effort, because they're likely going to fail anyway.

The author must have had some amazingly bad professors. -I've had a grand total of one professor that didn't write her own lectures. Most have a set of lecture notes that they've developed and refined over the years. -Sure. -Some changes, sure. Sometimes you want to teach something in a different way and see how it changes students' understanding of the material, which means reusing questions. Sometimes a question is good and you want to keep it. Sometimes something went horribly wrong (like engineering majors not knowing differential volume in spherical coordinates), so you tweak the question a bit (and hope they understand Cartesian). -Rearranging the questions is really about the best you can do without running the risk of being unfair. Sometimes it turns out that what seem like simple variations on questions result in dramatically different student performance. -What's wrong with using material you wrote? Isn't that what you were just demanding? -Agreed. Lots of TAs are crap. They're frequently new to the country and have little experience speaking English. As for cheating, we try to do the best we can while avoiding false positives. That means that lots of people slip through and action is only taken in the most egregious, obvious cases, like students who turn in the exact right answers to a different test form, or students who turn in identical wrong solutions.

When you make things up, you look like an idiot. Microsoft has published their support lifecycles for all versions of Windows. Here's the one for vista. Mainstream support ends in 2012, while extended support ends in 2017. So that's a little over 5 years for mainstream support, and a little over 10 years for extended support. Compare that to WinMe (here) which was only supported for 3 years and 5.5 years.

Pretty damn well? The registry cleaned up the mess of .ini files thrown everywhere (not unlike the giant pile of files in /etc (or whatever other location a particular installer decides to put its config info in)), and the b-tree structure means keys leftover by old apps have negligible impact (despite the alleged "winrot" that so many drone on and on about).

The code to actually do something useful *is* frequently distributed fairly widely. Pretty much every major instrumentation company provides Labview support along with lots of example programs that you can use. That's the kind of stuff that requires a PhD to do. We're willing to pay NI a good bit of money for Labview because it works and it's easy. As experimentalists, we have neither the time nor the funding to sit around coding up drivers and writing our own programming languages. Labview makes it easy - easy to get data, easy to make multithreaded apps. Easy is good when you need to get papers published.

Exactly. It's a neat step forward, but surface plasmon resonance is nothing new at all. Lots of companies make SPR devices (Biorad, Texas Instruments, etc etc) and publications about using gratings date back at least twenty years.