I would argue that certs with practicals (CCIE, JNCIE, RHCE, etc) tend to hold their value much better than those that can simply be gotten by taking tests.

Largely because of the first-sale doctrine, which codifies property rights sanity: if you sell me something, it is now mine, not yours. I can do whatever I want with it. Use my spatula as a screwdriver? Use a thermos bottle for a hammer? Watch scenes in a movie out of order? It's none of your business. I bought it. It is now my property, and I'm free to do with it as I please.

(Averting pedantry: of course that doesn't involve violating copyright. Straw men will be ignored.)

Hell yeah, I'll admit that I am King of the Geeks. Talk nerdy to me.

OK, OK. I'll double-check with a calculator that's not "bc" before publishing. I've done enough physics work, though, to trust that 1) calculations showing explicit conversions are almost always correct, and 2) calculations that don't almost never are.

Like when Bell Labs developed C to write Unix? There's a long tradition of major companies coming up with new languages to scratch an itch. Thank God is hasn't died. How boring to live in a time when we'd decided that there was nothing left to innovate?

For nontrivial math, I don't always trust Google's interpretation of the question to be the same as mine. That page is a little short on details of what it's actually doing. On the other hand, WolframAlpha is really good about showing its work. I just always forget that it's there.

In either case, yeah, I like doing it the hard way. Or as I call it, "learning" or "practicing".

Moore's law describes the number of transistors in a package, not linearly their size. Doubling the number of transistors means that each one has to be 1/sqrt(2) as big as the old version, and that to-the-11th would be about 1/45th as large, not 1/2048th. That's still pretty dinky.

This is because America is without responsible civilians. Or non-civilians, for that matter.

We think it's great because, among other things, it has first-class functions and a very high code:boilerplate ratio. This lets us write very concise, readable, and maintainable code.

Eww, no. I've never seen good Python code that asserts types because it's not the idiom for you to care. For instance, suppose you write a function like:

def get_contents_of(obj): return obj.read()

In this case, obj might be a file, or a web request, or a string (via the StringIO interface). Who knows? Who cares? As long as obj.read returns something, it works. BTW, this is supremely nice for unit testing when you don't really want to hit an SQL server 1,000 times in a tight loop.

Now, you could write something like assert isinstance(obj, file) to guarantee that you're only dealing with file objects. Of course, that lays waste to the whole concept of duck typing and people will laugh at you for doing it. So dropping that bad idea, you could write assert hasattr(obj, 'read') to ensure that the object has the needed methods. But why? Python gives you that check for free when you try to call the method. Let it do the heavy lifting and concentrate on the parts of the problem you actually care about.

Yeah, that would be terrible. You almost never use them in Python like that, partially because Python tends to have a vastly shallower call stack than, say, Java (largely because you don't need 10 layers of abstraction between bits of code thanks to the duck typing we just talked about).

I think it boils down to you not knowing idiomatic Python. That's OK. I'm ignorant about lots of things, too. But I think you'd find that you enjoy it more if you stop trying to write C or Java in Python, because that almost never works out well.

The cleanup will consist of getting those bioconcentrating mushrooms to grow across the contaminated region, then harvesting and isolating them as nuclear waste.

Well, that sounds cost-effective. I'm sure the shareholders will be pleased. Unless they're not on the hook for this cost, and we taxpayers are.

Does not the fuel, inside the wreckage of the reactor building, continually produce new I-131, Cs-134, Cs-137, Sr-90, Radon, Xenon, and a host of others? The fuel is no longer critical, but fission of the enriched Uranium is ongoing. That's why the stuff is still physically and radiologically hot. Production of radionuclides will be ongoing for many thousands of years.

It's true that the nuclides will be more or less contained inside the reactor building, but some of it will seep out, because it's not hermetically sealed like the reactor containment was. If you seal it (ie. make an airtight sarcophagus) - then heat from decay will build-up. It requires air circulation to cool, and that's the whole point of the sarcophagus. It will still allow some byproducts to get out, and will affect all of those who are nearby; and this will continue until the fuel cools enough for workers to get in there, physically remove it, and put it into sealed storage.

It's the ongoing release and exposure of these materials that is the health-hazard. And it will remain so for thousands of years. This is precisely why safety systems for nuclear plants are so highly engineered. This was the scenario they never wanted to happen.

Chernobyl wasn't ever dangerous, since only a few dozen people were killed.

I agree, but remember that Python is interpreted in exactly the same way that Java is: both compile high level code to bytecode and run it on virtual machines. PyPy selectively uses LLVM to compile that bytecode into assembler for some enormous performance boosts, much as the Java JIT compiler does.

Whatevs. I co-built a web service on Python that handled 250,000 requests per second with a horizontally scaleable design. We could bump that up to 1,000,000 requests per second by deploying 4 times the servers (which isn't as easy as it sounds because most things don't scale out well like that). I left that company and went to another employer that handled "only" 80,000 requests per second, averaged over a month. If you can ditch the chattiness of HTTP, well, I've written single-threaded UDP servers in Python that could handle 200,000 requests per second per server. How fast do you want it to be?

Unless you're seeing extremely low numbers or your design requires vertical scaling because it was architected in 1965, choice of language isn't all that important. Ruby is slow, too, but Heroku manages to shovel the data pretty well.