So, the plan is to immerse a huge number of optical detectors into the deep sea for an extended period of time. Talk to any biologist or oceanographer and they'll tell you what happens to things like that - they become completely encrusted with plant and animal material. It's called "bio-fouling" and it's one of the biggest problems with putting anything in the ocean (aside from extreme pressure). I just don't see how they'd keep a system like that a) operational and b) calibrated.

I run my Monte Carlo simulations of photon propogation through water on AWS. I'm a gradute student, so I'm pretty price sensitive, but with AWS I can "rent" a 8 core (roughly 20% faster than my i7 920 computer in the lab) with 7 GB of ram for ~$0.25/hour. That's the "spot price" so it could fluctuate, but it's still *way* cheaper than the $0.89/hr fixed price. I have a machine instance that has all my tools (MATLAB, Dropbox) and I just click a few buttons and BAM I've got a machine I can remote to and work.

You need to consider the cost of storage, as that actually costs me more than the price of the machine. My bill last month was $130 for 677.640 GB-Mo of storage (forgot to shut down some EBS locations) and 200 hrs of the High-CPU Extra Large instance. I also transferred OUT 130.669 GB of data.

The best response I've heard to someone saying, "why do you care if you don't have anything to hide?" Is to say in return:

When you and your wife have sex, do you close the curtains?

Now, occasionally you might get some pervs, but in general it illustrates the point that privacy is a good thing and we should be jealous about guarding it!

For those that didn't RTFA, this paragraph, on the small team approach, is golden:

It is this small-team approach that, of necessity, results in important capabilities being left out of the first release. The payoff, though, is that Steve ends up with a central core of perfectly-integrated functionality instead of a rambling labyrinth of disjointed “features.” This design framework is so well conceived that it can be built upon for years, even decades, without being stripped out and restarted. Compare that with the history of Windows, with false start after false start, resulting in their repeatedly beginning design anew.

It probably isn't. Equalizers or adaptive equalizers (for when you don't know your channel characteristics, which you do in this case 'cause you can measure the ear's response first) are straightforward systems. It's basically a digital filter that inverts the channel's frequency response. Pop open MATLAB and you can have one up and running in minutes. There's plenty of source code on the 'nets for generating a equalizer for a DSP or FPGA. The hardware/software isn't hard at all. That only leaves testing and regulation for attributing to the high cost.

For those that miss the (hilarious) reference. Here's the SNL clip on Hulu.

>generally trust Amazon more than I do the small fry sites they 'affiliate' with.

I think you're a bit misguided here. The "small fry sites" you're referring to are sites, like mine, that link to Amazon products in exchange for a cut from Amazon. It's huge marketing for Amazon, and a tidy revenue for me and others. But not now. I'm in NC and I got screwed. Amazon hasn't killed people *selling* products, they've just cut off people that are doing free advertising for them.

What about Dr. Horrible? I'm rather partial to him and his glorious voice. All hail the Evil League of Evil!

I think the poster, and myself included, found the 30 second moral message at the end to be very forced. Next time I watch the end, I think I'll stop it as the camera pans away from Adama and Rosalyn's grave. Oh great ... now I've gotta go grab some tissues *sob*