Management and their crony lawyers could have given up their entire salary and worked pro-bono all year, and it *still* wouldn't have been enough to bring the company out of the red. Employee salaries and pensions, however, are probably at *least* a billion dollars per year (if it's only a third of revenue, which I would guess is on the low end). So making cuts to salaries/pensions would actually do something.

Your article doesn't have total amounts, but let's be generous and say that management gave themselves and their crony lawyers an extra $10 million per year. Sure, it's an insult and a slap in the face, but it's not enough to really impact the bottom line significantly.

If $10 million in management excess is the reason the union employees voted the way they did, then they cut off their nose to spite their face.

You often criticize poor hardware designs (e.g. that ACPI is "a complete design disaster in every way") and complement good ones (AHCI). If you were given the opportunity to directly influence the design of important hardware standards in the future (let's say, if Intel gave you veto power), would you do it?

I recommend incorporation for the tax advantages, not for the lawsuit liability protection. From what I've read, and what my CPA has told me, a lawsuit can make your personal assets just as vulnerable in a single person corporation as they are in a sole prop. The difference is that it's *possible* to make contracts between your corporation and others, rather than you personally. But just because it's possible doesn't mean it's likely -- until you build up some significant business credit rating (e.g. a good D&B report) you're going to have to personally guarantee everything (e.g. lines of credit, business visa, etc.) anyway.

But mdadm *does* beat at least some of the enterprise $700-$1500 ones as well. My LSI MegaRAID SAS 9261-8i cost me about $900 (the battery alone was around $300) and it's slower than snot.

I was raking in 800 MB/s seq with mdadm on an empty 8-disk RAID-50 using a bunch of $30 "cheapy" SATA HBA, but when I switched the exact same drives to hardware raid, the most I could get was 250 MB/s (seq) on an empty array and 160 MB/s at 85% full. Not to mention the random read I/O of 1 MB/s (yes, one MB per second -- not a typo). This is after spending a few weeks optimizing things: stripe-aligned partitions, block-aligned stripe sizes, and both controller and disk cache enabled. The latter of which I'd prefer to have turned off (even with a battery).

I certainly wont make that mistake again. Of course, it's partly my fault for buying something without waiting for reviews (several other newegg buyers found it to be ludicrously slow as well), but I thought that after all these years it was a sure bet that *anyone* could turn out a decent hardware raid card if you give them over a grand. Apparently not. And I should have really researched the raid-5 write hole more before blowing $1200 on a supposed fix for the problem when a much better solution is to just use RAID-6 and the write intent bitmap (or ZFS).

Of course, I'm not trying to say *all* hardware raid cards are bad. I'm sure that most of them are just fine. But I just don't see any benefit to them any more. Linux has mdadm, *BSD/Solaris have ZFS. The only reason for hardware raid is if your operating system's software raid implementation is completely braindamaged. In other words, it's for Windows.

Funniest comment in a long time.

Many of the ads that I see make me *hate* the vendor. Before seeing the ad, I would have been fine buying from them. After, I avoid them like the plague, tell my social circle about why their ad made me dislike them, start a boycott, or maybe just leave flaming bags of doggy doo on their front steps. So by blocking ads, I'm actually boosting the vendor's sales. They should be paying *me* to block their ads.

In reality, the above is a bit silly. In most situations, the ads that drive me nuts are for products/services that I (and most in my social circle) would have never bought anyway, and the advertisers know that. The very things that make the ad so annoying to me are precisely what makes it effective on the actual target demographic.

For example, I was watching TV with some extended family when one of those supremely annoying used car dealership commercials came on, with the "M-M-M-Monster Sale! Friday! FRIDAY! Fri-day! We're going craaazy!" and some family members said something to the effect of "Sounds like a great sale! We should seriously get down there!" I was shocked.

Before you decide, be sure to watch "Portlandia", a recent TV Series (on a torrent near you). It's a documentary about how much fun it is to live in Portland. (Heh heh). Where else can you see stuff like this on a daily basis? :D

I've tried the computer-translated captions on several youtube vidoes, and they are even more hilarious than normal computer translations. They don't even bear a *slight* resemblance to what is spoken in the video. But I'm sure that the videos with clear audio (i.e. approximately the same audio quality that Dragon Naturally Speaking requires) work a lot better.

Duplicity backup software: $5/mo (donated to EFF)
FDCServers Atom + 2TB HDD: $45/mo
Comcast Internet with 20 Mbps upload: $130/mo
Running into your 250 GB transfer cap in just 24 hours: priceless.

There have been news items all year about how the E5 was going to usher in a new era of low-cost 10 GbE LOM (LAN on motherboard). Even today's news stories are talking about it. But where's the beef? I've looked through about 30 motherboards from Supermicro, Tyan, etc., and the only 10 Gb LOM I've found is on a proprietary Supermicro MB and it's not even ethernet. Sure, system integrators have them, but I'd rather build my own box.

Anyone have an idea where they are?

Sorry, I'm not trying to be obtuse. Maybe I can put it more simply. In 24-bit color (8-bit per channel), there is a maximum of 256 levels of brightness (for any and all colors): 0 is the darkest, 255 is the lightest. That is nowhere near the number of distinct brightness levels we can see with our eye. Hope that helps.

Actually, it is. On a 10-bit-per-channel display (with 10-bit software, 10-bit O.S., and 10-bit cable), you can display 1024 distinct values, each increasing value brighter than the next. If you display the exact same image on an 8-bit-per-channel display, there will only be 256 levels. There are a variety of techniques for compressing them with the smallest amount of image degradation (such as dithering to avoid artifacts from quantization error), but no matter what technique is used, the human eye can *definitely* see the difference.

That's all fine and well as far as colors go, but what about *range*? Even the layman can easily distinguish more than the paltry 256 levels provided by 24-bit video. It's very difficult to compress 12-14 stops of dynamic range (a typical DSLR raw file or film negative) into 256 levels. You can spread the levels out so that each doubling of intensity has about 18 levels to it, but it looks terrible ("low contrast"). Of you can compress a huge number of stops into just a few levels, so that you preserve the lion's share of the levels for midtones to have an attractive image ("high contrast", "pop", etc.). Or you can just cut the head and feet off the image until it's down to just 7 or 8 stops, which is easier to fit inside the 8-bit range limitation.

I disagree. Do you really believe that human vision is limited to seeing just a paltry 256 values for the entire range of light to dark? Not even close.

I love the column on video, where the 1995 columns says "24-bit", and the 2012 column...oh wait, we're still 24-bit. Everything else has advanced by several orders of magnitude, but we're still limited to just 8 bits per color channel (RGB = 24 bits in total) going out over the DVI cable (and the display itself). Sure, now you can drop a few G's on a 10-bit (30 total) monitor (if your software can even make use of it), but it's kind of sad that progress has been so slow.