At some point, if there are enough consumers living beyond Earth orbit to commercially matter to normal web sites (or their future equivalent), the most logical step would be to allow web applications to bundle themselves and their data in a way that allows the web app ITSELF to be automatically deployed to a remote virtual server closer to the end user & sync its mirror of the database with the authoritative one back on earth (multicasting most of the replicated data from earth to elsewhere & caching it blindly at the local end to the greatest extent possible).

The basic technology already exists today (think: .war files and databases with replication, load-balancing, and hot failover), but the number of users (say, passengers on a cruise ship) is too small relative to the expense & currently-insurmountable licensing barriers (Oracle isn't going to start making "replicant" licenses for Oracle and MySQL available that would be remotely cost-effective anytime soon).

A lot of people forget that the annoying habit some websites have of using AJAX to send data nonstop keystroke-by-keystroke is a plague of fairly recent origin. There's no reason why web applications HAVE to be written that way. If latency is high, but bandwidth is abundant, it makes more sense to satisfy the request by bulk-sending everything the requestor is likely to want anytime soon, and do it in a way that allows it to be cached locally for the benefit of others who might make the same request later.

The thing is, if cable/satellite went to true a-la-carte, any short-term savings would be rapidly neutralized by skyrocketing per-channel costs. Instead of paying $89.95/month for 200-300 channels, you'd be paying $84.99/month for 17 specific channels. Someone who literally watched only one or two channels might come out ahead, but the cable industry would make sure that most of us were at least as fucked as we are now, and would use it as an opportunity to fuck other customers just a little bit harder. It's depressing, but deep down inside, we all *know* it's true.

There's another category nobody in the industry even wants to TALK about (besides complaining about the costs, then passing them straight along to customers) -- the fees paid to local channels that insist on payment in lieu of "must-carry". IMHO, the FCC should amend the "must-carry" rule to REQUIRE cable/satellite providers to break out the total amount they'd otherwise have to collectively pay the local channels, and allow customers to opt out of paying that fee in return for having to use an antenna to watch local channels. The probably couldn't get away with demanding a breakdown of channel-by-channel fees due to nondisclosure rules, but if they disclosed their aggregate cost, it wouldn't really reveal much. I can almost *guarantee* that cable/satellite customers who don't live in New York City would begin opting out EN MASSE if they started seeing line items on their bills like "Local Affiliate Rebroadcast Fees - $17.47 (optional)". In the short terms, the local affiliates would probably raise prices for the remaining customers who don't want to bother with an antenna... but the higher they raised them, the more cable/satellite customers would be motivated to opt out of the fees and switch to OTA reception for them.

> And why shouldn't we count satellite and wireless?

Satellite is broadband for the damned & desperate.

I really wish cable/satellite would adopt "Chinese Menu" pricing for their mid-tier, and allow people who don't care about Disney*.* or ESPN*.* to pay the same price, but substitute HBO and/or Showtime instead (ie, pick two out of four... Disney, ESPN, HBO, Showtime... 3 for $10 more, all 4 for $18 more). I believe it would mostly be revenue-neutral for the cable/satellite companies, and would go a long way towards softening the sting of my monthly cable bill by letting me substitute two channels I don't currently pay for, but would LOVE to get instead of two expensive blocks of channels I never watch.

The problem with many of OCZ's drives (like the Vertex2 and Agility2) was that traditional RAID wouldn't save you because whatever killed drive #1 could (and often did) kill one or more of your OTHER drives, too.

The fault lies 100% with OCZ. They omitted the supercapacitor that Sandforce's engineers intended to keep it powered up if it unexpectedly lost power during a write, and they compiled their drivers to NOT use the multi-step write strategy that a drive without backup power needed in order to write safely and recover gracefully from power loss (because multi-step writes killed performance).

I don't know about him, but in my case, I happen to like having 5GHz 802.11ac with beamforming. The last time I checked, every open firmware in existence for the AC68U has broken support for 5GHz 802.11ac and its advanced radio features.

It's not so much "tough" as "the goddamn solder paste has to be kept refrigerated until you're ready to use it, then goes bad within a few days of warming it up to room temperature". Throwing away mostly-full syringe after syringe of solder paste after you used maybe 0.5mL of it to solder one chip gets expensive after a while.

LOL, if you think Digikey and Mouser are expensive, check the prices (including shipping) from Newark or Farnell. Jesus God naked on a Harley, Newark absolutely *rapes* you on the shipping charges. I can't even count the number of times I *almost* bought something from Newark, then called it off once they revealed their criminally-expensive shipping charges at checkout time.

DigiKey's shipping is expensive, but they have a huge advantage -- if you absolutely MUST have something tomorrow, their cutoff time for next-day delivery is something like 10pm most nights, and they'll happily ship via USPS Overnight so you can order on Friday night or Saturday and get your part by Saturday or Sunday.

Mouser's shipping isn't *cheap*, but their rates for Priority Mail are relatively reasonable & they don't have a minimum order amount like DigiKey does.

I still have occasional fantasies about Radio Shack closing MOST of their retail stores, but keeping at least one store per metro area open that's well-stocked with just about any component hobbyists might want to buy (say, everything that SparkFun and AdaFruit sells). Back in the 80s (when Radio Shack's parts department occupied the rear third of the store), people paid INSANE prices for components because you could walk in, grab what you needed, and go home with it in hand.

Ah... but were they able to do it with a chip that was available in DIP form, with useful amounts of flash & ram, and relatively relaxed power & I/O design on a breadboard?

We've gotten spoiled by $7 Arduino knock-off boards from China, and a lot of us have forgotten that just 5 years ago, Atmel literally couldn't make the ATmega644p fast enough for stores like DigiKey and Mouser to reliably keep it in stock. For those who weren't into AVR microcontrollers a few years ago, the 644p was Atmel's beefiest AVR that you could buy in DIP form. The next step up from a 644p was a 1280 or 2560 (the 2560 is used in the reference design for Android ADK), and they easily cost $30-50 (~$30 for a board that was literally just a bare chip soldered to a breakout board, $50 for one that had most of the same hardware that's the norm for an Arduino board). The 1280 and 2560 themselves were fairly cheap... I think the 1280 was around $10-12, and the 2560 was around $15. But the act of having someone solder it to a board to make it something YOU could deal with (unless you had a hot air rework station & didn't object to buying solder paste that had to be kept refrigerated, warmed to room temperature over the span of a day, and went bad a few days later when the flux separated out) basically doubled or trebled the purchase price.

Back in the same era, it was almost UNHEARD of for people to buy breakout boards for Atmel's smaller chips, like the Mega168 (unless they were rank n00bs buying their first one), because a DIP Mega168 cost around $4, but a Mega168 soldered to a dev board with Arduino-like hardware ran about $25.Back then, the hardest problem every N00b had to solve was "how the fuck do I connect the 3x2 or 5x2 header from the AVRISP to MISO/MOSI/SCK/~RST/Vcc/Gnd on the breadboard (I used to endlessly wish somebody would make a breadboard whose pins from one side were extended by one into the middle , so you could stick a 2xN header straight into the breadboard and wire away (for some inane reason, breakout boards to convert 3x2 and 5x2 headers to breadboard-spacing were always outrageously expensive, and stayed that way until about the eBay floodgates from China opened about 2 years ago).

As a thought experiment, how much linear resolution/sample bandwidth would you need to capture a VHS videotape "straight through" in a way that allowed you to digitally-reconstruct the path of virtual rotating heads AFTER the raw linear capture was completed? Say, if you had the means to somehow fabricate a custom high-density matrix of read heads that massively oversampled the tape as it followed a linear path... assuming such a beast doesn't already exist for something like DLT?

Why? Because all present restoration technologies that involve rotating heads produce results that are very much dependent upon the player's ability to iaccurately track the helical paths at capture time. What I'm proposing is capturing a linear 2-D signal map of the tape that's so precise, you could analyze it & synthesize an optimal virtual path & instantaneous velocity to extract higher-order data in non-realtime that's not even visible when played back via normal means.

The biggest problem with most VHS-to-DVD transfers is inadequate sample resolution or encoding bitrate. VHS is low-res... but it's also noisy, and noise compresses badly. Encoding captured VHS video in a way that preserves it EXACTLY the way it was read from the VCR (preserving higher-order information like chroma shift and luminance noise) so it can be further restored later requires a MINIMUM 704x480 resolution and 6-8mbps bitrate. In English, that basically means you'll get about an hour of high-quality captured video from VHS on a single-layer DVD. Two hours might be do-able at good quality if you're digitizing 24fps content, but a straight VHS capture encoded to 4mbps or lower is probably going to look worse than the original from the tape did.

A good way to think about it is to envision taking a photograph of an early-80s videogame's CRT display. The nominal resolution of the game might have only been something like 320x200 or 256x192, but other factors like the electron gun alignment, mask pitch, bleed, bloom, etc enter into the equation as well. If your goal is to just capture an approximation of the display, you could resample it down to its nominal resolution... but if your goal is to preserve every nuance of the video's appearance as it appeared on the CRT (so you can literally emulate things like CRT alignment bleed and color halos on a 2560x1440 LCD), your photograph of the CRT will have to be MUCH higher resolution than its nominal resolution.

The tapes you'll have the hardest time transferring with a new VCR are the ones made with a cheap VCR after ~1999 (after everyone had basically switched to DVD for watching movies, but VHS still existed as a nasty kludge for time-shifting prior to DVRs becoming affordable & common). Old tapes recorded with expensive VCRs generally play fine unless they were stored someplace humid. New tapes recorded with $99 VCRs might not be playable by anything besides the VCR that recorded them... and even THAT's questionable. 1980s tapes were built to last, and VCRs were precision devices built to exacting standards. Early-2000s tapes were designed to cost a dollar to manufacture & last for a year or two, and the VCRs were as mechanically shoddy as they could get without outstripping the capabilities of the DSP chip.

In audiocassette terms, last-gen VCRs & tapes had CATASTROPHIC problems with what would have been called "wow & flutter" on an audiotape. Basically, VHS depends upon having a precise match between the tape speed and head rotation speed, and the last-gen models were UNBELIEVABLY sloppy with it. Often, as belts aged, they'd lurch and slip.

Big tip: forget everything you've been told about the "real" resolution of VHS tapes. If your goal is to preserve them in digital form forever, without making them worse, capture at 720x480 & accept the fact that noise compresses badly, so you'll need a fairly high bitrate (6mbps BARE MINIMUM). And don't be too eager to throw away YUV color information -- keep it 4:2:2 unless you literally don't ever intend to try restoring it. Remember, most video-restoration tools depend upon exactly the kind of "higher-order" that aggressive compression throws away & mangles.

Most importantly, never forget that the only thing Nyquist guarantees is that a sample rate less than double the information rate is guaranteed to fail... it says NOTHING (directly) about the minimum sample rate that will actually preserve higher-order detail.

The really bad news: at the moment, there's no compression scheme that's mostly lossless (like lossless h.264 with x264) AND compliant enough with Blu-Ray standards to be directly playable by a random Blu-Ray player. So... don't get too hung up on near-line playability. Capture your minimally-compressed gold copy and carve it into (non-LTH BD-R) stone, then re-encode a DVD-playable copy (704x480 MPEG-2 6-8mbps, AC3 or MP3 audio, possibly with long GOPs and/or VBR if you know your player can handle them).

> Samsung is planning a 2560*1440 cell phone.

The Oppo Find 7 already *has* one :-)

http://forum.xda-developers.co...

For those who've never heard of it, it's the phone everyone at XDA-developers.com has been having wet dreams about for months -- a top-shelf, best-of-breed Android phone that makes no hardware compromises & ships unlocked with Cyanogenmod. There have been officially "open" phones in the past, but they were always last year's hardware or lacked important features like microSD (when you reflash a lot, microSD makes your life several orders of magnitude easier & more convenient) and/or LTE.

The last time I checked, Oppo's plan for the Find 7 is to make it directly available in the US as a retail product with US warranty by June, but AFAIK if you're dying to get one now, the GSM international model is hardware-identical to what you'll be able to get from stores like Newegg, Amazon, and Tiger Direct. If 1700MHz AWS HSPA+ is important to you (ie, T-mo in a market that hasn't been refarmed yet), you might want to double-check support for AWS.

One caution... the mildly-rare model M13 with pointer stick (manufactured around 1995) has mouse buttons that are pretty fragile & become flaky after just a few years... and the printing on the black one with black keys rubbed off mine within a couple of years (a known issue with white printing on black keys, and the reason why the black ones are particularly rare). Also, the beige plastic insulation on the cable turned brittle and started disintegrating sometime around 2009 (purchased as new-old-stock sometime around 1999). For some reason, this didn't happen to the black keyboard's cable.

I know I could get the buttons fixed (I own two M13 keyboards... a beige one with white keys, and a black one with black keys), but the keyboards themselves DO seem to be having occasional issues now that they're approaching their 20th birthdays, and the original Trackpoint (I) is a little lacking in the resolution/sensitivity department... it was developed in an era where 1024x768 was a physically huge huge hi-res display, and it's kind of painful to depend upon as the only navigation device in a 3-monitor setup.