Monday, 26 April 2004
8:14 - I just posted the first draft of Building a Small Form Factor (SFF) PC on the subscribers' page.
Here's something interesting, at least to me.
And here's what I wrote:
So I decided to estimate the status today, five years later. The dozen or so active systems around here have:
Not all that long ago, that would have defined a supercomputer. Even today, if I ran all the systems around here as a Linux cluster, I'd have a "micro-supercomputer".
In the five years from 1999 to 2004, the computing power available to me has increased by a factor of about 30 for hard drive space, about 10 for RAM, and about 10 for processor. If that growth curve holds, in 2009 I'll have:
And, in 2014, I should have:
Relative to my projections five years ago, hard disk space has increased dramatically, processor has doubled over the previous projection, and RAM has increased a bit less than projected. All of this assumes that Moore's Law remains valid, of course, which I think it will. Doubling everything every 18 months roughly translates to bumping things by an order of magnitude every five years, and that's how it's worked out.
By 2019, or 2024 at the latest, I should have enough local storage to store the sum total of distilled human knowledge. Every book and magazine article ever written, every scholarly journal, every movie and sound recording ever made, every photograph ever taken. Everything. And I should have the memory and CPU power to take advantage of that massive amount of data. There are two questions, though. First, will the software be there to allow me to mine this overwhelming amount of data? And second, will that data actually be on my local network, or will the copyright monopolists have locked things up so tight by then that none of really have anything to store locally? And a third question. How will we back all this stuff up?
Tuesday, 27 April 2004
9:31 - When Barbara got home from the gym yesterday evening, she announced that she wanted a digital audio player. I thought she'd been using her portable CD player, but she says it skips when she bounces it around while exercising. I thought it was supposed to have anti-skip buffering, but apparently it doesn't work very well. She's taken that to work to listen to music there, and is using her old cassette tape player at the gym.
So I went off in search of portable digital audio players. She wants something small, light, and relatively inexpensive, say under $100. I'd about decided on a Creative Labs NOMAD MuVO NX. NewEgg sells the 128 MB version for $97. Then I noticed that it supports only MP3 and WMA. I really wanted something that supports OGG Vorbis, both because its sound quality is better at a given bit rate than MP3 and because OGG Vorbis uses an open format. But I can't find an OGG-compatible player at anything near our price level.
I'm actually in the process right now of ripping Barbara's rather large CD collection to .WAV files using Exact Audio Copy. I'd intended eventually to compress all of them to OGG files and store them on the server. Ideally, I'd like to be able to have Barbara simply pull down whichever ones she wants and store them in her portable digital audio player. I really don't want to have to compress the .WAV files in two formats and store both.
Unless there's a sub-$100 OGG-compatible digital audio player I don't know about, it looks like I'll have to choose between MP3 and WMA. I think I know that WMA provides better sound quality at a given compression level, and I think it's possible to compress .WAV files to WMA without including any DRM. I also think I read recently that someone has added WMA support to a Linux distribution, although I think they have to pay royalties to Microsoft. So perhaps my best option would be to compress the .WAV files to MP3 at 256 or 320 Kb/s and have done with it.
I confess that if I use a good encoder like LAME, I can't tell the difference between CD audio and a 256 or 320 Kb/s MP3 file, even using good speakers or headphones. Barbara can tell, though, particularly with classical music. Perhaps I should store both the WAV files and the compressed files. I do have a server sitting here with 320 GB of RAID 0+1 storage, which I could easily reconfigure to 640 GB of RAID 1. That'd give me room for lots of uncompressed CD audio, along with the compressed audio files, video files, and so on.
I'm also not sure which application to use to manage all those audio files. EAC has a pretty good database/library function. So does ATi's Multimedia Center. I'm hoping that the index data is easy to transfer among different library management programs. As it is, I'm using CDDB as I rip the CDs. I assume that if I later use a different library management program that it will be able to use that index data without me having to re-rip the original CDs, but I don't know for sure.
Any advice would be appreciated.
Lots of comments on yesterday's journal entry, both on the messageboard and via private email. Most people suggested replicating data to multiple hard drives and multiple systems on the local network, which of course I do already. Several people pointed out that if every PC contains the entire database of the world's knowledge, there's no need to back it up. A few people sung the praises of tape technology or external hard drives. Here's a representative one:
Yes, I'd reached the same conclusion. In fact, I did a sidebar in the "Building a SOHO Server" called "Backing up the Beast". In a SOHO environment, backing up 320 GB or 640 GB is not trivial. I concluded that the best solution was to use USB 2.0/FireWire external hard drives, which I suggested that readers look at, not as hard drives, but as "funny looking tapes".
My real concern with hard drives versus tape is the unrecoverable bit error rate. For hard drives, that's typically 10e-14. For tape drives, it's more like 10e-19. In other words, hard drives are about 100,000 times more likely to produce an unrecoverable bit error storing a given amount of data than a tape is. Those numbers seem tiny until you consider the amount of data being backed up. If you back up 100 GB to hard drive, that's roughly 10e12 bits, so there is a small but very real possibility that an unrecoverable bit error will exist on that backed up data.
The answer, of course, is to do two backups. Unless you're backing up one 100 GB file--or at least a bunch of huge files--that second backup greatly reduces or eliminates the chance of the same file having a corrupted bit on both copies. Of course, if you're backing up a huge database, that may not be an adequate guarantee, because a corrupted bit anywhere in the database may be a serious or fatal problem. Of course, by making two copies, you can reduce the overall probability of any unrecoverable bit error from something on the close order of 1 in 100 to something like 1 in 10,000, which is probably good enough for most people.
Wednesday, 28 April 2004
10:04 - Thanks to everyone who sent suggestions about audio encoding for the portable digital audio player Barbara wants. The award for most unusual suggestion goes to Brian Bilbrey, who suggested that I put a battery in the SFF PC I just built and give it to Barbara as her portable digital audio player. Barbara said the wrist strap would be the problem...
All of these little flash memory players seem to support MP3 and WMA. A few support other formats, but those are the more expensive models. Barbara wants something inexpensive to use primarily at the gym. She doesn't want to have to worry about it getting eaten by the treadmill or something. In the sub-$100 range, the players provide at most 128 MB of storage and the format choices are MP3 versus WMA.
As it turns out, OGG support is moot, because at the quality level I'd want to compress the primary copies of our CD audio data, not much would fit into a 128 MB player anyway. So, after doing a bit of research, I've concluded that the best option is to compress the .WAV files twice, once at 128 k/s for Barbara's digital audio player, and once at very high quality (and a correspondingly high bit rate) for storage on the server and distribution throughout the house. As to format, I'll probably use Ogg Vorbis at its highest quality setting or nearly so for compressing the copies that'll reside on the server, and Windows Audio 9 to produce 128 k/s WMA files for Barbara's portable player.
In fact, I may use WMA for the high-quality files as well. My concern about .OGG files is compatibility with different players. Although it'd certainly be no problem to play the .OGG files, I may want to use a player that doesn't support them (e.g., the player that ATi bundles with their RADEON All-In-Wonder adapters). WMA is not an open format, but I can play it on any Windows machine or on a Linux box using mplayer, so access to the files isn't a problem.
I thought about trying to batch-convert compressed files from high-quality to 128 k/s quality, but I understand that method is plagued by "moire" issues, no matter what format I use. Instead, I'll simply run the conversions twice, each time from the original .WAV files.
13:57 - And speaking of digital music, here's an article about a Russian site that claims to be selling copyright-cleared digital music legally for $5 per 500 MB. You specify which tracks you want, which format, and the bit rate. They rip and compress them on-the-fly and deliver them to you. You can even get the music uncompressed if you're willing to pay for the extra bandwidth. That amounts to about $1 per CD's worth of music at normal compression levels, and about $5 per CD for uncompressed music.
I haven't tried using this service, and I won't. There's no indication that it's a scam, but I dislike providing my credit card number even to ordinary vendors, let alone an unknown Russian outfit.
16:04 - Here's a change of pace.
Hmmm. I haven't played tennis seriously in 30 years, but I suspect the fundamentals are still the same. I never had much trouble hitting overheads, but I know a lot of people do. If I had to come up with a top-ten set of rules for hitting overhead smashes, they'd be (in no particular order):
1. Don't let the ball get behind you. A lot of people are so excited when they get a short lob that they rush toward the expected point of impact and end up overrunning where they need to be. They then backpedal furiously to try to get back under the ball. If the ball is too far behind you, at best you'll hit a weak overhead. You'll probably hit it into the fence. Or, if you really overran it, you may not be able to get to the ball at all.
2. Bounce the ball. A lot of people think it's more impressive to hit the smash without letting the ball bounce, but unless you're forced to do so it's a stupid move. After the bounce, the ball is falling from a much lower altitude and therefore moving much more slowly. I've had people argue with me that bouncing the ball gave my opponent more time to prepare. My response was, "so what? He's not going to get his racket on the ball even if he has a week to prepare."
3. Hit it with topspin. I used to serve flat "cannonball" serves for both first and second serves, but I usually hit overheads with topspin to help pull them back down into the court. Because you're normally inside the baseline, any but the shortest player can nearly always hit down on an overhead, which means you can hit the ball very hard and still use topspin to pull it down and into the court. To get topspin on the ball, simply place it further to the left (if you're right-handed) than you would for a flat serve.
4. Contain your excitement. If you hit many overheads into the net, especially into the bottom of the net, it's possible that you're too far behind the ball. But it's more likely that you get overexcited about killing the lob and just can't wait long enough to "pull the trigger". Treat the lob like a service toss, and tell yourself that you have all day to hit the ball.
5. Don't aim for the corners or baseline. You can hit an overhead smash harder than any other shot, including your first serve. You don't need to hit corners or the baseline to make your overhead a clear winner. Aim well inside the lines. If there's an opponent at net, aim for him. (Unless the "him" is a her. During a mixed doubles tournament, I once nailed a woman at net, and she came after me swinging her racket. I think she'd been watching Rosie Casals.)
6. Track the ball with your free hand. As you position yourself to hit the overhead, use the index finger of your off hand to point at it. It sounds simple, but it's the best way I know to get yourself positioned properly relative to the ball.
7. Be moving forward as you strike the ball. Position yourself a bit behind where you think you're going to hit the ball, and then move forward at the last instant so that you're actually approaching the net as you strike the ball. Your overheads will have a lot more power on them, and you'll find that you never end up with the ball behind you.
8. Practice hitting overheads. For some reason, few people practice hitting overheads, so when an opponent tosses up a lob during a match, they're hitting a shot that they almost never hit except for real. When I was using a ball machine, I used to hit 100 or more overheads in a row as a routine part of practice. During warm-up, I'd always toss up at least half a dozen lobs for my opponent, and expect him to do the same for me.
9. Don't overhit the ball. My playing style was to hit just about every ball as hard as I could. I didn't even have a second serve. I hit two first serves. The one exception was with overheads, which I hit at maybe 80% of maximum power. It's pointless to overhit an overhead. You already have all the advantages. Your opponent is already on the defensive. You're normally well inside your baseline, which means the whole court is open for you to hit into. You're hitting down on a slow-moving ball. Overhitting simply means you'll miss a higher percentage. Hitting at moderate pace means nearly all your overheads will go in, and nobody is going to get them back anyway.
10. After you hit the overhead, don't stand there admiring your shot. If you hit a hard overhead, about 90% or 95% should be outright winners, but that still leaves 5% or 10% that are going to be returned. After you hit the overhead, continue moving forward toward the net. You can close in on it more than you would for an ordinary volley, because any return is likely to be weak, allowing you to volley it off at a sharp angle. If the return is a lob, as it usually will be, it's likely to be very short. If your opponent does manage to lob deep into your court, it's going to be high, giving you plenty of time to get to it (unless you're playing my brother, who is the only person on the planet I have ever seen other than Ilie Nastase who can consistently hit backhand topspin lob half-volleys.)
Thursday, 29 April 2004
9:44 - Our dogs really don't like each other, which I suppose is normal for two males.
Whenever anything bad happens to Malcolm, he blames it on Duncan. For example, if I yell at Malcolm for something, he immediately accosts Duncan, looking for fight. Duncan isn't as obvious about it, but he does the same thing. This morning, as I was connecting the leashes to the dogs in preparation for taking them out the front door, I accidentally stepped on Duncan's paw. He yelped loudly, and immediately went after Malcolm.
I really hope they make peace with each other before Duncan gets much older. Duncan is nine years old, and Malcolm is six months short of his fifth birthday. They weigh about the same, although Duncan is tall and slender and Malcolm shorter and squatter. Until a year or so ago, Duncan almost always came out on top when they got into it. Lately, Malcolm has been winning. Duncan has some weakness in his rear end now, and when Malcolm jumps on him, Duncan collapses. I'm afraid that as Duncan continues to age, he's going to get hurt.
The only good thing is that Malcolm is a much gentler dog than Duncan, although a casual observer would think the converse. Malcolm is aggressive in confronting Duncan, and willing to fight at the drop of a hat. However, Malcolm's attacks are almost exclusively threat displays, with no real attempt to harm Duncan. Malcolm will occasionally end up with a mouthful of Duncan's fur, but almost never draws blood.
Duncan, conversely, is a very laid-back dog. He doesn't like to fight, and it takes a lot to get him to fight. But once he starts to fight, he's fighting seriously, and fighting to win. Duncan almost always draws blood, and he won't give it up. If Barbara and I separate the two, Malcolm is willing to call a truce. Duncan, on the other hand, tries to get to Malcolm to continue the fight.
It's a pure dominance thing. Both want to be top dog, and neither is willing to settle for second place. Their belligerence is directed solely at each other. Neither of them would dream of harming a person, and they even get along with nearly all other dogs. They fight only with each other.
My inclination has always been just to let them fight it out. That's the natural way. Both of them would probably be bleeding when the fight was over, but one of them would have conceded to the other. It's unlikely that there'd be any serious damage. Dogs instinctively fight within the pack using very strict rules about how much damage is permissible. Dogs don't understand political correctness. I think that separating them simply draws things out because they're not able to come to a resolution in the only way they understand. I'm just afraid that with no conclusion and ongoing fights that Duncan is going to end up being seriously hurt one day.
17:30 - I hope the penultimate sentence in this article contains a typo. The article is about liquid body armor, which sounds strange but is apparently practical. Liquid body armor depends on Shear Thickening Fluid (STF), which is a suspension of very hard solid particles in a liquid polyethylene glycol base. The mixture remains liquid until something impacts it at high velocity, at which point it essentially turns solid. The article states,
"The transition happens very quickly, a millisecond or quicker."
Let's see now. Assuming a high-velocity rifle bullet strikes liquid body armor at 3,000 feet/second, that means that bullet will only be able to continue for 3 feet before the armor solidifies. In the front, through the victim, and out the back.
Hmmm. Sounds like they need response time about three orders of magnitude faster. In one microsecond, that 3,000 fps projectile would travel 0.003 feet, or about one millimeter before the vest hardens. Even two orders of magnitude wouldn't do it, because the projectile would travel 10mm, or nearly half an inch. Ouch.
Friday, 30 April 2004
9:51 - Fred Reed has posted another column worth reading. In it, he talks about what he calls the cognitive elite, and their assumption that everyone is, fundamentally, like them.
... because she does and everyone she knows does. Which is how I see the world around me unless I explicitly stop to consider. I've written about this before. I interact almost exclusively with very, very bright people, both face-to-face and via the Internet. The mean IQ of my readership is so far to the right side of the bell curve that one needs a magnifying glass to separate it from the baseline. Most of the people I speak with every day are genius level or higher. I've exchanged email with half a dozen people so far this morning, all of whom are probably at least 3.5 to 4 standard deviations above the mean. One of them is probably six.
When one is surrounded almost exclusively by such people, the natural tendency is to begin thinking of them as "normal", which they clearly are not. It's a small step to make the tacit assumption that everyone else must be like them. Of course, when I think about it I know that's not the case, but it's an easy error to make.
As Fred says, like seek like, which means that in modern society there is an increasing stratification based on intelligence. Smart people tend to associate only with other smart people and stupid people only with other stupid people, and neither group understands the other. Smart people don't realize just how stupid stupid people are, and stupid people don't understand how smart smart people really are.
If this is carried to its logical conclusion, I have to wonder if ultimately we're looking at a bifurcation into two sub-species that are capable of interbreeding, much like H. sapiens sapiens and H. sapiens neanderthalensis. The parallels are not exact, but interesting nonetheless.
10:43 - I just posted the "final" version of the "Building a Small Form Factor (SFF) PC" on the subscribers' page. That's the last of the "project system" chapters. Now I have to finish the introductory chapters. Crunch time approacheth, so you probably won't be hearing much from me during the month of May.
9:09 - The start of a new month. I'd hoped to have the book at 100% completion as of yesterday, but it was not to be. As things stand, I have three chapters plus the preface left to complete. All of those are in progress, but need a lot more work. That means there won't be much posted here until I finish them.
Dinner with Paul and Mary last night at their home. I wanted to look at their DBS system. They had Time-Warner analog cable at their old place. When they moved to their new house a year ago, Time-Warner installed analog cable there for them. But when Paul called Time-Warner to ask them to add HBO, they told him they wouldn't do it unless he upgraded to digital cable. Apparently, when they installed analog cable at the new place it was a mistake. They're supposed to install only digital cable for new installations. Their policy is that anyone who has analog cable can keep it, but they won't upgrade it, add premium channels, and so on.
Digital cable is a lot more expensive than analog cable. Paul and Mary, faced with either going without HBO or paying an extra $30 a month for digital cable, decided to drop cable service entirely and go with satellite. Like digital cable, satellite uses a digital feed, and I wanted to look at digital. I actually called Time-Warner and asked them if I could get digital cable installed in addition to our existing analog cable, but they refused to do it. The only option they gave me was to discontinue our analog cable service before they'd install digital cable. Having done that, we'd be stuck with digital cable, because they won't re-install analog cable. So I needed to look at the installation of someone with digital cable or DBS.
Their set-top box has a feature that makes recording programs a heck of lot easier than it was formerly with digital systems. It has a 16-event timer, much like a VCR timer. You can specify time, channel, and duration, and the set-top box handles turning itself on and off and changing channels. All you need to do to record is set up your VCR to record on channel 3 at the proper times, and the set-top box timer takes care of changing channels for you. This seems like such a simple and useful feature that I'm surprised all set-top boxes don't provide it.
Oh, and Paul comments with regard to my journal entry yesterday that he doesn't think bifurcation will be a problem as long as there are horny Smart teenage boys and pretty Stupid teenage girls. He may be right.
Sunday, 2 May 2004
Copyright © 1998, 1999, 2000, 2001, 2002, 2003, 2004 by Robert Bruce Thompson. All Rights Reserved.