Monday, 5 April 2004
9:30 - More on Firefox/Thunderbird. I spent a while over the weekend trying to get Firefox/Thunderbird wiped off my system and replaced with Mozilla 1.6. I couldn't make it work. I uninstalled everything--Firefox, Thunderbird, and Mozilla 1.6--and then wiped out all reference to any of them. I deleted the profiles. I deleted all the directories. I deleted all references in the registry. Once I was sure everything was cleaned out, I re-installed Mozilla 1.6. It wouldn't run. I click on the icon or on the program name, and nothing happens. Nothing.
And then I encountered something really strange. On a whim, I decided to uninstall the non-functional Mozilla 1.6 and install Mozilla 1.5. Mozilla 1.5 installed normally, and ran normally. Urk? So I un-installed Mozilla 1.5 and installed Mozilla 1.6. It still wouldn't run. Uninstalled 1.6 and re-installed 1.5. It ran fine. What the hell is going on here?
Can a system be too fast? I think it can. In particular, WebWasher doesn't operate properly on very fast systems. It munges the rendering in Mozilla, IE, and Opera, and the amount of munging varies from page to page and even refresh to refresh. The typical symptom is a bunch of script code at the top of the rendered page. That can be anything from a fraction of a line to most of a screen.
I first noticed this problem when I replaced Barbara's system with a faster one (a P4/3.0G, I think). I really notice it on the faster system I'm building lately, which are mostly 3.2 and 3.4 GHz systems. There are other symptoms as well. For example, this page lists moonrise and moonset times for Pilot Mountain, where we observe frequently. On my main desktop system, which runs a 1.7 or 1.8 GHz P4, the page renders properly. On my den system, which is a P4/3.2, the page is butchered. It doesn't matter which browser. It butchers the page badly in Mozilla or IE. It doesn't matter if WebWasher is running, either. Again, the problem seems to occur only on very fast systems.
How can this be? I don't know. Perhaps a subroutine is executing so quickly that it finishes before another subroutine expects it to. I do know that these systems are extremely fast. For example, even after the system has been freshly booted, FreeCell loads instantly. And I do mean instantly. There is no perceptible delay between double clicking the icon and the program being displayed. None at all. Same thing with large programs like Word. Double-click the icon and, bam!, there it is.
And that's by no means the fastest system in the house. I have the Kick-Ass LAN Party PC sitting on the library floor at the moment. It's faster still, with a P4/EE processor, an ATi RADEON 9800 XT video adapter, and so on. I'll soon build a system that's faster still--the Personal Workstation--with a P4/EE, RADEON 9800 XT, and dual 15K RPM Seagate Cheetah SCSI drives. That system should load programs almost before I ask for them. But I have to wonder if my hardware is becoming too fast for my software.
Perhaps I should look into an alternative for WebWasher.
9:41 - Oh, yeah. I should have mentioned that I uploaded the second version of Building a Kick-Ass LAN Party PC to the subscribers' page on Friday, and the third version of Building a Home Theater PC yesterday.
11:03 - I got an interesting question from Ron Morse, who's one of the tech reviewers on the new book. I was going to go out and research it myself, but I figured one of my readers would know the answer.
To be honest, I haven't been following this very closely. As the December 2006 deadline for switching off analog broadcasting draws nearer, the programs being broadcast continue to get worse. Just in the last couple months, Barbara has stopped watching NYPD Blue and ER. She's down to watching just Left Wing now. By 2007, I suspect the only things we'll care about on TV are the Weather Channel and news, and we can get those from the Web.
I confess that I have no idea what happens to cable. Will our local cable system continue delivering analog versions of digital channels? Is that even feasible? Or does everything go 100% digital and 100% HDTV as of 1 January 2007? If so, we may not bother buying a new TV.
Perhaps I'll just post this on my journal page and let my readers tell me what's going to happen.
12:33 - Here's what looks to me like a pretty good answer.
15:26 - I am generally uncomfortable with child pornography laws. It seems to me that they're flawed on several bases, not least that they make a presumption of guilt rather than innocence. I might support a law against producing child pornography if "child" were defined reasonably, say a person under age 12, but it has always seemed to me that criminalizing mere possession of so-called child pornography is wrong. It is particularly wrong as it is done now, because there is no burden of proof on the state to prove that the models are younger than the statutory age, or even that the models exist.
Yes, it's true, one can be convicted of child pornography for possessing images that were created on a computer without using a live model, and one can certainly be convicted without the state being required to produce the "child" in question and establish that that "child" was under the statutory age at the time the image was made. People have in fact been convicted of child pornography for possession of written fiction that features sexual acts by children. Crimes of the imagination can be prosecuted ferociously nowadays.
But even when justice is perverted on that scale, it seems bizarre to me that a child, at least as the law defines it, can be charged with child pornography. I read about this case several days ago, I think on CNN or FoxNews. A girl, age 15, has been charged under child pornography statutes for producing and distributing pornographic images of herself.
Now, this girl needs a good talking to, no doubt. She probably needs to be spanked. But charging her with felony child pornography seems a bit outré.
Tuesday, 6 April 2004
10:00 - Still heads-down writing/editing. I have to take a break soon to do my taxes, which'll put me in a bad mood for the rest of the month.
14:52 - I just posted the third and "final" version of Building a Kick-Ass LAN Party PC on the subscribers' page. I say "final" because as far as I'm concerned I'm finished with it. O'Reilly editors will now have at it, and I'll get PDF page proofs back to do a final check on it.
Wednesday, 7 April 2004
9:09 - An interesting question has arisen. One of my technical reviewers on the new book commented that I've forgotten in all of the project systems to use an audio cable to connect the optical drive to the motherboard or sound card. Years ago, I used to forget to do that all the time and it was a problem because audio back then didn't travel across the bus. Recent optical drives and motherboards, though, transfer audio digitally across the bus. Of the dozen or so live systems I have here, not one of them has an audio cable connecting the optical drive to the motherboard, and the audio on all of them works just fine, at least for the purposes for which I use audio.
So the question is, what am I missing? The Plextor PX-708A, for example, has not one but two audio connectors on the back. One is a standard MPC analog audio connector, and the other a SP/DIF digital audio connector. Plextor doesn't supply a cable for either, nor do the motherboards I use. I realize that I could connect the digital audio connector to a SP/DIF-in on the audio card, but I'm not sure what that buys me relative to just allowing the audio to be transferred across the bus. And I sure can't figure out what good the analog audio connector on an optical drive is nowadays.
Any comments would be appreciated.
11:32 - Everyone from AnandTech to The Inquirer has commented about Intel's new PR numbering scheme, and they all wonder what happens as Intel processors get faster. The fastest Pentium 4 models currently on the roadmap are the 570 (3.8 GHz) and the 580 (4.0 GHz). That puts the 4.2 GHz version at model number 590, and the 4.4 GHz version at model number... uh-oh. What, everyone wonders, is Intel going to do when they "run out" of 500 numbers.
Well, Intel could start using increments of less than 10, as they do for the Celeron, but that's not what's going to happen. Intel's announced roadmap is their subtle way of telling everyone that there will never be a 4.4 GHz Pentium 4, and possibly not a 4.2 GHz model either. Instead, Intel will introduce the Pentium V--I've suggested to Intel repeatedly that "V" looks a lot better than "5", kind of like the old V-8 logo; we'll see if they take my advice--probably at 4.2 GHz or 4.4 GHz, and probably as the "Pentium V Model 510". Of course, that brings up the ugly specter of Pentium 4 Model 580s going up against Pentium V Model 510s, but that's simple enough for Intel to address. The PR numbers are good only within a product line, don't you know, and of course even a slow "V" is faster than a fast "4".
Of course, I could be wrong. Intel could decide to make the Pentium V the 9-series, or, if they want to stick with prime numbers, the 11-series, although that would put them in the odd position of having desktop processors with higher model numbers than their server processors. But whichever they do, remember, you heard it here first.
15:18 - I just posted the third and "final" version of Building a Mainstream PC on the subscribers' page. I say "final" because as far as I'm concerned I'm finished with it. O'Reilly editors will now have at it, and I'll get PDF page proofs back to do a final check on it.
16:05 - There's an interesting article over on NewsForge about the reliability of self-funded studies (or the lack thereof). Reading it reminded me of the one time I contracted with a certain very large software company to write a white paper for them. The pay was to be excellent. As I remember, they were willing to pay $30,000 for a 30-page white paper. There was a lot of work involved. I think it would have taken me a couple months to finish. Still, excellent pay.
The white paper was to be used internally, and not provided to customers. The subject was a detailed comparison of that large company's relatively new network operating system versus Novell NetWare, and the contact at the large company told me repeatedly that they wanted a "completely unbiased comparison" of the two products. As he said, it was for internal use only, and it was important for their sales and marketing people to understand both the strengths and weaknesses of their product versus NetWare. He even commented at one point that it was more important to point out the weaknesses than the strengths, so their people could figure out how to address those weaknesses. (At the time, I thought he meant address them in technical terms, i.e., fix them, but as things progressed, I began to believe he meant "address" them as in figure out how to talk around them.)
I got started on the project, having been assured that I was to write what I thought without concern for anything but accuracy. I'd put a couple of weeks' preliminary work into the project, outlining, deciding what we'd cover and in how much detail, and so on. Then I started actually doing the comparisons. I had two servers running each OS up in my office so that I could test things as I wrote about them. I started with a comparison of Novell's NDS directory service versus the large company's version. There simply wasn't any comparison in terms of functionality, performance, scalability, or any other aspect of the two directory services.
The large company's so-called "directory service", for example, enforced a flat name space. That is, one could have only resource with the same name throughout the directory service structure. If you had an employee John.Smith in your Dallas office, the guy named John Smith in the Milwaukee office had to use some other identifier. John.Smith was already taken, and that meant it was taken for the entire organization. I was flabbergasted. I spoke to the technical folks at the large company, and they confirmed that what I believed was true. (It may still be true for all I know, although the version I looked at was about 0.1.)
So I wrote the comparison of directory services, and I told the truth. My primary contact at the large company freaked out. The conversation went something like this:
Him: "You can't say this!"
And the conversation went downhill from there. For a long time, I thought the large company wouldn't pay me at all for the work I'd done at their behest. In terms of the agreement, they probably owed me $10,000 to $15,000, but I never expected to see any of that. Months later, my agent told me that he'd negotiated a kill fee with the large company. As I remember, it was $2,000 or $2,500. Not much to show for the amount of work I'd done, but better than nothing.
And to this day I refuse to work for that large company, despite the fact that they continue to pay extremely well for white papers and other technical writing. Perhaps it's foolish pride, but I really don't want my name on something unless it honestly represents my own beliefs. I might have stuck out that project and given them what they wanted, but I insisted that if I did that I wouldn't allow my name to appear on the white paper. They insisted my name must appear, so we had nothing left to talk about.
Thursday, 8 April 2004
9:44 - We're not the only ones. Some time ago, Time-Warner screwed up a network upgrade and left users with older Toshiba 1100 modems, including us, without service for several days. Now I see that Cox Communications in Washington, DC has done something similar. They attempted to upgrade the firmware in users' Toshiba 1100 modems. The process died in mid-upgrade, leaving those modems useless. But at least their modems aren't bursting into flames.
Mark Huth responds to my comments yesterday about writing a white paper for a certain large software company.
Yes, Microsoft has always been good that way. I remember when OS/2 and Windows were duking it out in a contest for which'd be the next mainstream OS. Microsoft was giving away their SDKs to anyone who could fog a mirror. IBM was selling their SDK, for $600 as I recall. OS/2 was definitely the superior product in a technical sense, but it had no chance with IBM treating SDKs and developer relations as a profit center. It reminds me of a comment John Dvorak made years ago, when AT&T was still in the PC business. He said that if AT&T bought KFC, they'd market the product as "cold, dead, greasy fowl." (Pournelle says the original comment was his, and that Dvorak picked it up and ran with it, but did give credit to Pournelle.) At any rate, at least back then, if IBM had bought KFC, they'd probably have marketed the product as dead, rotting bird parts.
Back to work on chapters. I've finished and posted three of the "project system chapters--"Building a Mainstream PC", "Building a Kick-Ass LAN Party PC", and "Building a Home Theater PC". Today, I'll incorporate comments from my editors and tech reviewers in the "Building a SOHO Server" chapter and get that finished.
These chapters are huge, and page-count is becoming an issue. I'd originally planned to include two additional "project system" chapters--"Building a Cheap (But Good) PC", and "Building a Personal Workstation". I've already written the first part of each of these chapters, but I may have to abandon them, at least for now. There may simply not be room for them. (If not, I'm hoping O'Reilly does a second volume of the book with more project systems; I have ideas for several other "must-have" PC projects).
We're going to do our damndest to cram a fifth "project system" in, but it won't be one of those I mentioned. Originally, I'd planned to include a chapter, "Building a Small Form Factor (SFF) PC", but I decided not to do that. I have qualms about existing SFF PCs like those from Shuttle. There are a lot of issues, but one of the biggest is cost. When I priced Shuttle systems, I was shocked. Paying $350 to $400 for a case, power supply, and second-rate motherboard is simply outrageous. These boxes are all at least semi-proprietary, which increases cost and reduces flexibility.
Come Antec with their new Aria SFF case, which address most or all of my concerns with SFF systems. It's inexpensive and I'm sure the build quality will be up to Antec's usual high standards. It uses industry-standard microATX motherboards, so all the proprietary issues go away. It includes an excellent Antec power supply, rather than the low-end units typically used in SFF boxes. Antec says it's not only the quietest case they make, but it's the quietest case available anywhere. That'll be a nice contrast to typical SFF cases, many of which sound like vacuum cleaners. (Yes, I know some aren't that loud, but those are typically the ones with wimp power supplies.)
So I'm going to pull out all the stops and build an SFF system for a new project chapter. Antec USA just got their first samples of the Aria case in yesterday. They're sending me two, which should be here by Friday. I'm really looking forward to building some systems around them. I've always like the SFF idea in theory. Now we'll see how it goes in practice.
11:35 - I just posted the third and "final" version of Building a SOHO Server on the subscribers' page. I say "final" because as far as I'm concerned I'm finished with it. O'Reilly editors will now have at it, and I'll get PDF page proofs back to do a final check on it.
14:26 - Jerry Pournelle posted a letter from a reader yesterday, complaining that Apple wouldn't ship to APO addresses. Jerry asked, "Does anyone know why this is so? Is is political correctness?" To which I responded with the following message, which Jerry published (except for one word).
I'm sure many vendors will tell you they don't ship to APO/FPO addresses because it's more difficult and riskier. Many resellers have standardized on UPS and/or FedEx, and it's a bit more hassle for them to ship via USPS. Many will also tell you that they run into more credit card issues--fraud, chargebacks, etc.--with APO/FPO addresses and they have little recourse when they do.
Still, if they could get their minds off their short-term bottom lines, they'd realize that mistreating our young men and women in uniform is penny-wise and pound-foolish. So they take a small hit expense-wise. Who cares? I'd think that'd be something they'd happily do to help our armed forces. It's little enough, certainly.
Which reminds me of a story I've heard from more than one WWII veteran. Both my father-in-law, who was a Marine in the Pacific, and a good friend of mine from college who landed as an Infantry sergeant at Omaha Beach and walked to Berlin have told me the same story. Apparently, the Red Cross and the Salvation Army both provided coffee and doughnuts to the guys who were coming off the front line. The difference was, the Salvation Army gave them the coffee and doughnuts. The Red Cross charged for them. No money, no doughnuts. Just off the front line and don't have any spare change? Tough shit.
Naturally, they both remembered the Salvation Army with fondness, and neither had any use for the Red Cross. Apple and other companies with military-hostile policies probably aren't even aware of the ill will they're causing. Or perhaps they don't care. Either way, I hope they reap what they've sown in the fullness of time.
We subsequently exchanged several messages, and Jerry made it clear that he didn't believe this had actually ever happened, that it was an early urban legend. As I told Jerry, I couldn't speak from personal experience--I wasn't born until 1953--but that I was merely reporting what others whom I believe to be credible had told me. I actually had it in the back of my mind that the Red Cross had been forced to charge for coffee and doughnuts, but I couldn't remember the details. Then comes this:
I probably didn't express myself well. I have nothing against the Red Cross. I admire them for what they do. I was simply using them as an example of the long-lasting resentments that occur when an organization is treating or is perceived to be treating our soldiers, sailors, or Marines unfairly. I'm sure that at the time the Red Cross executive board must have realized what a public image disaster this would be for them, so I'm surprised they went along with FDR's orders. It seems to me that a better idea might have been to provide free coffee and doughnuts to all allied servicemen regardless of nationality. And, although I do admire your even-handed policy in general, if it were me I would make an exception for Osama bin Laden or Adolph Hitler.
9:18 - It's Friday, or, as the dogs call it. Barkday. We have the recycling truck and the garbage truck today. When you live with dogs, you get to understand their vocalizations. For example, the "Barbara's home!" bark is completely distinct from anything else. Even on the rare occasions when Barbara comes home unexpectedly (unexpected by me), I always know it's her by the bark.
They even have different barks for the mailman and the UPS truck. I suspect they have different barks for UPS and FedEx, but my ear isn't sensitive enough to distinguish between the two. There are distinctly different barks for the recycling truck versus the garbage truck, although in both cases I can sense their fury. "Bob! Bob! They're stealing our stuff! If you let us out, we'll put a stop to this outrage!"
Here's something different...
Barbara was not amused. I thought it was funny. Tasteless, but funny.
Thanks. I've thought about building a mini-ITX system, but every time I consider doing it the poor bang-for-the-buck gives me pause. If one wants to build a PC into a Teddy bear or a cigar humidor, there's little alternative to mini-ITX, but for SFF and larger systems, there are a lot of better alternatives. The key thing about this Antec Aria SFF shoebox case is that it accepts microATX motherboards, which opens up a whole other world of possibilities. Instead of being stuck with the pathetic VIA processors, I can use a standard Celeron, P4, or Athlon, any of which blow the doors off the VIA. With a mid-range or slower processor, I can use a fanless heatsink like one of the Zalman units to produce a system that's almost silent, so the VIA has no advantage there. I can think of some applications for mini-ITX, for example where power consumption and heat production are key, but there just aren't that many situations where mini-ITX makes any sense to me.
15:32 - I see articles all over the place today saying that the era of $0.99 music downloads may be over. The RIAA copyright-pigs aren't satisfied. They want to increase it to $1.49 to $2.99 per song. Give me a break. They currently sell low-bitrate files, hag-ridden with DRM for about the same price per song as a CD sells for. Of course, from their point of view, they have to boost that, because very few CDs have more than one or two pieces that people would actually buy ala carte.
Everyone has his own level of pricing comfort for downloadable music. Mine is at $0.10 per song, assuming CD-quality audio, absolutely no DRM of any sort, and I have to own the music in the same way that I own the music on a CD. That is, it's mine forever. I can play it whenever and however I want, for as long as I want, as many times as I want. I can transfer it to different media. In fact, I must have all of the Fair Use rights I have with music on a CD, including the right to make personal copies and give them to friends. That's what it will take for me to consider buying downloadable music, and I'm a better prospect than most people, who simply download it for free from the Net.
FedEx showed up a while ago with two Antec Aria cases. Various on-line vendors are already advertising them for sale for $99 or so, but I know for a fact that they don't actually have any to sell. Antec USA got in a small shipment Wednesday. The two they sent me are from that shipment. None are yet in the channel. I shot a few quick images of the Aria sitting on the kitchen table. On the left is an overall view. The center image shows the front panel, with the USB, FireWire, and audio ports, as well as the bundled 8-in-1 card reader and the bezel for the 5.25" external drive bay. That one confuses me. I suspect it's designed to allow using a standard optical drive without worry about its bezel color. But the hinged bay cover on the Aria looks to me to be hinged the wrong way. Oh, well. I'll figure it out. The right-hand image shows the rear panel of the Aria, which is dominated by the huge 120mm case fan.
My first impressions are that this case is up to Antec's usual high standards. I suspect they'll sell a million of them. The retail price is $129, which means you'll eventually see them advertised for $99 or so. My guess is that they'll be in very short supply for a while.
The Aria is a bit larger than a standard Shuttle SFF system. It's 15mm taller, 69mm wider, and 35mm deeper than the standard Shuttle 185mm X 200mm X 300mm. That small increase in size buys you a lot--four PCI slots (versus one); three drive bays (versus two); a 300W power supply (versus the typical 200W or 220W), and, most important, the ability to use standard microATX motherboards. I think Antec has a real winner on its hands with the Aria.
Saturday, 10 April 2004
Sunday, 11 April 2004
Copyright © 1998, 1999, 2000, 2001, 2002, 2003, 2004 by Robert Bruce Thompson. All Rights Reserved.