|
Week of 3/8/99 Friday, July 05, 2002 08:16 A (mostly) daily journal of the trials, tribulations, and random observations of Robert Bruce Thompson, a writer of computer books. |
|
Monday |
[Monday] [Tuesday] [Wednesday] [Thursday] [Friday] [Saturday] [Sunday] Monday, March 8, 1999 I absolutely have to get more work done on my book. Actually, it's not the amount of work that's the problem, it's the amount of output. I'd originally hoped to get the first draft 100% complete by the end of February, and certainly by the end of March. I'm working very hard on the book, but here it is the second week in March and I haven't reached 50% complete yet. Hardware books are much slower to write than software books. I don't know what to do to make it go faster. I guess it's going to take as long as it takes. But I'd better get back to work on it now. More here later, perhaps. * * * * * This from Chuck Waggoner [waggoner@gis.net]:
All of this is disturbing to me. As is the "we didn't know we were doing this, but now that you've pointed it out, we'll stop doing it" routine. As a matter of policy, I *never* register software, just because I don't want the software vendors having that information about me. But now I wonder what's been going on behind the scenes. When I visit the Microsoft web site, or simply when I'm connected to the Internet, are Microsoft applications sending information about me to Microsoft without my knowing that they're doing so? I don't think I have any paranoid tendencies, but this has to make me wonder. * * * * * This from The Moores [themoores@the-i.net]:
Yep. That just confirms the NY Times article. I think it would only be fitting if someone tracked down and published the home addresses, telephone numbers, credit card numbers, credit histories, tax returns, etc. of all the Microsoft and Intel executives. * * * * * This from Bruce Denman [bdenman@ftc-inet] [http://web.infoave.net/~bdenman]
As far as Tom and Anand, I've never meant to imply that they were intentionally biased, merely that accepting advertising money raises at least the appearance of impropriety. I'm sure they're both very nice guys, although I've never met either of them. In fact, I just read something on Tom's Hardware the other day that showed a distinct lack of bias. To give you an idea of the amounts of money involved, a video card manufacturer was offering up to $15,000/month to major web sites that would promote their product. Tom "outed" them by publishing the memo that contained the offer, and they were pretty upset about it. I have had hardware vendors offer me payment to review their products, although never on anything near that scale. I always tell them no. I'll take free eval units--that's standard practice--but I won't take money from them for reviewing their products. They may dress it up as de minimis payment to help offset the time and costs involved in reviewing their product, but I look upon such payments as bribes, pure and simple. As far as choosing someone to host your web site, I really can't make any recommendations. I'm happy with pair Networks, but that's based on my requirements (which may be very different from yours) and very little actual experience. I was happy with BigBiz, too, until their service started going down hill. But I will say that you should be able to find a satisfactory web hosting company with setup charges of $0 to $25 or so, and monthly charges in the $10 to $30 range. The numbers you quote are outrageous for the level of services provided, but are typical for ISPs that offer web hosting. Better to go with a company that specializes in web hosting. There's no reason to tie the two things--ISP versus web hosting service. They're completely separate. A month or two back, I mentioned several web sites that concern themselves with choosing a web hosting service, rating web hosting services, etc. You can probably find that reference using the search function. Alternatively, visit the web site for Peter Kent's excellent book, Poor Richard's Web Site. If you don't have that book, get a copy before you make any decisions about starting your own full-blown site. You can pick it up at your local bookstore, or from Amazon.com by clicking here.
|
|
Tuesday |
Tuesday, March 9, 1999 We were supposed to have a couple inches of ice and snow when we woke up this morning, but it hasn't shown up yet. The Weather Channel says there's a 100% chance of a winter storm today, so we'll see. * * * * * Bo Leuf mentions Internet faxing on his
site this morning. It's something we've talked about in the past, so
he solicted my comments. He mentions the eFax
service, but it appears to me to be just another fax gateway service.
The sender still has to send the fax using existing dial-up fax protocols.
What I want is for IETF to define a standard for IP faxing. This would
allow a sender to address a fax to an IP address or hostname (ideal for
recipient companies who have full-time Internet connectivity) or to an
email address (ideal for individual recipients, either at home or at their
desks at work). * * * * * This from Scott Kitterman kitterma@erols.com:
Yes, I'm a big advocate of APC UPSs. I've been buying them for probably a decade, and have bought hundreds of them for myself, my clients, and my former employers. In all that time, I remember getting only one or two units that didn't function perfectly right out of the box. When I needed a new UPS last summer, I decided I might as well try something else, just to see what was out there. I ended up buying a TrippLite 675 unit, which blew up the first time I connected it to mains power. I mean that literally. It made a pop loud enough to hurt my ears, and sparks came shooting out the back. Tripp was very good about replacing the unit immediately, but that could have been a very bad situation if I'd connected the UPS to the equipment before connecting it to the power receptacle. I bought the Tripp because it was about 10% or so cheaper than the equivalent APC model. Don't get me wrong, the Tripp is not a bad unit by any means. But the specs aren't quite as good as the APC, and the APC gives the impression of better construction quality, at least to my eyes. There are really only three things that matter about a UPS: battery, inverter, and switching circuitry. APC uses high-grade components for all of these. Many of their competitors use components that aren't quite as good. Paying 10% or 15% more for the APC gets you a better piece of equipment, one that's likely to last longer and is less likely to fail when you need it. And the fact that APC batteries are user-replaceable is a major benefit. If you'd bought something other than an APC, you'd probably have had to ship the UPS back to the factory to get the battery replaced. In my opinion, that more than makes up for the small price premium that APC units sell for. In fact, you could have kept a spare battery on hand, eliminating down-time, which is not an option with most of APC's competitors. The Back-UPS line is decent low-end power protection, but I recommend that people step up to the Back-UPS Pro if at all possible. Actually, the ones I prefer are the Smart-UPS, which provide true sine-wave output in the larger sizes. APC positions the Smart-UPS for protecting servers, but I find them useful for ordinary client protection as well. I have a Smart-UPS 1000 under my desk right now, with three computers attached to it. Anyone who uses his computer for work rather than just play is crazy if he doesn't have a UPS on it. Nowadays, I just flat-out tell people, "Buy APC."
|
|
Wednesday |
Wednesday, March 10, 1999 Here's an interesting web site I just stumbled across this morning. BookCloseOuts.com specializes in remaindered books. They've even got a special section of $2.99 computer books, although I'm not sure how much value an old computer book would retain. I still like Edward R. Hamilton, Bookseller, from whom I've purchased remaindered books for years, but these new guys appear to be worth a look if you like buying bargain books. I may go over there and buy a dozen or two mysteries. * * * * * I've begun to wonder about the long-term viability of AMD. I saw in the paper yesterday that AMD is cutting 300 jobs, and that they seem unable to deliver K6-III processors in numbers even remotely close to meeting demand. And they're up against formidable competition from the Celeron, on both price and performance. As is obvious from the model name, AMD is trying to position the K6-III against the Pentium III. But the truth is that, at least in terms of published performance benchmarks, the K6-III actually competes against the Celeron. They're roughly equivalent in integer operations, and the Celeron blows the doors off the AMD at floating point. The one advantage AMD has is 3DNow!, but few applications support it, and it seems likely that most software manufacturers will focus their attention on supporting Intel's SSE first and 3DNow! only as an afterthought, if at all. And the other major concern is chipset support. If the K6-III is to succeed, it must be a simple drop-in replacement in existing Super7 motherboards. And yet, I've seen reports from several sources that the K6-III doesn't function properly in motherboards that use the VIA MVP-3 chipset. If that's true, it's a disaster for AMD. When I checked the AMD approved motherboard list this morning, I found that only two motherboards (one GigaByte model and one ASUS model) are certified for the K6-III/450, and both of them use the ALI Aladdin V chipset. There are a handful of motherboards certified for the K6-III/400, all but one of which use the Aladdin V. The AOpen AX-59Pro uses the VIA MVP-3, and is certified at 400. So apparently, reports that state that the K6-III doesn't work in MVP-3 motherboards are wrong. The question is, is the AX-59Pro simply an exception to a general rule? All of this can't be good news for AMD. They're trying to sell a chip for $400 to $500 that is outperformed in most applications by a $100 to $150 Intel Celeron. They can't deliver the chip in reasonable numbers, and there are questions about its compatibility with existing MVP3 Super7 motherboards. I may be wrong--I often have been--but I sniff a disaster in the making here. My guess is that AMD will eventually solve their yield problems and be able to deliver the K6-III in reasonable numbers. But I suspect they'll soon be forced to cut prices down to the level of the Celeron, say $100 or thereabouts for the K6-III/400 and perhaps $125 or $150 for the 450 MHz part. I'm supposed to have some K6-2 and K6-III processors coming in, so I'll do my own tests. I don't trust what some others have had to say about them. There are AMD-lovers/Intel-haters and AMD-haters/Intel-lovers out there, and it's not always immediately clear to me which camp, if either, a person who publishes benchmarks falls into. It's easy enough to diddle benchmarks, so I'll see for myself. * * * * * This from Robert Morgan [robert@morgan.pair.com]:
Exactly. As a matter of fact, I was over on Anand's page reading that review when your mail came in. Kryotech seems to be specializing in making pointless components, and Anand seems to be specializing in praising them with faint damns. As far as I can see (and I'm going on what Anand says, because I've never seen any Kryotech products), this kludge accomplishes exactly nothing that a $50 Peltier cooler in a decently ventilated case wouldn't. Anand seems preoccupied with the fact that a Peltier device exhausts heat into the case. But that's what case fans are for, to remove warm air from the case and replace it with air at ambient temperature. This appears to be yet another expensive, pointless product from Kryotech. * * * * * And Robert Morgan [robert@morgan.pair.com] sends this follow-up:
I didn't even notice the typos. Hell, I don't even worry much about my own typos. I do sometimes go back and fix a typo that jumps out at me, but I tend not to edit readers' mail much. If I'd noticed those, I probably would have fixed them, though. As far as astronomy, I did have some interest in it back when I was a teenager. Enough to grind my own mirror. I'm still interested, but I don't have the time for it. That and the fact that I live in Winston-Salem, which has the light pollution you'd expect of any decent size city. Although I'll admit I was tempted when I noticed someone on the net selling 18" pyrex blank/tool pairs that had been pre-ground to f/8 spherical. That'd sure short-cut the process of grinding a big mirror. Of course, I'd probably want to take it down to 1/10 wave or something, which'd take me years.... Congratulations on your new scope. I envy you. * * * * * Mid-afternoon: FedEx showed up this morning with a box from Intel that contained some very interesting items, including a PPGA Celeron/400 and a BI440ZX Socket 370 mainboard to install it in. Slot 1 Celerons should by rights have killed Socket 7. They damaged it badly, but they didn't sink it, thanks mainly to AMD's Super7 initiative. But Socket 370 really should be the final nail in the Socket 7 coffin. Intel positions this Socket 370 board as a low-end solution, and it appears to me that it and other Socket 370 mainboards should dominate the low-end segment. We're talking about $100 mainboards and $60 to $125 processors here. They're low-end in terms of cost, but they provide performance at least comparable to (and usually better than) the mid- and high-end Socket 7 stuff. All of this gives the AMD K6-* and the Cyrix MII a real problem. I suspect that's why AMD will abandon Socket 7 for Slot A with its K7, and Cyrix will abandon Socket 7 for Slot 1 with its Jalapeno. One way or another, Socket 7 really is finally dead. There will still be a lot of Socket 7 CPUs selling into the upgrade market for a while, but my guess is that Socket 7 motherboard sales are probably drying up even now. And, yes, I know that Gateway recently jumped onto the Super7 bandwagon. I think that's a decision they'll rapidly regret. My guess is that Super7 will be a very short-term fling for them. The box also contained a Pentium III, which I plan to test exhaustively once I have some SSE-enabled applications available.
|
|
Thursday |
Thursday, March 11, 1999 I've gotten several messages concerning my comments on AMD. I won't bother to print them, because they all say pretty much the same thing. That I'm "unfair" to AMD, "don't like" AMD, or am "biased in favor of Intel." None of that is true. I've requested some AMD processors, and once they arrive I'll give them a fair evaluation. I have an open mind. As I said earlier, I don't trust a lot of the stuff that's been posted on the web about AMD, pro and con, so I'll do my own testing and form my own opinions. AMD versus Intel seems to be a religious issue. If it's not at the same level of acrimony as NT versus Linux, it at least comes close at times. So, for the record, here are my thoughts on the matter: 1. I like both AMD and Intel. Everyone I've spoken with at both companies is professional, and they all seem to be nice people. 2. I'm glad that AMD exists. It is because they exist (and to a lesser extent because companies like Cyrix and IDT/WinChip exist) that Intel ever introduced the Celeron. If it wasn't for the competition provided by AMD, Cyrix, and IDT, Intel would probably have never felt the need to ship such low-cost, high-performance CPUs. Competition is a Good Thing. 3. I think a great many of AMD's current problems exist because they feel forced to bring products to market faster than they otherwise would to counter Intel. If Intel didn't exist, AMD would not have shipped the K6-III so soon. They'd have taken another six months or whatever to solve their yield problems before even announcing the K6-III. Of course, if Intel didn't exist, the K6-III would never have been developed. Again, competition is a Good Thing. 4. Many of AMD's problems arise because they are married to the Socket 7 platform. Think what you will, but Socket 7 is an inferior technology relative to Slot 1. There's only so much that can be done within the framework of the obsolescent Socket 7. AMD has done a superb job of pushing Socket 7 to its maximum potential. But even AMD recognizes that Socket 7 is not viable in the long term, as their shift to Slot A with the K7 indicates. If Slot 1 versions of the K6-2 and K6-III existed, they would provide serious competition to Intel CPUs. But with Socket 7, AMD is fighting with one hand tied behind its back. 5. For now at least, I buy Intel processors for myself and recommend them to others because they offer both the highest performance and the best bang for the buck. But I keep an open mind. If AMD sends me some samples, I'll evaluate them on their own merits and report what I learn, good or bad. * * * * * This from Dave Farquhar [farquhar@freewwweb.com]:
I'd never thought of the power issue, but I suppose you're right. I seem to recall that the refrigerator has a separate power cord, so I suppose you could plug it and the PC into different receptacles, but that's a good point just the same. I don't know anything about Anand's economic situation. I believe his father is a college professor, so I'd guess his family is probably middle-class to upper-middle class. I think he runs his servers locally on an ADSL line, but that's just my impression. He's generating enough traffic to his site that I'd guess his ad revenue is more than enough to pay his costs now. As far as running components cold, I don't think that's a problem. Electronic components are happier at lower temperatures. Most disk drives are rated to run within pretty broad ranges. I think 32F/0C to 104F/40C is a pretty common range. And I agree that Anand needs to be asking Kryotech some hard questions. But he is just 17 years old, after all. * * * * * This from Robert Morgan [robert@morgan.pair.com]:
Slot 1 is a proprietary Intel technology, and I don't know all the issues surrounding it. Given my personal philosophy, I obviously think that Intel is entitled to do as they please with Slot 1. They invented it. It's theirs. But in practical terms, I think AMD could probably license Slot 1 in exchange for a royalty. That's pretty common practice in the industry, and I know that Cyrix plans a Slot 1 and/or Socket 370 version of their Jalapeno CPU, presumably under license from Intel. I'm glad that you and millions of others choose to buy AMD processors. Without that market, AMD could not provide credible competition to Intel. And you're right about the performance issues. All modern processors are so fast that it's just a matter of degree. In fact, I recently swapped the Pentium II/450 in my test-bed system for a Celeron/333. I'm sure the system is demonstrably slower now, but the truth is that I can't tell much difference subjectively when running my normal applications. Same thing with AMD versus Intel. The AMD processors have excellent integer performance, and a given AMD-based system may in fact be faster in integer performance than a given Intel system. For those few that use floating point heavily, the Intel processors are noticeably faster. As far as the technical issues for Socket 7 versus Slot 1, I'm not a CPU engineer, but I've been told by people who should know that Socket 7 is inferior electrically at high speeds. Running a 100 MHz FSB pushes Socket 7 to its limits. But both Slot 1 and Socket 370 support GTL, which Socket 7 does not. GTL as implemented currently can easily support FSB speeds of 133 MHz or more. You're correct that the K6-III brings L2 cache onto the processor itself, eliminating the CPU-memory bus bottleneck that crimps the performance of other Socket 7 processors. But the real drawback to Socket 7 at this point is that Intel ceased developing chipsets for it. The 430TX is their latest Socket 7 chipset, and it doesn't incorporate many modern features, such as AGP. VIA, ALI, and others produce full-featured Socket 7 chipsets, but their performance and stability is a step down from Intel's best. Even the Socket 7 fanatics admit that Socket 7 motherboards are less stable than Slot 1 motherboards. And you can't blame Intel for abandoning Socket 7. I'm sure that part of their reason was to shift the market towards their proprietary Slot 1 motherboards, chipsets, and CPUs. But that's only part of the reason. The real reason that Intel introduced Slot 1 was that they'd about reached the limits of what it was practical to do with Socket 7. * * * * * The following from someone who requests anonymity:
I read that book right after it first hit print, which must be at least twenty years ago. As a matter of fact, I'm pretty sure I have the hardback first edition still around here somewhere. Your message motivated me to go over and check on Amazon.com. I was expecting to find that the book was long out of print, but I find that there's now a 25th edition dated March, 1998. Although I haven't re-read the book in many years, I remember thinking highly enough of it that I can recommend it without reservation. If you want to buy a copy, click here to order it from Amazon.com. * * * * * This from Adam Lieber [aslieber@Princeton.EDU]:
Yes, I believe that's correct, although it's by no means certain that Cyrix will ship the Jalapeno as a Slot 1 CPU. I understand that they are considering licensing Socket 370 to avoid the high component costs involved in producing Slot 1 parts.
|
|
Friday |
Friday, March 12, 1999 And another sign of Y2K madness is that the confidence men and scam artists have embraced it. These crooks telephone their victims, usually elderly people, and convince them that the account their money is currently in is not Y2K-compliant. To make sure their money is safe, they have to transfer it into another account which is Y2K-compliant. That account, of course, belongs to the scammer. I think the death penalty is about right for the scum who prey on elderly people like this. Or at least keelhauling. Laws that protect children exist because children are presumed to be innocent and easily duped. But many elderly people are just as trusting. Swindling an elderly person out of the money he needs to put food on the table is inexcusable, or should be. * * * * * This from Robert Morgan [robert@morgan.pair.com]:
Well, I'm not sure what to make of your statements. Open standards are fine, but are you saying that you are entitled to force Intel to give up their proprietary, patented technology simply because you feel strongly that it should be an open standard? That doesn't sound reasonable to me. In fact, it sounds like armed robbery. I don't know whether or not Intel would agree to license Slot 1 to AMD, or indeed whether AMD has even explored that option. But even if AMD wants to produce a Slot 1 processor and Intel refuses to grant them a license, so what? Slot 1 is Intel's intellectual property, and they should not be compelled to license it if they don't want to. It's not like there aren't alternatives. AMD is perfectly free to create their own competing standard (as they are apparently doing with Slot A and the K7), and to keep that standard proprietary or to make it open. I mean, AMD may be David to Intel's Goliath, but AMD is nevertheless a huge company that sells millions of processors. They single-handedly ramrodded the Super7 initiative through, and Super7 is now broadly supported. Personally, I think it was a mistake to focus on Socket 7 instead of spending the same effort on a better technology, but that was AMD's decision. And I'm not sure how the fact that Slot 1 is proprietary has harmed anyone. There are more processor choices than ever before, they're cheaper than ever, and it doesn't look like anything is slowing down innovation in processor architectures. How has Slot 1 hurt anything at all? * * * * * This followup from Robert Morgan [robert@morgan.pair.com]:
Well, other than by force majeur, I don't see any way that Intel could be compelled to give up its property. As far as erecting barriers, that's what businesses do. All businesses. That's the whole point behind copyright, patent, and trademark law. The idea is that a person or a company who devotes resources to developing a marketable idea should be be granted a monopoly of limited term to enjoy revenue from that innovation. Without that protection, there is no economic incentive to innovate. The fundamental purpose of Intel, AMD, Microsoft, or any other business is to create revenues and profits. Why would they not do everything they can to protect their innovations and their turf? I think there are indeed many more processor choices than there used to be. Five years ago, NT ran on Intel, MIPS, Alpha, and PowerPC chips (I'm not sure what your reference to NEC means.) NT still runs on all four of those platforms. It is true that Microsoft will not support the MIPS or PowerPC chips in NT5, but that's a market decision. No one (almost literally) was buying NT for those platforms. And I'm not sure what those chips have to do with anything. They're certainly not mainstream processors by anyone's definition. But look at the mainstream chips. Ten years ago, you had one choice, the 386. Five years ago, you had two choices, the 486 or the Pentium. That was it. You ran on an Intel architecture, whether that processor was made by Intel or another company like Harris or AMD. Nowadays, you can choose among many variants of the Intel P6 processors, the AMD K6, the Cyrix MediaGX or MII, the IDT WinChip, and a couple of other new players which haven't yet made their marks. These are all different processors, not just second-sourced processors made from Intel masks. And if you don't remember the Pentium at $1,000, I certainly do. Everyone introduces their newest high-end chips at the highest price they can charge to take advantage of early adopters. That's just the nature of the business. When Intel ships the Merced or AMD ships the K7, you can bet they'll be $1,000+ processors (probably $3,000 in the case of Merced). That helps pay off the huge up-front costs that they incurred to design the chip and build the fab. And Slot 1 has indeed done things, if indirectly, to make processors both cheaper and faster. Even disregarding the lower value of the dollar now versus a few years ago, today's processors are much cheaper. Today, you can go out and buy a Celeron for $70 that has literally 90% of the power of Intel's fastest and most expensive desktop processor. Five years ago, $70 wouldn't have bought you nearly that much in relative terms, let alone absolute terms. * * * * * This from Dave Farquhar [farquhar@freewwweb.com]:
I agree with your observations on Anand, but what it really boils down to is that Anand runs a hobby site. Running a web site isn't exactly cheap, but it's not really that expensive, either. At least in terms of providing the required infrastructure via a web hosting service. Before professional web hosting services like pair Networks became available, it cost several thousand dollars a month to host a real web site, most of that for connectivity and equipment. Nowadays, you can host a pretty serious site with a lot of traffic for $25 or $50 a month. Something on the scale of Tom's Hardware, which recently passed 1,000,000 page reads/day, probably costs several thousand dollars a month to host. But that's with several co-located web servers at pair. Just as an example of what can be done with a fairly inexpensive web hosting account, in terms of disk storage and traffic allowance, Pournelle's site would probably fit in the pair Networks Advanced category, which costs less than $20/month and provides 40 MB of disk space and 200 MB of daily traffic. If not, he'd certainly fit in the Webmaster category, which costs about $30/month, and provides 80 MB of disk space and 400 MB of daily traffic. * * * * * Barbara found a place that gets my vote for the strange web site of the week, and maybe the year: You have to have audio and animation enabled to get the full effect.
|
|
Saturday |
Saturday, March 13, 1999 Yesterday evening Pournelle and I got down to serious work on our book. I sent Jerry a first-draft chapter. He went through the first five pages or so, annotating it, making adds, changes, and deletions. Like any co-authors who've never worked together before, the first thing on the agenda is developing a working relationship that suits both of us. I suspect that won't be too hard. We're both pretty flexible, and our writing styles seem reasonably compatible. Barbara looked over Jerry's comments and changes, and agreed with all of them. I did too, come to that. This morning, Jerry sent me another copy of the chapter with more work completed. Barbara and I laughed out loud at some of the war stories he'd added. Pournelle's perspective is priceless. This is going to be a great book. People are going to use it as a reference, certainly, but a lot of people will sit down and read it cover-to-cover like a novel. Pournelle is that good. * * * * * I see mention on Tom Syroid's page this morning that Microsoft may be porting Office to Linux. That seems bizarre to me. Office is a major cash cow for Microsoft, so it is understandable that they want to sell as many copies as possible. But Office is also the crown jewel that locks people into Windows. Two major factors prevent Linux from becoming a mainstream desktop operating system. One, obviously, is that Linux apparently attended the Muggers' School of User Friendliness. An average user can't even format a floppy disk under Linux, let alone do something as complicated as connecting it to the Internet. But that's relatively easy to address via a graphic shell and wizard-like scripts, and there are any number of those in development right now. The real barrier to widespread adoption of Linux for the desktop is the dearth of mainstream productivity applications--read Office--available for it. Microsoft would be foolish to make Office available for Linux. They'll not sell many copies. The mindset among Linux users is that software should be free, and I suspect a Linux version of Office would be pirated very heavily. Right now, Microsoft makes most of its Office revenue from bundling deals with computer vendors, and that's obviously not a factor with Linux. So it seems to me that if Microsoft wants to reduce sales of both NT and Office, the best way to do it is to port Office to Linux. Somehow, I don't think that's going to happen. * * * * * Bo Leuf talks about the joys of dealing with the Swedish tax authority on his site this morning. He requested a minor change in his tax status, and they sent him a form. Some time later he got another copy of the same form. They keep sending him the same form over and over, and can't explain why that's happening. But it's not just governments that do this. I had a similar experience
almost 15 years ago when I was working at an Entré Computer Center.
Ashton-Tate had some kind of promotion for computer store employees. You
had to read their literature about the newly released dBASE III Plus and
then take a simple quiz about product features. You then faxed your
answers to AT, and they were supposed to send you a free live copy of
dBASE III Plus. At the time it cost something like $400, so it was worth
spending a few minutes to get one's own copy. They had a similar deal for
a long-forgotten integrated product called Framework, which competed with
Lotus's long-forgotten Symphony. * * * * * This from Bruce Denman bdenman@ftc-i.net [http://web.infoave.net/~bdenman]:
Immediately followed by:
Well, despite the fact that Tom posted a TNT2 update today, you have a good point. I think the last time his site was updated in any significant way was on February 22. Oh, if you do a refresh every morning, you'll find that the date changes to today's date, but the only change you usually see is a couple of lame "news" items. As far as I'm concerned, Tom's Hardware is like the old joke about soap operas. You can visit it once a month and not miss much. And then:
Well, the link may not have been there the first time you visited. I seem to remember that it wasn't there the first time I checked. It appeared later in the day. * * * * * This from Dave Farquhar [farquhar@freewwweb.com]:
There are a lot of things that can cause spontaneous reboots, but by far the most common are memory problems and power supply problems. All chips prefer cooler temperatures to warmer ones, and higher voltages to lower ones (within limits, of course). Although an overheating CPU can certainly cause reboots, I doubt that's what's happening here. Of course, it's worth checking to make sure that ventilation isn't blocked, the CPU fan is running, etc. before trying anything else, but I think the most likely cause is the power supply. Cyrix chips are used most often in inexpensive systems, and those systems tend to have cheap power supplies. There might be a couple things related to the power supply that could be causing the problem: First, of course, it may simply be failing. That's not surprising by any means, because many mass-market systems built on Cyrix chips use the cheapest power supplies available. Another possibility is that it simply can't supply adequate amperage at one or more voltages. You can't simply look at the gross wattage rating to determine this, either. You have to look at the individual amperage ratings for different voltages, which many cheap power supplies don't even list. Even if you can find this information, it may not do you much good, because you also need to know the amperage required by each component at each voltage. But if your friend has added an expansion card or other peripheral, it may be that the combined load now exceeds the ability of the power supply on one or more voltages. You can do a field expedient check on this by removing all cards and disconnecting all drives that aren't absolutely necessary and running the system without them. If it runs without rebooting, chances are that you simply have too much installed for the power supply to support. Before you swap out the power supply, however, it'd be worth checking the other likely cause. If you have a known-good memory module, try swapping it in and seeing if the problem goes away. If it does, the memory may be failing. Generally, memory lasts forever, but it's possible for a power spike to damage memory without destroying it completely. Note that a damaged memory module may function properly when plenty of power is available, but begin to act hinky when the power supply is stressed. * * * * * And Dave then had this to say:
Umm. I think I see your problem. There's no way that any 235 watt power supply, let alone an inexpensive one, is going to handle three hard drives, a CD-ROM, a floppy, a cage full of expansion cards, a fan or two and whatever else you have connected. That poor power supply sounds like it's running at 100% (or more) of its capacity. I try to run power supplies at no more than 60% of their rated capacity, and preferably less. They run cooler, provide cleaner power, and last a lot longer. When I built my current main system, I was going to buy the PC Power & Cooling 235 watt unit. When I told Larry Aldridge that I might run as many as three hard drives, he told me he didn't recommend their 235 watt unit for that application. I ended up with a TurboCool 300. Try disconnecting a couple of the hard drives and pulling as many cards as you can. I'll bet the spontaneous reboots stop. Then again, with all that stuff in the case (I presume it's a mini-tower), you may indeed have a cooling problem.
|
|
Sunday |
Sunday, March 14, 1999 I see in the paper this morning that Crayola is changing the name of one of their crayon colors for only the third time ever. They changed the name of Prussian Blue in 1958, apparently because most kids didn't know where Prussia was. In 1962, they changed the name Flesh to Peach, belatedly acknowledging that not everyone is Caucasian. Now they're going to rename the color currently called Indian Red. Apparently, some people are upset, believing that this crayon color refers to the skin color of Native American Indians. According to Crayola, the color was named after a dye produced on the Indian subcontinent of Asia. But they don't want to offend anyone, so they're holding a contest to rename the color. I've already suggested Politically Correct Red, so that one is taken. * * * * * Before I forget, does anyone know a URL or other contact information for the company that makes the ATX test-bed chassis that was mentioned on Pournelle's site some months ago? He can't remember it, and I can't find it. This is a specialized testing chassis that allows swapping motherboards, drives, and other components in and out easily. I'd like to get one of these, but I've been unable to locate any information about it. * * * * * Internet Explorer 5 is about to drive me mad. I have to keep reminding myself that it is just a beta, and a pretty good looking beta at that. But the one thing that I simply can't work around is the way it handles multiple instances. While I'm writing, I frequently have multiple web pages--sometimes as many as a dozen--active in separate instances of IE. When I want to do another search or hit another web page, I fire up a new instance. The problem is that IE insists on loading the new web page on one of the older instances, replacing what I wanted to keep there. This doesn't happen if I actually type the URL into the box. It happens only when I use a Favorite to get to the new site. For example, if I want to call up a new instance of NorthernLight, I can start a new instance of IE5, type in www.northernlight.com, press Enter, and everything is fine. Northern Light loads in the new instance of IE5. But if I call up a new instance of IE5 and click the Northern Light favorite menu entry, the new instance of Northern Light comes up in an older instance of IE5. ARGGGHHHH. Same thing if I start a New Window instead of double-clicking IE on my desktop. The new window comes up with contents duplicating the old Window. If I then click on a favorite, it loads, but IN THE ORIGINAL WINDOW RATHER THAN THE NEW ONE. But playing around with IE5 got me to thinking about using its Subscription feature, which is now called Off-line Browsing. I believe that IE4 had pretty much the same feature, but I never used it because I never did figure out what subscriptions were. I remember thinking that they must have something to do with Channels and push technology, which I wasn't even slightly interested in. But the name change to Off-line Browsing made it clear to me what this feature was all about, so I tried it. I played around with it a bit on Friday, but yesterday I got serious about it. I decided to set up Chaos Manor, Tom's Hardware, and AnandTech for off-line browsing. After the downloads finished, all 115 MB of them, I had complete copies of those three web sites on my local hard disk. I told the off-line browsing process to go three levels deep on each site, which is the greatest depth it allows. So I'm not sure if it got everything on all of the sites, but it got quite a lot. Just out of curiosity, after the scheduled updates completed, I decided to re-run them immediately to see how much (or how little) would be required to update the sites immediately after I'd just done so. I learned a couple of interesting things. The first very interesting thing is that if you want your site to be effectively unusable by off-line browsers, you should use dynamic content rather than static content. Pournelle's content, for example, is completely static. His server doesn't generate pages on the fly. The original update totalled about 20 MB from his site. When I turned around and re-ran an update immediately, it took only a minute or two to complete, and transferred only about 20 KB worth of data. Obviously, what's happening here is that IE5 is loading and parsing the site tree, checking file names and timestamps (and presumably file sizes) against some sort of hashed database that it maintains locally. I was more concerned about what would happen when I tried updating Tom's Hardware and AnandTech. Rightly so, as it turned out. AnandTech is essentially all dynamic content. His server actually generates each page on the fly and sends it to your browser. As I suspected, when I started the re-update process for AnandTech, it started to reload every single page it had just loaded an hour or less before. I feared the same would happen with Tom's Hardware, but it didn't, or at least not completely. Apparently, much of Tom's Hardware actual content is static. But he has a lot of links to banner ads and so forth, which are generated/rotated on the fly. The upshot was, an immediate re-update of Tom's Hardware took about 15 minutes and resulted in downloading about a megabyte, most of that useless banner ad data. What I'd hoped to do was set up a scheduled update for each site that would suck down only the new information every morning about 6:00 a.m. That'd work fine for Jerry's site, marginally well for Tom's Hardware, and not at all for AnandTech. So much for dynamic versus static content. I guess the fact that dynamic content is essentially unusable for off-line browsing might be considered a Good Thing from some people's point of view (particularly banner advertisers, who'd be paying for a lot of click-throughs that no one would ever actually see). But it's a Bad Thing from my point of view as a user. * * * * * My friend Steve Tucker recently brought up his own web site, Wakeolda.com. We were sitting around talking the other night about how best to drive traffic to a new site. That's important to me as well, and I suppose that I really should be doing more to increase the traffic to this site. At any rate, I told Steve that the key thing is unique content. A bunch of family pictures and a set of links isn't going to attract anyone to speak of. Steve's in a unique position to create content for which there is great demand. He works in Sports Marketing Enterprises for R. J. Reynolds. Steve used to be responsible for the NASCAR Winston Cup auto racing series. A couple of years ago, he moved to a position where he deals not just with Winston Cup, but with all of the RJR sports marketing programs, including drag racing, motorcycle racing, and so on. For years, Steve has been on a first-name basis with all of the folks involved, from drivers to owners to sponsors. Steve and Suzy are long-time friends of NASCAR Royalty, people like the Pettys, Earnhardts, Gordons, Waltrips, and Jarretts. I told Steve that a lot of people would love to read what he has to say about things from his perspective on the inside. He's started a journal, and my guess is that he'll become as addicted to writing it as the rest of us are to writing ours. If you have any interest at all in auto racing, check out Steve's page. He's just now getting started, but my guess is that it'll develop into something fascinating. * * * * * This from Roger G. Smith [rgsmith@c-gate.net]:
Thanks for the kind words. I was just thinking the other night how things work out. In 1979, I was helping my brother move from one apartment to another. When we arrived at the new apartment, it was completely empty with one exception. There was a copy of Lucifer's Hammer lying in the middle of the living room floor. I'd never heard of Pournelle. Being an inveterate reader, once we got everything hauled up the stairs and into the apartment, I flopped down and started reading the book. I was delighted to have found a "new" author that I enjoyed reading so much. I remember thinking--lying there on the carpet using a cardboard box as a pillow--that I would like to someday write a book with this man. Even then, I wanted to write for a living. I don't know why I fixed on Pournelle rather than Niven. The unique name, perhaps. I'd never read anything by either of them, so they were both just names to me. So here it is twenty years later and I'm writing a book with Pournelle. Not fiction as I imagined then, true, but a book nonetheless. * * * * * This from Joshua Boyd [jdboyd@cs.millersv.edu] http://catpro.dragonfire.net/joshua
Good points, and I actually used to use emacs. But that's not really an alternative for me any more. When I was young, I used to laugh at people whose cars had automatic transmissions and air conditioning. What wimps. I drove a Jeep CJ. Nowadays, I've joined the automatic transmission/air conditioning crowd. But I have nothing but respect for people like you, who grab Linux, wrestle it to the ground, and make it do what they want it to. But I just don't have the time to deal with it, much though I'd like to. I can still sit down in front of a Linux system and function at some minimal level, but nowadays cat, ls, and pwd are more my speed than recompiling a kernel. * * * * * Another from Joshua Boyd [jdboyd@cs.millersv.edu] http://catpro.dragonfire.net/joshua
Well, yes, refrigeration has always been used to improve CPU performance. I remember the old Cray *MP series. They were basically enveloped by a refrigeration unit. And you're correct that software licensed on a per CPU basis could make refrigeration a cost-effective solution. But it's not a PC technology. Of course, the PIII serial number may yet allow PC software manufacturers to license by CPU type, speed, and number. That's what worries me about the PIII serialization more than anything. I can't really speak to gaming chips, because I'm not a gamer. However, I'd think that an overclocked $70 Celeron 300/A would give you a lot more bang for the buck than a $500 K6-III. As far as gaming, AMD lives and dies on 3DNow!, and I understand that they're not having a lot of success getting game software manufacturers to include it. My guess is that the next generation of Celerons will be based on the PIII core, will include SSE, and will be well supported by game software. But again, this isn't my area of expertise. And I agree about SMP. For my purposes, at least, a dual Pentium II/300 box blows the doors off a single-CPU Pentium II/450 or any other single processor box you care to name. But that's not always true. Windows NT itself does a pretty good job of spreading the load across multiple processors, but if you're running a single program that is not multi-threaded with SMP in mind, it'll run faster on a fast single processor machine than on an SMP NT box. As far as RAID, take a look at the Promise FastTrak IDE RAID controller. I'm doing a long-term evaluation of one right now, and will have a report about it posted Real Soon Now. I'm currently running four 10 GB Maxtor DiamondMax Plus 2500 drives in a RAID 0+1 arrangement. It's wicked fast, and all data is fully mirrored for safety. And at less than $1,000 for the controller and all four drives, it beats traditional RAID solutions all hollow on price. As far as Ms. Miller, I posted her URL because after she savaged Pournelle's site I thought some people might get a giggle from seeing what she considers to be a well-designed site. * * * * * This from Bo Leuf [bo@leuf.com]:
You may well be right. I confess that I don't know as much about Linux and Open Source issues as I probably should. But I suspect that Microsoft would have a very difficult time co-opting Linux. In the past, they've had an identifiable target to attack. Linux is too diffuse.
|