|
Week of 4 October 1999 Sunday, 10 October 1999 12:42 A (mostly) daily journal of the trials, tribulations, and random observations of Robert Bruce Thompson, a writer of computer books. |
|
|
Monday, 4 October 1999 [Last Week] [Monday] [Tuesday] [Wednesday] [Thursday] [Friday] [Saturday] [Sunday] [Next Week] Gray skies this morning, with thunderstorms on the way in. The dogs won't be happy. I arrived in my office this morning to find another fatal error from Diskeeper, this time accompanied by a Dr. Watson message. The real solution to this problem would be to backup, fdisk, re-install NT, and restore. But I don't have time to do that right now, so I'll limp along as is until I have my new system built. This is all taking time that I don't have. * * * * *
Now there's an idea. I'll forward this on to Larry and see what he has to say. 18:30: I got quite a bit of writing and research done today. I ran out of gas about 5:00 p.m, which isn't surprising given that I was working on about 3 hours of sleep last night. The good news, I suppose, is that turning off logging in WinGate reduced the shutdown time on my main NT box from about 45 minutes to about 17 minutes. Or I suppose that's what caused it. Not that that's really much of an improvement. Barbara is off to the Dixie Classic Fair with her parents and sister tonight. I can't imagine why she'd want to go there, but they all seem to enjoy it. I'd planned to get more writing done tonight while she was away, but I think this'll be about it for me. * * * * *
Okay, I guess, but I'm not sure what the point is. I consider the fragmentation levels you mention to be unacceptably high. If you're suggesting that Linux filesystems are any less subject to fragmentation than NTFS filesystems, that's just not right. In fact, Linux filesystems are demonstrably inferior to NTFS in several respects. That's one of things holding back Linux adoption in server space. * * * * *
I have now, thanks. But I don't think it signifies. The Athlon is perilously close to being classifiable as a failed product, and mere product announcements don't mean much. I'd guess that Compaq and IBM are simply hedging their bets. Now, if Compaq and IBM were shipping Athlon systems in quantity, that'd be a different story. But they're not, and they probably never will. At this point, buying an Athlon looks like a sucker bet to me. Although I appreciate the people who buy AMD processors because it keeps the price of Intel processors down for the rest of us, I sure wouldn't buy one myself (or let my sister marry one). Buy a Pentium III, or wait for Coppermine. I've been wrong before, and I may be wrong this time, but I don't think so.
|
|
|
Tuesday, 5 October 1999 [Last Week] [Monday] [Tuesday] [Wednesday] [Thursday] [Friday] [Saturday] [Sunday] [Next Week] I see in the morning paper that a convicted murderer taunted the victim's family, who expected him to be sentenced to death. He was sentenced to life imprisonment instead, and proceeded to nyah-nyah the victim's family at the sentencing hearing. He then somehow escaped, no one is sure quite how, and murdered someone else. I think we should slightly modify the procedure for murder trials. All it would take is a trapdoor, a large tank of piranhas, and some minor electrical wiring. Seat the defendant on a trapdoor over the tank of piranhas. Give each juror a single-pole/single-throw switch, labeled "Guilty as Hell" in the closed position and "Not sure yet" in the open position. Wire all twelve switches in series, and set all twelve open at the start of the trial. At whatever point a juror becomes convinced that the defendant is guilty, he throws his switch to the closed, Guilty as Hell position. If and when all twelve switches are closed, a circuit is completed that drops the defendant into the piranha tank. This would also cut down on appeals. Of course, the problem would be that there really aren't all that many murder trials, so the piranhas might have to go too long between feedings. To keep them happy and healthy, I'd suggest extending this method to trials of child molesters, con-men who prey on the elderly, armed robbers, muggers, and dishonest politicians. On second thought, we'd better not use the piranhas for crooked politicians. The last thing we need is a bunch of fat, overfed piranhas. For crooked politicians, telemarketing people, and spammers, we could instead use a deep pit filled with punji stakes. 11:30: Barbara had a dentist appointment this morning, and stopped at OfficeMax on her way home. She found something that solves my nagging problem about what to do with the CDs I burn. I hate jewel cases. I despise jewel cases. They're expensive, clumsy to use, crack easily, and often scratch the CD when they break. And so I buy CD-R blanks in spindles of 100. But that begs the question of what to do with them once they're burned. Barbara found a neat little desktop CD storage tray from Laserline. It stores up to 100 CDs or DVDs in fifty double-sided sleeves. The sleeves are clear plastic with a soft backing, and can be purchased separately. The whole thing with 50 sleeves cost under $10. 12:20: This is the third post today, which must be some kind of record. On the other hand, they haven't been anywhere near as large as some of the monster posts I've done in the past. * * * * * Tom Syroid announced on his web site this morning that he'd finally found a solution to his problem. For months, he'd been unable to get USB to work under Windows 2000 on his dual-CPU EPoX motherboard. The solution turned out to be very simple. All he had to do was enter BIOS Setup and change MPS version from the default 1.4 to 1.1. I sent him the following message: Geez, I probably shouldn't admit this, but the MPS issue crossed my mind when you first mentioned the problem. I should have mentioned it then, which would have saved you months of aggravation, but you'd mentioned that you'd brought the problem to EPoX support's attention. I thought that if it was something that simple they'd surely have told you.
There's not much difference as far as I know. My understanding is that MPS 1.1 limits a machine to one PCI bus, while MPS adds extended configuration tables that allow MPS 1.4 systems to bridge multiple PCI busses, as may be required in a larger server. As far as I know, there are no performance implications on a simple dual-CPU system. * * * * *
I've never thought there was much point in worrying about which order you take tests in. A lot of people think the Networking Essentials test is the easiest one, and so take it first to "dip their toes in the water." You should definitely take workstation and server close together, because there's a lot of commonality between those two tests. In fact, I wrote a series of web-based MCSE training courses for DigitalThink that cover both exams together. Of the tests you mention, most people consider SITE the toughest, so it may make sense to leave it for last. Paul Robichaux has just finished his MCSE (and also wrote the SITE course for DigitalThink), so he may have some advice that runs counter to mine. If so, I'd be inclined to listen to him rather than to me. Also, I'd recommend that you pick up copies of Michael Moncur's two _MCSE in a Nutshell_ books. They're not intended for primary test preparation for folks who don't know NT, but many find them very useful for a final polish right before taking the exams.
|
|
|
Wednesday, 6 October 1999 [Last Week] [Monday] [Tuesday] [Wednesday] [Thursday] [Friday] [Saturday] [Sunday] [Next Week] Another cool morning. We're really into Autumn now, it seems. My favorite time of year. Daytime temperatures in the middle 70's (24C) and nights in the lower 40's (5C). I wish it were like this all year 'round. * * * * *
Well, one can hope. I really hope that AMD succeeds, but I don't think they will. Some of my readers perceive me as being "anti-AMD" or hoping that AMD will fail. Nothing could be further from the truth. As I've said before, the presence of AMD as a viable competitor to Intel is good for all of us, including ultimately Intel. Whether or not AMD can pull that off is the question. I don't think they can, but I hope they will. * * * * *
Interesting. I'd never heard of this product until I read Dvorak's column and the First Looks article on PC Magazine's web site. I agree that anyone who has an always-on connection needs to have this product or something like it. Of course, those running Linux can do something very similar without buying firewall software, and those running Windows NT Server can use the Routing & RAS upgrade to do some pretty serious filtering. For that matter, WinGate allows you to lock up things pretty tightly. But something is certainly needed. BlackIce may be a good choice, although I haven't had a chance to look at it. * * * * *
Well, probably on the theory that it's better to stomp on a small potential threat than wait until it become a big threat. As things are now, the Athlon is no real threat to Intel at all. It might be, if Athlon motherboards were available in quantity and actually worked. But that's not the case. After I got your message, I checked around to see if I could find Athlon motherboards actually for sale. I couldn't. There are a lot of people advertising them, but very few motherboards actually available to be shipped. The places I trust not to represent that they have product available for shipping that isn't really in stock all have notices up like "100 due in", "50 on order", "Backordered through 11/15" and so on. The other problem, of course, is that those Athlon motherboards that are actually shipping (in however limited quantities) are garbage. They're prototypes rather than serious products. Anand, certainly a pro-AMD journalist if there is one, tells people not to buy them. The FIC motherboard he reviewed, for example, won't even fit a standard case because the I/O template is non-standard. So, for now at least, I stand by my advice. Don't buy an Athlon. There are only two motherboard manufacturers I really trust. Intel and EPoX. Obviously, Intel isn't going to ship an Athlon motherboard, so that leaves EPoX. If and when EPoX ships an Athlon motherboard, then I'll consider changing my position. * * * * *
Hmm. I'm not sure. I don't recall ever seeing two Power Switch leads on a PPC case, which is not to say that there weren't two on cases I've used. Sometimes I figure out a situation like this and then forget all about it. I checked the PC Power & Cooling web site, and found the following list of front-panel connectors: Green/Black-Power LED which seems consistent with my memory of what was there the last time I built a computer on a PPC case (a month or so ago). I think the best solution would be to call PC Power & Cooling and ask their tech support people.
|
|
|
Thursday, 7 October 1999 [Last Week] [Monday] [Tuesday] [Wednesday] [Thursday] [Friday] [Saturday] [Sunday] [Next Week] If it isn't one thing, it's another. Shortly after Barbara went back to bed to read last night, my mother called upstairs to say that Kerry, our 12 year old Border Collie, had a nose bleed. We checked out the dog medical books, and apparently a nosebleed in a dog can be non-trivial. Barbara was concerned enough that we called our vet, Sue Stephens, at home. Fortunately, Sue is also a friend, although Barbara doesn't like to take advantage of the friendship that way. Barbara wanted to find out if it was really necessary to take him to the emergency 24 hour vet, with whom we've not had good experiences. Sue said to keep an eye on him, take him to the emergency vet if the bleeding persisted or got worse, but otherwise to just bring him in to see her tomorrow if the situation hadn't cleared up. Fortunately, he appears to be okay this morning. Barbara has a narrative with photographs of her recent trip to Mackinac Island. She shot the pictures with traditional 35 mm gear, and we spent an hour or so yesterday scanning the prints. She still has some work to do arranging images and text flow on the page, but it's good enough now to look at. I did some more experimenting last night with MP3s, ripping audio CDs (digital audio extraction or DAE), and so on. I got some interesting results. I tried ripping with both a Plextor SCSI CD-ROM drive and several ATAPI CD-ROM drives. The Plextor was much faster, of course, but that wasn't the really interesting part. Just on a hunch, I did binary comparisons on the .wav files I'd ripped against the original tracks on the CDs. The .wav files I'd ripped with the Plextor compared exactly. Those I'd ripped with the ATAPI drives failed the comparison. I confess that I can't hear any difference between files ripped with the Plextor versus those ripped with the ATAPI drives, but it appears that the Plextor drive makes exact digital duplicates, whereas the ATAPI drives do not. I also played around a bit with compressing MP3s at different levels. Native CD-DA audio CDs play back audio at 75 sectors/second. Those sectors are 2,352 bytes. In other words, a 44,100 Hz sampling rate times 16-bit samples times two channels equals 176,400 bytes/second. Dividing that by 2,352 bytes/sector yields 75 sectors/second. That means that ripping a 3 minute audio track yields a .wav file of just over 30 MB, and ripping a full 60 minute CD turns yields just over 600 MB of .wav files. MP3 compressors allow you to choose among various bit rates. Most of the MP3s you can download from the web are compressed at 64 Kb/s, which compresses the raw audio data by about 21.5 X. That bit rate compresses a 3 minute audio track down from 30 MB to only about 1.5 MB, but the quality is terrible. Most so-called high-quality MP3s are compressed at the 128 Kb/s bit rate, which yields a 3 MB file. I've been using 256 Kb/s, which yields a 6 MB file, but I'm wondering if I should be using a higher bit rate still. I don't want to rip and compress a couple of hundred CDs and find out later that 256 Kb/s wasn't fast enough. Basically, I can easily tell the difference between 64 Kb/s and 128 Kb/s even on typical computer speakers, but I can't tell the difference between 128 Kb/s and 256 Kb/s. That changes when I use a decent ($50) set of headphones. With those, I can tell a difference between 128 Kb/s and 256 Kb/s, although the difference is hard to explain. The difference is particularly noticeable when listening to something like a Bach concerto, less so with vocals, and almost indiscernable for hard rock. Disk space to store the MP3s isn't a major issue, but I'd prefer not to waste huge gobs of it. So the question is, what bit rate should I use when compressing the ripped .wav files? Would using a rate higher than 256 Kb/s be worthwhile? Should I compress different types of material at different bit rates? Bach at 320 Kb/s, for example, and Louie, Louie at 16 Kb/s? * * * * *
I assume you're referring to the fact that you sent me a couple of messages about Linux versus NT file systems that I haven't published or responded to. No, you didn't upset me. Please understand that I get something like 100 substantive email messages a day. I read all of them, but sometimes I don't have time even to acknowledge messages, let alone respond constructively or publish them. I used to respond to all of them, if only with a simple "Thanks." but that drew enough flak that I stopped doing it. Understandably, people who take the time to write a detailed message expect something more than a one-word reply, but there's only me.
19:15: This is now a GIF-free site. A pox upon Unisys. I've converted every GIF file on this site to PNG format. If you have trouble viewing PNG graphics in your browser, please see this site. If you have a web site of your own, please see the Burn All GIFs web site, and consider participating in Burn All GIFs day, Friday, 5 November, 1999.
|
|
|
Friday, 8 October 1999 [Last Week] [Monday] [Tuesday] [Wednesday] [Thursday] [Friday] [Saturday] [Sunday] [Next Week] I learn something new every day. Enabling DMA on Barbara's system, theodore, made a huge difference. According to the benchmarks, sustained disk throughput has about tripled, from 5 MB/s to 15 MB/s. And that difference appears to be confirmed subjectively. I copy and move a lot of big files--WAV files, Pournelle's 8 MB daily web stats, etc.--from theodore, which is our main network data store, to kerby, my main system, for processing. With DMA enabled, copying/moving those files seems to take about one third the time it did when theodore was running PIO. But that wasn't good enough for me. I want theodore to be a real server. I'd decided from the first that it'd run SCSI, but I didn't have the SCSI host adapter or the drive in hand when I built theodore, so I installed a 7,200 RPM ATA Maxtor DiamondMax Plus as a temporary measure. I now have a Seagate Barracuda 18LP 7,200 RPM U2W SCSI drive and an Adaptec 2930U2 host adapter for theodore. When I opened the host adapter package and read the documents, I noticed that they said the 2930U2 was optimized for systems that had only one hard drive. The 2940U2W that I have sitting here ready to go into my new system says that it's designed for use in systems with multiple hard drives. I wasn't sure why that was, so I sent the following message to Adaptec: Here's what I hope is an easy technical question: I'm reading the documents for the 2930U2, and I notice that they say that this host adapter "was designed to perform best in a single hard disk drive, multiple non-hard disk drive configuration. If you're connecting more than one hard disk drive, we recommend that you use either the SCSI Card 2940U2W or the SCSI Card 3950U2" I understand that one shouldn't install non-U2W devices on the U2W bus, but I don't understand why the U2W bus on the 2940U2W is able to support multiple LVD drives and the 2930U2 isn't. Aren't they the same U2W bus, or does the 2940U2W have a better SCSI controller or something? The system I'm building around the 2930U2 will initially have a Seagate Barracuda 18LP LVD drive, which I believe has sustained throughput of about 22.5 MB/s. Will that single drive saturate the bus on the 2930U2, or would I be safe in adding a second LVD drive later?
Thanks for your fast and detailed response. I was aware of the SCSI disconnect feature, but not of the performance implications. Your crystal-clear explanation just taught me something about SCSI. * * * * *
I agree. I wouldn't consider buying an Athlon for the next three to six months. By that time, there may be usable, debugged motherboards available. The trouble for AMD is that anyone with any sense is going to wait a while before buying an Athlon. That means two things. First AMD isn't going to be selling many Athlons (they sold something like 100,000 last quarter, probably about what Intel sells in a *day*). Second, this period when no one is going to be buying Athlons is, unfortunately for AMD, their window of opportunity. By the time it's "safe" to buy an Athlon, Intel will have moved on and AMD will no longer be competing against the now-current crop of Intel processors. * * * * *
I'll defer to your obviously greater experience. I certainly didn't make a scientific survey of the MP3s out there on the web, so perhaps it was simply coincidence that most of what I found was 64 Kb/s. I'm not sure what lossless compression you're referring to. I notice that .wav files, which are an exact representation of the digital audio stream stored on a CD, don't compress at all to speak of with ZIP or similar compression schemes. Is there some version of MP3 that uses lossless versus lossy compression, or are you talking about another method entirely? * * * * *
Yes, I do think that the Athlon is a good processor, from what I know of it. My doubts have nothing to do with the processor itself. As I've said repeated in these pages and elsewhere, AMD has always designed good processors. Their problems historically have been (a) their inability to execute by delivering product in volume in a timely manner, (b) their inability to get third-party manufacturers to support their products properly (e.g. the slow rampup of Super Socket 7 and now Slot A motherboards), and (c) their inability to market their way out of a paper bag. When I say that the Athlon is in danger of becoming a failed product, I'm not talking about technical issues. I'm talking about the fact that by the time Athlon motherboards are available in volume, Intel may have trumped the Athlon ace with new products of their own. "Too little, too late" has been the sad story of AMD for several years now, and I'm afraid the same will happen with the Athlon. See this article in The Register as an example. As far as motherboard availability, NECx is one of the sources I checked. When I looked, they had no Athlon motherboards listed as in stock. They showed "on order" as the status. * * * * *
Thanks. But I don't think that's what he was talking about, or at least I hope not. For an AT power supply, the four wires you refer to are power leads rather than signal leads. They should be heavy gauge wire and terminate in spade lugs which connect to the power supply, as opposed to the signal leads, which are light gauge wire and terminate in header pin connectors which connect to the motherboard. There's no reason that the power switch can't have both sorts of leads, but if that's the case, it should have six wires coming out of it: four heavy wires with spade lugs (blue, white, black, and brown); and a two-wire blue/white cable terminated with a header pin connector. I have a PPC Solid-Steel Tower sitting open on my kitchen table right now. I just looked and I don't see any wires that terminate in spade lugs coming out of the power switch. Perhaps the Personal Mid-Tower is as you describe. I have more than one of those around here, but it's not convenient to pop the lid on any of them right now. * * * * *
I'm like you. I'm not sure what they're planning to burn. I took it metaphorically anyway. But I've removed all GIFs from my site, and don't plan to use one ever again. In converting from GIF to PNG, I noticed that none of the PNGs were larger than the GIFs they replaced. In some cases, particularly on very small images, the PNGs were about the same size. Most of the time, there was a noticeable shrinkage from GIF to PNG, e.g. 15 KB down to 12 KB. In several cases, the shrinkage was dramatic. For example, the copy of my tartan shrunk from 20 KB in GIF format to 2 KB in PNG. Converting everything with IrfanView took me less than an hour. A lot of webmasters won't take the time to do the conversion, but I'd sure encourage anyone who has the time to do it. I don't like being threatened, particularly when I haven't done anything wrong.
|
|
|
Saturday, 9 October 1999 [Last Week] [Monday] [Tuesday] [Wednesday] [Thursday] [Friday] [Saturday] [Sunday] [Next Week] Although the leaves aren't really changing color much yet, they are starting to fall. And there's rain forecast for this afternoon and tomorrow. That means that, oh joy, I get to climb up on the roof and blow out the gutters this morning. Every year, Barbara suggests we get gutter caps installed, and every year I put off doing it. I don't much like them, but perhaps one year soon I'll decide to go ahead and get them. Climbing up on the roof every week during leaf season to do gutter maintenance isn't as much fun as it used to be. Back to building my new system this afternoon. I could probably build the whole thing in an afternoon, but I need to photograph and document the process step-by-step. That means it'll take several days to get it done, but there really isn't any hurry to complete it. * * * * *
Well, in theory, it *should* be possible to use lossless compression on music and get some results. There are, after all, quiet passages, periods when only one instrument is playing, and other points where the full bandwidth of the constant-rate data stream is not being used. But I fear that the percentage compression would be relatively low, particularly on complex music like classical. * * * * *
Well. I'm glad to hear that Unisys lost a big one. I'm hoping that this whole GIF mess will turn into a firestorm that results in the government scrapping the whole idea of software patents. That whole idea is bizarre anyway. Consider some of the "innovations" that have been deemed worthy of software patents: using a Save-As function to allow saving a file under a different name; using colors to highlight important information in a list; saving data to a volume located on a remote machine. I am not making any of this up. Incidentally, this whole thing made me remember a discovery I made a decade or so ago when I was writing an RFP. I was using WordPerfect for DOS (5.1 as I recall) and ran a spell check on the document. It didn't recognize "Unisys". As a replacement, it suggested "anuses". How prophetic. * * * * *
Hmm. The only thing I can think that it might be for is a system with dual motherboards. I seem to remember that PC Power & Cooling makes (or made) cases designed to contain two motherboards, although I've never used one.
|
|
|
Sunday, 10 October 1999 [Last Week] [Monday] [Tuesday] [Wednesday] [Thursday] [Friday] [Saturday] [Sunday] [Next Week] Barbara let me sleep in this morning. I didn't wake up until almost 9:00. As much as I enjoyed sleeping in, that means I'm behind. I still have to do today's update, process web stats for my own site and for Pournelle's, read over some material Pournelle sent me, set up next week's page, etc. etc. I also have to do laundry, and I'd like to get some time to work on my new system. I'm getting too old for this seven days a week stuff... * * * * *
Thanks. I wasn't aware of that program. The 40% compression makes sense. I wouldn't expect any lossless compressor to as much as halve the size of CD-DA data. As far as quality, you confirm my impressions. I may go with 320 Kb/s instead, but I suspect I wouldn't be able to tell much difference between 256 and 320. * * * * *
I must admit I'm confused. The CD-DA data stream is 75 sectors/second, with each sector being 2,352 bytes (or 176,400 bytes/sec). That converts to 635,040,000 bytes/hour. Compressing that at 4:1 yields 158,760,000 bytes per hour, rather than 800 MB/hr. A 256 Kb/s MP3 file contains 32,768 bytes/sec or 117,964,800 bytes/hour, which amounts to 5.38:1 compression, presumably the 5.3:1 rate you mention. The 4:1 compression would record 44,100 bytes/second, which corresponds to an MP3 compressed at about 344.5 Kb/s. So I guess it might make sense to compress at 320 Kb/s. I don't think I have any rate higher than that. Thanks for the information. * * * * *
I hope you had the good sense not to buy one. The FIC motherboards are the ones that Anand says have non-standard I/O templates that don't fit cases. Also, I know someone just published a story yesterday saying that the FIC motherboards were incompatible with the Athlon running at some speed (which FIC confirms). I think it may have been 650 MHz rather than 600, but I'm not sure. At any rate, the fact remains that there are no usable motherboards for the Athlon, and won't be until after the first of the year. Don't take my word for it. Read Anand or any of the other pro-Athlon folks. They'll tell you the same thing. By the time all the parts are in place to build real Athlon-based systems, AMD is likely to have missed their window of opportunity.
[Last Week] [Monday] [Tuesday] [Wednesday] [Thursday] [Friday] [Saturday] [Sunday] [Next Week] |