|
Week of 3/22/99 Friday, July 05, 2002 08:05 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 22, 1999 If you didn't read the updates last weekend, check back to last week. I posted quite a lot of interesting new stuff Saturday and Sunday. * * * * * I see that Pournelle's back, and upset again about the Web Pages That Suck discussion about his site. That's my fault. I stopped by that page the other day to see what new garbage had been posted. The discussion had devolved from what had been largely an attack on Pournelle's site to an attack on Pournelle himself. I probably should have let that sleeping dog lie, but I mailed Jerry to tell him what had been going on over there since the last flare-up. He wrote an essay last night about the situation, which can be found here until later today, when the new week's view will change the address to this URL. That's why I keep two copies of my current page, one as thisweek.html and the second as the named weekly file, in this case 0322RTDN.html. That way, when I need to insert a link, I can point to the latter file, which won't change. The sad part is that I agree with some of the substantive criticisms they made about Pournelle's site. In fact, I've been telling him some of the same things for months. But because of the way these criticisms were presented, as attacks rather than helpful suggestions, Pournelle has dug in his heels, and I don't blame him. For example, they hate his email blimp. I'm not crazy about it, either. But because of their attack on it, Pournelle is determined to keep it. One of things they dinged him for was for having the nerve to put a link "Click here to send email" under the blimp. That one's my fault, too. A few months back during a private email exchange, Jerry expressed surprise that anyone wouldn't realize that clicking the blimp would send him email. I told him that I'd never realized that that was what it was for. With all the other graphics--exploding computer, lava lamp, grendel, etc.--scattered around his site seemingly at random, I just figured the blimp was another one of those. So he added the label and got savaged for it. I'd actually considered offering my own site for criticism, but after reading the comments I realized that I wouldn't give much weight to any criticism most of those people might make. But I am always interested in hearing how my site can be improved, and I do try to incorporate suggestions that make sense, so please don't hesitate to criticize. * * * * * I'd always wondered about the commodity versus name-brand memory question. Commodity memory sellers say that their memory is just as good as name-brand memory, meets all specs, and is cheaper. Name-brand memory sellers say that their memory is higher quality and that using it rather than commodity memory contributes to a more stable system. I decided to find out the truth. I already had a bunch of commodity memory sitting around, so I called Crucial Technology and asked for samples of their PC100 memory. Crucial is a division of Micron, and is a true memory manufacturer. That is, Crucial/Micron actually manufactures the memory chips and the printed circuit boards themselves, and then assembles them into DIMMs. One step down the food chain are memory assemblers like Kingston. They purchase the RAM chips from one of the few companies that actually make them, and then assemble them into finished product. Still further down the chain are memory resellers, who purchase memory modules on the open market, presumably test them to ensure that they meet specifications, and then put their name on the modules. At the bottom of the chain are the nameless companies who simply buy memory modules on the spot market and remarket them as unbranded commodity memory, without making any attempt at quality control. At any rate, I wanted to see how name-brand memory like Crucial would compare with some of the no-name PC100 memory I have sitting around here, so I spent some time this week-end playing with memory. The first problem was coming up with a reasonable testing method. If you're evaluating processors or hard disks, you can test performance simply by running standard benchmarks like WinBench. That's not possible with memory. It either works or it doesn't. If a memory module is fast enough to work at a given memory bus speed, replacing it with a faster memory module doesn't make the system run any faster. And all of the memory I tested, commodity and Crucial, worked fine at a 100 MHz FSB. So the obvious next step was to run the run the memory faster than its rated speed. I won't give all the technical details, list the motherboard and processor I used, etc. because this was a very unscientific first cut at testing the memory. At 103 MHz, all the memory continued to function properly. At 112 MHz, I started getting random lockups with the commodity memory, but the Crucial memory continued to run without problems. At 124 MHz (!), the commodity memory didn't work at all. The Crucial memory had frequent lockups, but was clearly more stable than the commodity memory. At 124 MHz, I was pushing the processor, the system board, and the chipset themselves pretty hard, so I can't say whether the lockups were caused by the Crucial memory or by other components. I tend to suspect the latter. You might be thinking that this data proves that commodity memory is just as good as name-brand, so long as you run it at its rated speed. After all, I'm not likely to run a production system with a 112 MHz FSB, let alone 124 MHz, and neither are you. But there's another issue here. All other things being equal, memory chips are happier at lower temperatures and higher voltages (within their specs, of course). But a heavily loaded system generates a lot of heat, and filling out all your expansion slots and drive bays may drop the voltage delivered to the memory modules to the lower end of nominal (or even below) particularly if you're using a cheap power supply. If you're using marginal memory, the combination of higher temperatures and lower voltage can stress the ability of that memory to function reliably. Using name-brand memory gives you an extra margin of safety that can make the difference between a system that runs rock-solid and one that experiences random lockups. So my conclusion, based upon this admittedly partial and unscientifically derived data, is that buying name-brand memory is well worth the small additional cost. Pournelle once noted on his web site that he always uses name-brand memory at least until he has a new system stable, but then he sometimes swaps the name-brand memory out for commodity memory. I'll go further than that. Based on what I've seen, I intend to run only Crucial memory in all of my systems. I'll keep commodity memory on hand for testing purposes, but I won't run it in any system that matters. Highly Recommended. * * * * * This from Tom Syroid. The "today" he refers to is actually yesterday:
Hmm. I just did a Find for any file or folder that matched "identi*" and found nothing. As you're running Office 2000 and I'm running Office 97, I can only conclude that the identities folder is created by the Office 2000 install rather than IE5 install. Thanks for the kind words about my essay. * * * * * This from Bo Leuf:
Hmm. Well, perhaps that folder is created only if IE5 is installed on a system that already has Office 2000 installed. Or perhaps I installed it differently than Tom and you did. If you installed directly from the Internet, that may explain it. I ended up completely removing IE5 from my system on Saturday. My friend Steve Tucker had sucked down a complete set of installation files from CNet and burned them onto a CD. I installed directly from that CD without connecting to the Internet. * * * * * This from Shawn Wallbridge [swallbridge@home.com]:
Thanks. For the sake of all Canadian computer users, I hope they decide not to apply that tax to CD-R media. I can't think of a better way to kill a good technology. * * * * * More server problems. I attempted to publish the above material to my web server this morning and got the following message:
Whatever that means. Obviously some necessary part of FrontPage Server Extensions is not running on the web server. When I clicked OK on the above message, FrontPage Explorer displayed the Web Publishing Wizard dialog, which uses ftp to transfer files directly to the web server. I didn't want to do that, because I was afraid it would screw things up. But upon reflection, I decided to give it a try. After I entered my ftp server address, target directory, username and password, FrontPage just sat there telling me it was attempting to verify the connection to the ftp server. After several minutes I gave up and cancelled the process. The problem may be because I'm running through a proxy server rather than with a direct dialup connection on this machine. Accessing ftp sites through IE works fine, but the proxy server settings are configured in IE Internet Options. I've configured FrontPage for access via proxy for normal publishing operations, but I see no way to configure ftp proxy access for the Web Publishing Wizard. I was hoping it was use the existing FrontPage settings for proxy access, but apparently it doesn't. So I can't publish until pair Networks fixes the problem. When I encountered this error, I immediately fired off email to urgent@pair.com, but my experience to date is that it takes pair Networks several hours to a day or so to respond to urgent mail. It's now been almost an hour and a half since I sent the urgent message, and I haven't even gotten the auto-reply message that confirms they've received my message. That got me to worrying about my mail, so I sent a test message to myself. It's been 15 minutes now, and it still hasn't arrived. I telnet'd over to port 110 on mail.ttgnet.com, logged in, and found no messages waiting. Hmm. Ordinarily, I wouldn't even think about a problem with mail, but weird things have been happening. I have the account barbara@ttgnet.com set up to autoforward to her POP account on Bellsouth.net. Most of the time, that works fine. But I've gotten several messages in my inbox that were addressed to Barbara. This happens sporadically, and with no apparent rhyme or reason. She'll have a batch of mail come in from one of the mailing lists she subscribes to. All of the messages are delivered properly in her mailbox except one or two in the middle that end up in my mailbox. Something serious is going on on Wawrra. In addition to the fact that the FrontPage Extensions are apparently not running, mail is hosed as well. I never received the standard autoreply message to the message I sent to urgent@pair.com an hour and a half ago about the FrontPage problems, and test messages I've sent to my own address are not showing up. I've sent messages to thompson@ttgnet.com, ttgnet@ttgnet.com, and ttgnet@wawrra.pair.com. None of them have been delivered, but I haven't gotten bounces on any of them. I had two friends send email to me, and their's hasn't shown up either. I telnet'd to port 110 on mail.ttgnet.com and everything appears to be normal except that there are no messages in the mailbox. My guess is that the SMTP daemon on Wawrra is down. I'm beginning to feel as though I've jumped from the frying pan into the fire. As aggravating as the billing error and non-responsiveness of pair's billing department was, at least I could console myself that the service was fine from a technical perspective. But I'm beginning to wonder about that as well. Oh, great. Now I can't get to the main pair Networks server. A refresh times out. I'm able to get to any number of other sites on the Internet, so the problem is not at my end. Argggghhhh. * * * * * Okay, I tried doing a ping and tracert on www.pair.com. The ping timed out several times. By the time I did the tracert, www.pair.com was back up, and the ping also worked. In the meantime, I also did pings and tracerts on several other servers, all of which worked fine, so there is apparently some problem at pair. As of 11:38 a.m. I got one of the mail messages I'd sent myself at 10:54 a.m. A second one that I'd also sent at 10:54 a.m. showed up at 11:41 a.m. And at 11:50 a.m. I received the autoreply from urgent@pair.com that was sent at 11:01 a.m. in response to my second message, which I sent at 11:00 a.m. So urgent@pair.com has received the messages in a timely manner. I keep checking the system notices page on their web site to see if anything is being done. Nothing so far. Enough of this. I've wasted an entire morning (yet again). Why can't things simply work the way they're supposed to work? I don't have the time to mess with this any more right now. I'll try publishing one more time, just in case they've fixed it without bothering to tell me, and then give it up as a bad job until later today or tomorrow. I can't express how disgusted I am. It's enough to make me seriously consider giving up running a web site and just write my books. * * * * * Afternoon: That time, the publishing succeeded with nothing out of the ordinary. pair Networks tech support replied to my messages to say that they were having some problems on the server wawrra, where my web site resides. They finally rebooted the server and everything appears to be back to normal. I now have mail coming in in dribs and drabs, so of which was sent as early as 8:00 a.m. But all of this has cost me more than half a day. I can't afford much more of this... * * * * * This from Mike Boyle mboyle@toltbbs.com:
Thanks for the data point. I'm not sure that anyone outside Microsoft knows what's really going on. Come to that, I'm not sure anyone inside Microsoft knows either.
|
|
Tuesday |
Tuesday, March 23, 1999 O'Reilly & Associates has created a unique relationship with its customers. I can think of no other publisher whose customers buy their books because of the publisher's name rather than the author's. People may go into a bookstore looking for the latest Clancy or Grisham, or the newest edition of John Levine's Internet for Dummies, but no one goes into a bookstore looking for the latest book from a particular publisher. No one, that is, except O'Reilly customers. You can see evidence of this in the Amazon.com "Customers who bought this book also bought" section under any O'Reilly title. Generally, those customers bought only other O'Reilly books. To see what I mean, check out my latest O'Reilly book, Windows NT TCP/IP Network Administration. The other books listed are all by O'Reilly. Click on one of those other books and you'll find the same is true for it. People who bought it bought mostly (or all) other O'Reilly titles. Then pick another computer book at random. You'll find no such publisher loyalty. In response to this extraordinary phenomenon, many brick-and-mortar bookstores have established special O'Reilly sections or kiosks where loyal O'Reilly customers can browse the latest O'Reilly titles. And now Amazon.com has done the same thing by creating a special O'Reilly section in their virtual bookstore. You can check it out at The O'Reilly Bookstore @ Amazon.com. I suspect a lot of Amazon computer book customers will bookmark this page and never even look at titles from other publishers. From my point of view, that's a Good Thing. * * * * * Regarding the discussion about Microsoft porting Office to Linux, it appears that won't be happening any time soon. The source of the story was comments made by O'Reilly author Simson Garfinkel during a radio talk-show interview. Mr. Garfinkel comments further in this Boston Globe article. * * * * * This from Tim Werth timothy.werth@eds.com:
I've not used the ASUS P2B motherboard, although I've heard numerous good reports about it. And the Intel SE440BX2 "Seattle 2" is an excellent motherboard, which I do have experience with. The only downside to Intel motherboards is that, as you might expect, they don't make provision for overclocking. However, if what you want is a solid, well-built, stable motherboard, Intel is an excellent choice. They don't provide direct end-user support for their motherboards, but that hardly matters. Their web site is so good and so comprehensive that I can't imagine anyone needing help that can't be found there. And, yes, I am a big proponent of PC Power & Cooling cases and power supplies. But I'm not sure that buying one of their foundation systems is a good idea. I note that the Turbo Cool 300 ATX power supply costs $109, and the Personal Mid-Tower case costs $69, for a total of $178. I think the auxiliary fan is included, but if not that increases the price by $12 to $190. That means that that SE440BX2 motherboard is costing you $179, which is not a very good deal. The SE440BX2 is available in versions with or without embedded audio. The website doesn't mention audio, so I'll assume that price is for the motherboard without it. You can find SE440BX2 motherboards on the web for $125 or less. NECx, for example, lists the no-audio version at $121 and the audio version at $135. So you'd be paying a pretty heavy price premium to have them install the board for you. Of course, a lot of people would think it worth $50 to get everything from one place and not have to mess with installing the board. As far as the case and power supply, I'd go with the PC Power & Cooling. Comparing their $70 case and $109 power supply to the typical no-name $70 case/power supply combination, you'll find that there's a world of difference in the power supply. The PC Power & Cooling case is nicer, too. Everything fits, sharp edges have been deburred, and so on. If money is a major factor, I'd recommend buying at least the PC Power & Cooling power supply, even if you stick it in a cheap case. And that takes us back to the motherboard. There's some doubt in my mind as to where Intel is going with Slot 1. They're pushing Socket 370 heavily now, to the extent that Celerons will soon be available only in Socket 370 form. It may be that Intel is simply trying to differentiate their product line, i.e. "Socket 370 = Low-end" and "Slot 1 = high-end", but it may also be that Intel is tending toward Socket 370 for all of their processors. There's no reason they can't ship the Pentium II and/or Pentium III in Socket 370 form, and it wouldn't surprise me to see them do so. Slot 1 parts, motherboards and processors, are expensive to produce. It may well be that Socket 370 is Intel's future emphasis. If so, that has to concern anyone who's committed to Slot 1 motherboards. At this point, there's not a lot of choice in Socket 370 motherboards. Intel makes a nice low-end product, the BI440ZX. I have a sample of it, along with a Socket 370 Celeron/400, and intend to put it to a long-term evaluation. EPoX also makes some Socket 370 motherboards, and I may well request eval units of them. But right now, Socket 370 is pretty much limited to the low end. If I wanted to buy a motherboard right now, I'm not sure what I'd go with. Probably Slot 1, but it's a tough call. * * * * * This from HARTMAN, JOE (JSC-CC) [joe.hartman1@jsc.nasa.gov]:
Well that's truly strange. I just hit my web site, clicked on this week from the home page, and did a refresh. The current version of the page popped right up. My best guess is that you may be behind some sort of proxy server that's caching requests. If you're not running one locally, perhaps your ISP is doing the caching. I'll go ahead and post your message and this response to see if anyone else has ever experienced a similar problem. And thanks for the kind words.
|
|
Wednesday |
Wednesday, March 24, 1999 If you hurry, you can be the proud owner of a pre-pack of 24 copies of Bill Gates' new book, Business at the Speed of Thought. The retail price is $720, but with the 30% discount you can grab one of these collectors items for only $504. Amazon had this listed until early this morning, when they apparently pulled the item. Barnes & Noble still has it, at least for now, so click here if you're interested. The only thing that isn't clear to me is whether this pack contains 24 printed copies of the book, or just one printed copy with a licence that allows 24 people to read it. * * * * * This from Dave Farquhar [farquhar@lcms.org]:
Good points. I know that the "slocket" adapters are now available from many manufacturers for $20 or so, and I suppose they're a reasonable way to go. My only reason for considering Socket 370 right now is that Intel's high-bang-for-the-buck CPUs are the Celerons. They've said that the Celeron/433 and /466 will ship only in Socket 370 form, and starting this summer they plan to discontinue Slot 1 Celerons entirely. So combining the extra $10 or so that a Slot 1 board costs relative to an identical Socket 370 version with the $20 or so cost of a slocket adapter means that you'll pay $30 or so for having the added flexibility of a Slot 1 motherboard. For some people, that'd be the difference between one version of the Celeron and the next faster one, or between a marginal amount of memory and an adequate amount. And you're also correct that using a dual Slot 1 board with two PPGA Celerons and two slocket adapters is an easy way to build a dual-CPU system on the cheap. I haven't tried it, simply because my audience is largely business users and computer-literate home users rather than geeks. And I don't use that term disparagingly. I consider myself to have some strong geek-like tendencies. So perhaps I should get a couple of Celeron/300 PPGA CPUs and a couple of slockets and build myself an overclocked 450/504 MHz dual-Celeron system. I have a couple of dual-CPU Slot 1 boards around I could use.
|
|
Thursday |
Thursday, March 25, 1999 The headline in the morning paper says "NATO bombs Yugoslavia." True, I suppose, but on a par with "Axis bombs Pearl Harbor." I suppose I should be grateful that the US government is not yet comfortable enough with its imperial actions to take full credit for them. And speaking of Pearl Harbor, when the Japanese bombed it without a prior declaration of war, it was "a date which will live in infamy." Why is it then that when the US bombs Yugoslavia without a prior declaration of war it is perceived by nearly all US citizens and by many others worldwide as an admirable and necessary action, however regrettable? I'm about the furthest thing imaginable from a pacifist. "Speak softly and carry a big stick" pretty much sums up my personal beliefs. I do speak softly, and I carry a 45 caliber stick. So I don't object to these imperialist actions because I have any finer sensitivities about the futility of violence. Violence settled Hitler's and Tojo's hash, thank you very much, and the threat of violence settled the Seventy Years War. But there is a difference between defending oneself against attack and attacking someone else because you don't like what he is doing. What is going on in Yugoslavia, reprehensible though it may be, is not the legitimate concern of the United States. If indeed action needs to be taken, it should be taken by the European powers, and then only after a formal declaration of war. Like a vampire, the US currently enters only when invited, and I think Europe may later regret extending the invitation. And, although I do not condone the actions of those currently in power in Yugoslavia, I at least understand something of why they are occurring. If you are not familiar with what the Chetniks did during WWII, read something about it. Suffice it to say that the savagery of the Chetniks often made hardened SS men vomit, literally. Wholesale crucifixions, for example, were one of the Chetnik's favorite activities. What goes around comes around, and all we can do by dropping bombs is shift the equilibrium. People have been slaughtering other people in the Balkans since at least Julius Caesar's time, and nothing we do is likely to change that. But we will kill a lot of people. * * * * * There's been some discussion over on Pournelle's site about page widths. He does things pretty much the same way I do. A narrow left column with links and the remainder of the screen width devoted to text. A lot of people have suggested that he go with the three-column format used by a lot of sites--a left column with links, a relatively narrow middle column that contains the actual content, and a right column that's essentially blank space. Various people have suggested that the narrow-text format is for readability, but I don't think so. Here's the message I sent to Pournelle yesterday about it: As interesting as the discussion of font sizes versus line
lengths is, I think it misses the real reason for the narrow pages on many
web sites. Neither of us much cares about "page request" counts
on our sites because we're not trying to earn advertising revenue. But
sites that are have a vested interest in keeping their pages narrow and
short for two reasons: * * * * * This from Lee Mandell [Lee_Mandell@iceinc.com]:
Thanks for the kind words. Perhaps I'll have to buy a couple of PPGA Celeron 300/A CPUs. I can't very well call up Intel and ask them to send me a couple more processors so that I can test ways to get around their restrictions on using the Celerons in a dual-CPU environment. Actually, things are starting to get easy enough that anyone should be able to run a dual-Celeron machine. With the Slot 1 Celerons, one literally had to drill holes in the processor package to make the modification. With the current generation of slocket adapters, one must use wire and tape or nail polish on the slocket to connect things that aren't connected but need to be and to disconnect things that are connected but need not to be. But at least that's easier than drilling holes in the CPU and using a wire-wrap tool or soldering iron. My sources tell me that a new generation of slocket adapters is due soon. They'll provide jumpers to allow you to configure the slocket to support dual Celerons without doing any fancy jiggery-pokery. At that point, building a dual Celeron system should be very straight-forward, and Intel is going to have a hard time stuffing the genie back into the bottle. One thing that may still be a problem is BIOS support. I know of several Slot 1 "Pentium II" boards that do not support the Celeron, usually because the BIOS expects the 512 KB L2 cache on a Pentium II rather than the 0 KB or 128 KB L2 cache on a Celeron. Still, motherboard manufacturers are no dummies, and if this slocket-based dual Celeron idea becomes commonplace, they're likely to modify the BIOSs on their dual-CPU boards to support the Celerons. And speaking of L2 cache, that brings up an interesting question. In typical desktop environments, the 512 KB of half-CPU-speed L2 cache on a Pentium II and the 128 KB of full-CPU-speed L2 cache on a Celeron is about a wash. The smaller, faster Celeron L2 cache is faster for some things, and the larger, slower Pentium II L2 cache for others. But it's nearly always close to a wash, with one or the other showing only a few percentage points of advantage. But the situation with a dual-CPU machine may be very different. A lot of applications where a dual-CPU box makes sense move a lot of data, and the larger Pentium II cache may show significant performance benefits in those applications. So I guess I've talked myself into waiting for a while before I do anything about this. My guess is that in three months at most, building a dual-Celeron box will be a matter of buying some off-the-shelf components and setting some jumpers. My only concern is how much longer the Celeron 300/A is going to be available. If you're looking for maximum bang for the buck, the Celeron 300/A is the way to get it. Intel has never built a processor as overclockable as the Celeron 300/A, and is never likely to again. * * * * * Here's an example of technology not working the way it's supposed to. I periodically check The Register, a UK computer news web site. But their server is often slow, and I tired of waiting for stories to download. So a couple of weeks ago, I signed up for their email notification service. Supposedly, they would mail me a digest once a day with abstracts of and links to current articles. I've been getting the messages sporadically, so I didn't think much of it when one showed up in my inbox just now. But I have Outlook set to preview mode, and the "news" that was listed seemed pretty old. So I checked the SMTP headers of the message. Here they are, with the relevant items bolded:
Ugh. The message was dated Tuesday a week ago, March 16, and they just got around to delivering it this morning. Even the US Postal Service usually does better than that. Also note the To: address. That's one of the advantages of having your own domain. You can create email accounts on the fly each time you sign up for something like this. If you get spammed later, you know exactly where it came from, and can configure your mail service to quietly dispose of anything sent to that address. * * * * * Here's an interesting commentary on prescription drug costs. Barbara just stopped at the drug store to pick up a prescription. When Barbara quit her job at the end of last year, our health insurance coverage under her group policy at work disappeared, and we signed up with Blue Cross/Blue Shield, which charges a $10 copay per prescription for generics and a $20 copay for name-brand. When she stopped at the counter to pick up the prescription, they rang it up for $99 and change. She pointed out to them that they hadn't run it through her insurance. They did so, and charged her $20. But what was really interesting was that the receipt shows the amount actually charged to BC/BS. Instead of $99, BC/BS got a total price of $56.57. Pretty interesting, huh? If you're an individual, you pay full price. If you're BC/BS, you get more than 40% discount. Something about that doesn't seem right. * * * * * And on the Celeron overclocking/dual CPU subject, I've been exchanging mail with Paul Robichaux [paul@robichaux.net], another O'Reilly author. He's been using a loaner server from Dell, but only diamonds are forever. The conversation started thus:
I think highly of Intel and EPoX motherboards. Are you literally wanting to build a "real server" (dual processors, embedded SCSI, ECC, etc.) or just a competent box that happens to run NT Server? If, as I suspect, it's the latter, do you want to make provision for dual processors, build the least expensive high-quality box possible, or what? If I were building a system today, I'd probably pick one of the following: 1. Integrated motherboard; high quality; mid-range video & sound -- EPoX EP-BXT (Slot 1, embedded Yamaha business audio and Intel i740 graphics, about $150 street). 2. Integrated motherboard, top quality; high (but not cutting edge) video & sound -- Intel RC400BX (Slot 1, embedded Soundblaster PCI and nVIDIA RIVA graphics, Pentium III support, $200 street, hard to find) 3. Inexpensive but high quality system -- Intel BI440ZX (PPGA (Socket 370) Celeron support, embedded Soundblaster PCI audio, no embedded video, $100 street). I have samples of all of the above, and they're good boards. As far as dual-CPU boards, Intel makes them, but they're very expensive ($500+). I have a sample dual Slot 1 EPoX KP6BS, but I haven't had a chance to test it yet.
EPoX actually makes at least two dual Slot 1 cards. The BXBS has embedded SCSI, which ain't cheap. The K6BS doesn't, and is reasonably priced for a dual-CPU motherboard. I know for a fact that the Promise FastTrak IDE RAID controller works fine in the EP-BXT, because I'm running it that way now with two Maxtor DiamondMax Plus 2500 10 GB drives attached to it. I've played with all combinations up to four drives. Everything works, with one exception. I still haven't gotten the thing to boot from the array. There's no problem with the way I have it set up now--an old 4.3 GB WD IDE as the boot drive, connected as Master on the motherboard's embedded Primary ATA interface. I talked to Promise. They say they've never heard of that problem, and I think they're telling the truth. It may be some sort of weird interaction between BIOSs or something. If you want to leave open the possibility of dual CPUs, check my web page today. Before long, you should be able to install dual Socket 370 Celerons without much trouble. If you run overclocked Celeron 300/A CPUs, that'd give you dual 450 MHz PII class processors for less money overall than one 450 MHz PII. Alternatively, you can play it straight and not overclock by using Celeron 400 or Celeron 433 CPUs. Heck even non-overclocked dual Celeron 300/A's are going to be as fast or faster with NT as a single 450/500 PII/PIII.
* * * * * So it appears that there is some substantial level of interest in this activity. I've also had other messages from readers who are interested in running overclocked dual-Celeron systems. I'll put a note on my (already overloaded) to-do list to try to get something done about this.
|
|
Friday |
Friday, March 26, 1999 I'd better put on my asbestos underwear, because I'll be flame-bait today. I sent a message to Pournelle in which I said some unnice things about Netscape Navigator, and observed that Internet Explorer is a much superior product. That means I'll be getting flame mail all day. It'll divide into two major categories, the "how dare you say bad things about Netscape?" messages, and the "how dare you say nice things about Microsoft?" ones. Some will touch both bases. Well, I can't help it. Netscape Navigator 4.0 is almost two years old now. Two years. What is that in Internet time? Navigator is slow, clumsy, buggy, crashes frequently, and has a poor feature set. Internet Explorer 4 was noticeably superior to Navigator 4, and IE5 just widens the gulf. There are a lot of things I don't like about IE5--and about Microsoft, for that matter--but on balance IE5 is the best browser out there. I keep Navigator 4 on my system, but only to check how pages look to Navigator users. And what's worse is that it appears that Netscape has lost the war. Their next-generation browser won't appear for at least six months and probably a year. When it does, it is not likely to be much better than IE5, if indeed it is even as good as IE5. And, of course, Microsoft will be well on the way to IE6 by then. * * * * * This from Francisco Garcia Maceda [maceda@pobox.com]:
I agree that a dual-processor system is very desirable under Windows NT, and also that the difference in speed between a dual Celeron/450 and a dual Pentium II/450 is a minor issue. You can, after all, buy a dual-CPU motherboard and two Celerons for considerably less money than one Pentium II/450 costs by itself. But the way you use your system makes all the difference. Anand did a test of dual-CPU systems recently, and concluded that dual processors don't offer much advantage. Which they don't if you only do one thing at a time and use single-threaded software to do it. A lot of people still use Windows pretty much as a program loader and task-switcher rather than as a multi-tasking environment. For those people, a dual-CPU system isn't going to run much faster than one with a single CPU. But those of us who keep a lot of windows open and have a lot of things going on at the same time will see a dramatic improvement from adding a second CPU. As you are aware, for someone building a dual PPGA Celeron system right now, the MSI MS-6905 slocket card is the best choice. But several other manufacturers currently make such adapters, and many of them are likely to provide jumpered versions in the near future. Right now, the slockets are specialty items and only real enthusiasts are likely to use them as a means to running dual Celerons. But the jumpered versions are likely to become standard stock items in mainstream computer stores before long. I expect they'll show up at CompUSA and similar places in the next few months. I wonder if the supply of dual-CPU motherboards will dry up when that happens. * * * * * Mid Afternoon: A surprising lack of inbound flaming messages. I wonder if people aren't reading this, my mail is not arriving, or perhaps no one cares. * * * * * This from Thomas Jenkins [tjenkins@juno.com]:
Well, that's not been my experience. Perhaps it's because I run Windows NT for everything but testing purposes. Perhaps I'd have different experiences if I ran Windows 9x, but I don't think so. I have both IE4/5 and Navigator 4 running on several NT machines, and the behavior is the same on all. Netscape crashes on me frequently, including on machines where it has been freshly installed. I frequently keep many separate incidences of my browser open while I work, so I find Nav's behavior there particularly annoying. With six or eight incidences running, Nav is very unstable. What's worse is that when one copy crashes, they all do. That's never happened to me with IE. I don't really like IE. In fact, I dislike it. Its bizarre behavior sometimes drives me mad, as I've said many times. But I dislike Nav much more. And I've tried Opera, but it seems feature-poor compared to IE and is no faster. So, if I had to characterize the current browsers, I'd say that IE is by far the best, but of a very bad crop. And thanks for the kind words. * * * * * This from Chuck Waggoner [waggoner@gis.net]:
Well, I'd never discourage anyone from running NT on a dual-CPU box, but the fact is that all you really need is NT running on even a single-CPU box. Never lose sight of the fact that Win9x is really just a shell running on top of MS-DOS, which was never designed as a multi-tasking OS. The surprising thing is not that it crashes so frequently but that it runs at all. Despite everything you might hear to the contrary, NT is an extremely stable OS. I abuse it to death and it just keeps on ticking. Most of the reports you read about NT crashing to a blue screen are from people who are running it on a miscellaneous collection of hardware of dubious origin and/or are using buggy drivers. In particular, when, with NT 4.0, Microsoft moved video from user mode to kernel mode to increase performance, they also implicitly made the decision that it was okay for a buggy video driver to crash the kernel. On real servers, I stick with the vanilla Microsoft drivers, and on my main systems I use video cards from manufacturers (like Matrox) that are known for providing stable drivers. If you take a little care in choosing which hardware to run NT on and which drivers to use with it, it'll run for weeks or months on end with no problems. I know because I do it. As a dancing-bear experiment, I once loaded Windows NT Server 4.0 on a 486/66 system with 40 MB of RAM. I then loaded every standard service available (except Services for Macintosh) and a bunch of optional and third-party ones, e.g. Services for NetWare, several Arcada BackUp Exec services, etc., etc. By the time I finished, I had more than fifty named services loaded and running. I then started running applications, trying to crash the system. By the time I gave up, I had more than a hundred application windows open, running everything from Word97 to a C compiler to a dozen instances of IE. By this point, of course, the system had slowed to a glacial rate. Just opening a new application window was taking several minutes. But the point is this: nothing crashed. Everything ran, albeit very slowly. So NT is inherently a very stable OS. My main workstation at the moment is a Pentium II/300 box built around an Intel Seattle system board and 128 MB of RAM. That's plenty of horsepower to run what I need to run. I often have a dozen or more windows open, and almost never crash NT. Applications crash, certainly, but that's not NT's fault. And, of course, when an application does crash, it has no effect on NT itself or on the other applications that are running. In my experience, the stability of NT and running applications increases with additional memory. I run NT in 64 MB minimum, and 128 MB seems to be the sweet spot. You'll also find that, given enough memory, NT is faster than Win9X. The break-over point seems to be somewhere between 32 MB and 64 MB. At 32 MB, Win9X is definitely faster than NT. At 64 MB, NT comes into its own. By the time you have 128 MB, Windows NT is noticeably faster than Win9X at running most applications, or so my experience has been.
|
|
Saturday |
Saturday, March 27, 1999 WARNING I was attacked last night by a macro virus called Melissa that arrived via a Word document that was attached to a mail message from Jerry Pournelle. Jerry and I send each other attached files, including Word documents, pretty frequently, so I didn't think much about it. The first strange thing happened when I double-clicked the attached file icon and Word popped up a warning that the file contained a macro. I foolishly told Word to go ahead and load the file, which was a rather banal list of "do's and don'ts" for creating a web page. Just at that point, an Urgent message arrived from Jerry, saying that the file was a Trojan Horse. I immediately deleted the message, exited Outlook 98, and rebooted NT. I downloaded the latest virus scanning files for Norton Utilities for NT and ran the scan, but they found nothing. I then used NT's Find from the Start button to list all files changed today, sorted by date/time stamp. I located a couple of cached copies of the Word document in various subdirectories of \WINNT. The only other changed file (other than Outlook.pst, of course) was C:\Program Files\Microsoft Office\Office\Templates\Normal.dot. I opened it with Notepad and found among all the non-ASCII garbage several fragments from the subject line and text of the Melissa virus. Fortunately, I'd done an Xcopy backup of all my current data to a network drive at 12:04 p.m. this afternoon, just before I started my backup to tape. That gave me two good recent copies of Outlook.pst--the Xcopy version from 12:04 p.m. and the one on tape from 3:17 p.m. I noticed a couple of strange things about the infected .pst file. First, it was 45 MB, which seemed rather large. When I looked at the 12:04 p.m. version, I found that it was only 24 MB. I knew that I hadn't gotten 20+ MB of messages today. When I subsequently re-opened and then closed the .pst file again, it instantly grew to 50+ MB. I deleted Outlook.pst and Normal.dot and restored them from tape, so everything is fine. I've probably lost at least a few messages from readers that I hadn't replied to. If you sent me mail and haven't gotten a reply yet, please re-send it. With my system dis-infected and Outlook working again, I called Jerry to let him know that the virus wasn't as benign as he'd thought, and to find out the details of what happened. It all started when he got what looked like a normal message with an attached file. When he opened it just as I did, it began sending the message to people from his Contacts folder. Fortunately, he caught it before it'd sent many and informed everyone that it had sent the message to. The original message was addressed not just to Jerry, but to John Dvorak and several other big-name computer journalists, so this thing is bigger than just a vendetta against Jerry by the sucky pages people, which is what I'd first assumed. Although I'd killed things quickly, I was concerned that perhaps this virus had sent mail to some people from my Contacts folder before I'd stopped it. As it turns out, it didn't. Jerry tells me that the outbound bogus messages all showed up in his Outbox, just as though he'd actually sent them intentionally. I didn't have any such message in my outbox, so apparently I caught it before it did any damage. I hate people who do things like this. Keelhauling is too good for them. As of this morning, I've download McAfee Viruscan, which doesn't even detect Melissa, let alone disinfect it. The virus made some changes to my registry, which I've left in place for now. They appear not to be doing any harm, and I didn't want to change them manually before running a disinfection program. Be very careful if you get an email message with an attached Word file, even if the message appears to be from someone you know and trust. Make sure that you have Word's macro virus protection (such as it is) enabled. If you open the attachment, make sure not to allow Word to run any macros. * * * * * Barbara picked up Larry Niven's latest book, Rainbow Mars, for me at the library earlier this week. I wanted to like this book. I really did. The collaborations between Pournelle and Niven are some of my favorites. But I have just never much liked anything that Niven writes under his own name. I didn't like his last one, either, although, like this one, it got excellent reviews. I found this latest book meandering and confusing and, yes, I do get all the "in jokes." Perhaps it's just because I've never much cared for fantasy. The only fantasy I ever recall actually finishing was Heinlein's Glory Road. I'm looking forward to Burning City, though, fantasy though it may be. Pournelle will be able to write a readable fantasy if anyone can. So I gave up on Niven's latest about half way through. That's extraordinary for me. Unless a book is just incredibly bad, I almost always finish what I've started. And this book is not incredibly bad, which makes my abandoning it in mid-read all the more unusual. Fortunately, Barbara made another library run today. She brought me back Gates' new Business @ the Speed of Thought and Burke's The Day the Universe Changed. I read the latter when it first hit print, but it's worth a re-read. * * * * * This from Tim Werth [timothy.werth@eds.com]:
Good points. I've been using NT pretty much exclusively since version 3.50 shipped. I used Windows 95 long enough to satisfy myself that it was garbage, which didn't take long. I have only one box in the house that runs Windows 95 and one other box that runs Windows 98, and both of those dual-boot NT. I have to keep a system running each flavor of Win9X available for doing screen shots for books and so on, or I wouldn't bother. My attitude is that if a particular software package won't run on NT, I don't want it. Navigator effectively falls into that category.
This from Tom Genereaux [entropy@lawrence.ks.us], whose original message on Pournelle's site was what got me started about Netscape yesterday:
It sounds like you should be running NT. I think you're one of the few people I've ever talked with who ran NT and then went back to Windows 9X. * * * * * And that got me started thinking about NT. I'm always surprised by how many smart people continue to use Win9X. When I switched to running NT pretty much full-time about four years ago, I felt at times like the Maytag repairman. I didn't know anyone else who was running NT on clients, and few enough who were running it on servers. When Windows 95 arrived and everyone switched to it, I began to wonder if I shouldn't switch too. It was undeniably prettier than Windows NT 3.5, but a week or so of playing around with Windows 95 cured me of any desire to use it for anything that mattered. When NT 4 shipped three years ago, most of my computerish friends quickly moved to it. Nowadays, nearly everyone I work with or correspond with regularly is running NT instead of Windows 9X, including my publisher's staff, and nearly all of my regular on-line correspondents. Even some of my non-computerish friends have started running NT at home. Right now, NT Workstation typically costs an extra $99 when you buy a PC. Even with this premium, a lot of people nowadays order their new systems with NT Workstation installed. If Microsoft had any smarts at all, they'd get rid of that premium and let people choose either Windows 98 or Windows NT 4 Workstation for the same price. I think they'd find that NT would soon have the majority of that market. Some might object that NT doesn't have Prug-'N-Pray or USB support. So what? About 99% of the time, you plug a drive or an expansion card into an NT box, install the drivers needed (if any), and it just runs. I've had more problems with PNP under Win9X than I've ever had configuring NT boxes without it. USB isn't making much headway right now, so almost no one is going to miss it. And there's always NT 5.0, or W2K or whatever they decide to call it. It supposedly will have both PNP and USB support. In fact, Microsoft has supposedly had USB support ready to ship for NT4 for more than a year now, but (if you believe the rumors) are withholding it as an incentive for people to upgrade to NT5. Win9X is a legacy product for Microsoft. They'd much rather you ran Windows NT. So why discourage people from running it by charging more for it? Novell made the same mistake when they shipped NetWare 4.0. Instead of charging the same for 3.12 and 4.0, they placed a hefty premium on NetWare 4.0, their "advanced" network operating system. People stayed away in droves, and many of those who decided they needed something more than NetWare 3.12 ended up taking a long look at Windows NT Server. Microsoft has the same problem. They don't want to give up the bundling revenue from Windows 9X, of course, but they really would prefer that people buy Windows NT. Eliminating the premium would encourage a lot more people to choose NT, which in turn would force third-party software vendors to port those few products that don't currently run under NT, and would also ensure that hardware vendors provided frequently-updated NT drivers. If Microsoft wants to drive NT sales and make it the ubiquitous stand-alone and client OS, the way to do so is not by charging more for it. It used to be that Microsoft had little choice. Back when memory cost $50 per megabyte, NT simply required too much hardware to be a viable mainstream consumer OS. But those days are long gone. Even the sub-$1K PCs nowadays run NT adequately, and spending an extra $50 or so to bump their RAM to 64 MB allows them to run NT faster than they run Windows 98. So why doesn't Microsoft get rid of that stupid $99 upgrade charge for NT? * * * * * This from Bill Grigg [bill_grigg@bc.sympatico.ca]:
Hmm. It's interesting. Of the people running Windows 95/98, about half agree with me about how bad Navigator is, and the other half think I'm crazy. Of those running Windows NT, almost everyone agrees that Navigator is a pale shadow of IE.
|
|
Sunday |
Sunday, March 28, 1999 WARNING If you receive an email message with an attached Word document, DO NOT open that document, even if the message appears to come from someone you know and trust. A new, extremely virulent macro virus called Melissa is making the rounds. You can read more about it at CNN or ZDNET. Messages that contain the virus currently have the subject line, "Subject: Important Message From <the name of the person who relayed the virus>" and the message body "Here is that document you asked for ... don't show it to anyone else ;-)." However, there is nothing to prevent someone from modifying the macro virus to use a different subject line and message body, and minor variants on the virus have already been seen. What a mess. And likely to become much more of one come Monday morning, when millions of people fire up their email and open infected documents. I've been doing a field-expedient virus scan on all my systems, because none of the commercial virus scanning products yet have Melissa in their databases. So I've been using Start - Find to do a search for all files on my systems that contain the text "elissa". So far, everything appears clean. I manually removed the changes Melissa had made to my registry, and re-enabled macro virus protection in Word. Although this virus is generally described as relatively benign, there are some aspects of it that are cause for concern:
* * * * *
|