|
Week of 24 January 2000 Sunday, 30 January 2000 08:55 A (mostly) daily journal of the trials, tribulations, and random observations of Robert Bruce Thompson, a writer of computer books. |
|
wpoison
Search [tips] |
Monday, 24 January 2000 [Last Week] [Monday] [Tuesday] [Wednesday] [Thursday] [Friday] [Saturday] [Sunday] [Next Week] School is open today, although delayed by two hours. Barbara says the roads aren't all that bad, and she's off on errands and a gym visit. We have more snow forecast for Wednesday, which means the grocery store shelves will start emptying rapidly. Wednesday is her usual shopping day, so to avoid the rush she's also going to stop by the store to buy Coke, milk, and other essentials. I see that the judge has ruled in favor of the movie industry with regard to DeCSS, ignoring the fact that it's impossible to unexplode a firecracker. EFF will appeal, of course, but the damage has already been done. If you're something the size of the movie industry, you can roll over the rights of little guys, and there's nothing they can do to prevent it. It's almost enough to make honest people want to go out and start pirating DVDs. I won't bother, because I regard watching movies as a waste of time. In the two hours or so that it takes to watch the best movie ever made, I could read even a mediocre book and be the better for it. I started to watch a tape of Arnold Schwarzenegger's Predator last night. I hadn't seen it, and it got very good reviews. What a waste of time. I gave up half an hour into it and started reading a mystery. I can't even remember the title or the author, but it was certainly a more enjoyable use of my time than watching a ridiculous movie. The movie and music industries are fighting a losing battle, and I can't wait to see them lose. Of the two, I have much more contempt for the music industry. The movie industry at least creates the product. The music industry creates nothing. They're leeches, sitting between the content creators and consumers. They're doing everything they can to make sure they can continue to grab the lion's share of the profits, but MP3 will ultimately do them in. I once suggested, only half in jest, that someone who wants the latest Tom Petty shouldn't buy the CD. Instead, he should pirate the CD and then put a dollar bill or two in an envelope and mail it to Tom Petty. Tom Petty ends up making more money that way and you get the CD cheaper. The only loser is the music industry, which richly deserves to go bankrupt. Morally, you've taken the high ground, although legally you're still a felon. I use Tom Petty as an example because he actually tried to post some of his music as MP3s on his web site, but was stomped down by the music industry. Bastards. * * * * *
If I understand what you're saying, someone was copying music and selling it to purchasers who already owned a copy of that music in another form. If so, that's quite different from what I was talking about, which is one person making copies, whether from his own copy or a borrowed copy, for his own personal use of music that he already owned. But I agree that copyright law is confusing. * * * * *
Thanks. * * * * *
Thanks. I couldn't verify your experiment with the NT4 version of sol.exe, because I don't normally install games when I install NT. W2K, however, installs the games automatically, so I had them on all my W2K boxes. I tried running Solitaire on each of them, and you're right. The cards bouncing off the stacks at the end of the game is much slower on all of the systems. That sure explains it, because the video on all of these appeared to be operating at normal speed except for Solitaire. As far as the behavior you describe later in your message, I assume that's just the fact that W2K supports Plug-'N-Play whereas NT4 does not. I assume that if your BIOS has a setting for "PnP OS?" you must have set it to "Yes". If you set it to "No", I suspect you'll find that W2K can no longer mess with the IRQs you've mapped to particular PCI slots.
|
|
Search [tips] |
Tuesday, 25 January 2000 [Last Week] [Monday] [Tuesday] [Wednesday] [Thursday] [Friday] [Saturday] [Sunday] [Next Week] We had more snow and freezing rain overnight, with still more forecast for today, tonight, and tomorrow. It looks like we have an inch or two (2.5 to 5 cm.) of new accumulation, with possibly another 4 to 6 inches due to arrive over the next day or so. Low temperatures in the high teens (about -8C) and high temperatures at or just above freezing for the next few days mean this stuff isn't going away soon. It may not have been my imagination when I reported earlier that Windows 2000 Professional seemed slower than Windows NT Workstation. The Register reports that Microsoft now concedes that W2KP is "comparable" in speed to NT4W on systems with 64 MB or less RAM, and slower than NT4W on systems with 128 MB or more. Quite a change from Microsoft's earlier position that W2KP was "up to 24% faster". I also question the reports of greatly increased stability with W2KP. It's not that I doubt that W2KP is very stable. I'm sure it is. It's that I doubt there's much basis for comparison, because NT4 is also very stable. I have had several systems running NT4 for years, and NT4 is about as stable as anything I've used. Some of my NT4 systems run for literally months between restarts, and those usually happen only when I need to power a system down for some reason like changing hardware. So, although I don't dislike W2KP, I don't really see that it's much of an upgrade from NT4W unless you're simply dying to use USB. Certainly, I've seen nothing that would make W2KP a "compelling" upgrade, as has been reported elsewhere. W2KP runs slower than NT4W (or, at best, the same speed), is no more stable, is quite expensive, still seems to have problems working properly with an NT4 DHCP server, and has problems running some software that works perfectly on NT4W. Microsoft hasn't even certified Office 2000 to run on W2KP (!), although it seems to work fine here. Admittedly, I'm in a small minority here, perhaps even a minority of one, but when I see a naked emperor I point him out. * * * * *
Interesting that, just as we're discussing this, the RIAA has taken MP3.com to court for their "my MP3.com" service, whereby anyone can upload his own albums in MP3 form to the MP3.com server, and then access them from anywhere. What may be significant is that one doesn't actually upload the music if it already exists on the server. Instead MP3 flags the account as having access to that music, and simply uses the same music that someone else had previously uploaded. Oh, well. RIAA delenda est!
|
|
Search [tips] |
Wednesday, 26 January 2000 [Last Week] [Monday] [Tuesday] [Wednesday] [Thursday] [Friday] [Saturday] [Sunday] [Next Week] Thanks to all of those who wrote expressing concern about the winter storms we've been having around here. Actually, Winston-Salem missed the brunt of it. We got only about an inch (2.5 cm) of new snow from the big storm that passed here yesterday. Kernersville, about 10 miles (16 km) east of us, got about a foot (30.5 cm). Raleigh, a couple hours east of us, got 20 inches (51 cm). Some areas not far to the southeast of us got 24 inches (61 cm). Winston-Salem is in good shape--roads clear, power working, and so on--but those to our east really got nailed. The Register posted another article this morning about Windows 2000 performance problems, this time with respect to video drivers. Apparently, many of the video drivers that ship with W2K are missing minor features such as support for hardware acceleration, D3D or OpenGL. It appears that some of the drivers shipping with W2K are enough to make your screen light up, but that's about it. That's a minor issue for Microsoft, of course, compared with the importance of meeting an arbitrary ship date. As I've said in the past, I recommend refusing to upgrade to Windows 2000 Professional until SP1 ships, and to Windows 2000 Server until at least SP2, and preferably SP3. SP1 is already scheduled for release mid-year, which should tell you something. Let the other guys end up with the arrows in their backs. I've encountered something truly strange when attempting to process web stats. I've been using Analog for more than a year to process the raw web logs that pair networks provides into usable reports, with never a problem. I did the first week of this year, 20000101 through 20000107, with no problems. But when I downloaded my daily log file for 8 January, it caused Analog to GPF. I extracted the data from the zip and tried again. Same problem. The log files for 9 January and 10 January processed normally, as did Pournelle's log files for the whole period, so I figured that the 8 January file had something odd about it. Then the same thing happened on the 11 January file. I mailed the author of Analog, offering to send him the files. He mailed me back to ask for them, but I haven't heard anything from him. Meanwhile, when I tried to run Pournelle's stats last weekend, Analog GPFd on his data. At that point, I was getting concerned enough that I emailed pair support and explained what was going on. They mailed me back yesterday evening to say that they'd run the reports there with no problem, using the raw logs that still resided in my directory on pair's server. They suggested that perhaps something had been corrupted during the download. Okay, that seemed easy enough to check, so I downloaded my 8 January raw log file, uncompressed it and ran it. Analog processed it normally. At that point, the problem seemed solved, other than trying to figure out why a file would have been corrupted during download. I re-downloaded the raw log file for 11 January, and Analog also processed it without any problems. I'd already run the stats for 8 January through 14 January, but without the files for the 8th and 11th, so I decided to re-run that week's report with the new, good versions of the files for the 8th and 11th. Analog GPFd. Okay, I *knew* that the files for 1 January through 7 January were fine, because I'd already processed them and looked at the report. I still had the actual uncompressed files for 1 January through 7 January on disk, so I ran the report again. Analog GPFd. At that point, the only thing I could think of was that perhaps Analog was timing sensitive. I had it running on two different machines which exhibited the same problem, but both machines were pointed to the raw data located on a network server. Perhaps the problem was due to the stochastic nature of Ethernet and would be solved by running Analog against raw data stored on the local drive. I copied all the expanded log files over to the local machines and tried again. It blew up again. At that point, I'd about decided that perhaps Analog 4.01, which I'd recently upgraded to, was buggy, so I reverted to Analog 3.11, which was the version I'd been using since last spring. The first couple of runs I made processed everything normally, so I'd about decided that Analog 4.01 was in fact the problem. Then I ran one last report, and Analog 3.11 GPFd. So at this point I have no idea what's going on. Analog seems to have taken a dislike to my data, my machines, or something. I also did some experimenting last night with a product called Teleport Pro, which is an off-line web browser. IE4 and IE5 have this function built-in, but it's very limited. IE5, for example, allows you to go only three levels deep from the start page you specify, and it doesn't deal at all well with database-driven sites. I read the features list of Teleport Pro, and it appeared to be just what I was looking for. They have a crippled version available for download, so I downloaded it and gave it a try. It kind of works, in the sense that it went out and downloaded the site I pointed it at (within its limitation of 500 pages), but the downloaded local copy of the site doesn't function properly. I wasted a couple hours messing with Teleport Pro, and I'm about to give up on it. Does anyone know of an off-line web browser that actually works, particularly with database-driven sites? It shouldn't be all that hard to implement one. After all, all it has to do is automate the process of clicking on links and then store the resulting pages. Search engine spiders do that constantly. For dynamically-generated pages, it could simply assign a random page name to store the returned HTML page as, and create a map of the names it assigned to the string used to generate the page (in case that same data was referenced from another link). For that matter, with disk space so cheap, it could download and store duplicate copies of such pages. What I want is a product that I can simply point at a web site and tell it to go get the whole site. Is there such a thing?
|
|
Search [tips] |
Thursday, 27 January 2000 [Last Week] [Monday] [Tuesday] [Wednesday] [Thursday] [Friday] [Saturday] [Sunday] [Next Week] Those morons at the DVD Copy Control Association (DVD-CCA) would be funny if they weren't so evil. In the latest news, John Hoy, the President of DVD-CCA submitted to the court a declaration which includes the material that the DVD-CCA is trying to suppress, apparently not realizing that the declaration itself would become a part of the public record. So yesterday the DVD-CCA filed an emergency ex parte motion to seal the records. Duh. More on the NT4 versus W2K speed issue in this article from The Register. It appears that it was not my imagination. W2K is slower than NT4, sometimes enough so to be noticeable. Microsoft has apparently been slanting the benchmark tests to make it seem otherwise. The fact that I noticed W2K's slowness most on one of my slower systems probably just means that the same percentage difference is much more noticeable on a slow system than on a fast one. After reading that article, the very next one I read was this one, reporting that RAMBUS is dead. After reading that, I just couldn't resist sending editor Mike McGee of The Register the following message: -----Original Message----- I read your page every day, but perhaps you should start reading mine. I declared RAMBUS dead in my daily journal for the week of 19 July 1999, and reported lower performance with W2KP build 2195 a couple of days before you guys broke the story. :) To which he replied:
And.
And speaking of The Register, I see they're reporting this morning that the snow storm that dumped up to two feet of snow on the east coast sneaked up on us because the weather forecasters believed an IBM weather forecasting supercomputer instead of their own eyes. Reminds me of a weather forecast I heard once from a weatherman at a Vermont radio station: "Snow today. Don't know how much. Hasn't stopped." * * * * *
If you did a standard W2KP installation (without choosing advanced options), W2K setup probably converted your NTFS4 volumes to NTFS5, which NT4 cannot read. That would account for your inability to access NT4. As far as the problems accessing Windows 98, I'm not sure. Perhaps one of my readers will have an idea, and can mail you directly. I'm afraid that you are faced with stripping down to bare metal and re-installing, though. * * * * *
I think you're misinterpreting the statements. It's not that W2K's efficiency peaks at 32 MB. It's that its performance relative to NT4 peaks at 32 MB. For example (using entirely imaginary numbers), at 32 MB W2KP may provide 1.00 TU/s (Thompson units per second) performance while NT4W may provide only 0.96 TU/s. At 64 MB, W2KP may provide 2.50 TU/s, and NT4W 3.00 TU/s. At 128 MB, W2KP may provide 3.00 TU/s and NT4W 4.00 TU/s, and so on. More memory helps either W2K or NT4, but it helps NT4 more. I haven't done any tests yet with W2KP, but NT4W appears to be usable in 32 MB, happy in 64 MB, and delighted in 128 MB. For most users, more than 128 MB yields diminishing returns. That's not true, though, if you run seriously memory-hungry apps (like Photoshop) or if you keep many, many windows open. More memory is always a Good Thing, unless of course you're running a Pentium system with a chipset that doesn't cache beyond 64 MB, in which case installing more than 64 MB will actually slow the system down.
|
|
Search [tips] |
Friday, 28 January 2000 [Last Week] [Monday] [Tuesday] [Wednesday] [Thursday] [Friday] [Saturday] [Sunday] [Next Week] I think I'm coming down with something. I've almost lost my voice. When I attempt to speak sternly to Malcolm, I squeak. I can't remember the last time I took a day off, so I think I'll do that today and just lie around and read. I do have to do the weekly backups, though. More tomorrow, or perhaps not, depending on how I'm feeling.
|
|
Search [tips] |
Saturday, 29 January 2000 [Last Week] [Monday] [Tuesday] [Wednesday] [Thursday] [Friday] [Saturday] [Sunday] [Next Week] Thanks for all the get-well messages. My voice still isn't working very well, but I think I'll live. Barbara took care of mom and the dogs all day yesterday, and I just lay around reading and drinking chicken-noodle soup. I polished off the latest W. E. B. Griffin and a couple of mysteries. In between books and soup, I also ran the weekly backups. Barbara's machine, theodore, is also the main network data store. It has a Tecmar Travan NS/20 tape drive. My own main machine, kiwi, has a Tecmar 3900 DDS-3 "DAT" drive. I've gotten in the habit of doing a local-only backup on theodore (because that's where all the data is) and a full network backup from kiwi. One reason I do it that way is that theodore runs Windows NT Server 4 and kiwi runs Windows NT Workstation. Tecmar included two versions, server and workstation, of Arcada/Seagate/Veritas BackupExec with the drives. The workstation version can back up any volume on the network that has a drive letter mapped to it, but will install only on NT Workstation (not NT Server). The Server version will install on an NT Server machine, but can back up network volumes only if they reside on machines running NT Workstation or Windows 9X (not NT Server). That means, for example, that I can't use Barbara's machine to back up data from kerby, which is my former main workstation and is currently the BDC for the network and my WinGate server. Volumes on kerby are visible, but grayed out. I guess that's why they call it "single server edition." After I'd started the network backup on kiwi, I sat down at Barbara's machine to retension a tape. For some reason, I got to wondering if the BackupExec workstation version checked the OS version only at install-time, or did a run-time OS version check. So I copied the BackupExec Workstation folder from kiwi over to theodore and ran the executable to see what happened. The splash screen appeared normally, but the program load aborted with a missing DLL. I copied that missing DLL over from kiwi, fired up BackupExec again, and it again blew up, this time with a different missing DLL. Being stubborn, I continued finding and copying those DLLs from kiwi over to the main BackupExec directory on theodore. After copying a couple of them, I found an executable, versions.exe, in the BackupExec folder and ran it. It provided a list of the installed executables and DLLs, with question marks next to the missing ones. That made it easier to track down which DLLs were missing, and I got them all copied over. BackupExec Workstation then fired up and ran normally on NT Server. Well, normally except for one thing. The drop-down list for destination showed "No devices found". After a short look around for configuration files, I concluded that that information must be kept in the registry, so I started spelunking that on both machines. Part of the problem is that these software versions apparently overlapped the change from Seagate BackupExec to Veritas BackupExec. The workstation version (on kiwi) had HKLM registry entries under Software\Seagate. The server version (on theodore) had registry entries under both Software\Seagate and Software\Veritas. The Seagate registry entry on kiwi had a Devices key, which was not present under either Seagate or Veritas on theodore. I couldn't find anything at all on theodore that was obviously tape drive configuration information. I could simply have copied the Devices subkey from kiwi over to theodore, except that the machines use different drives. Short of removing the Tecmar Travan NS/20 tape drive from theodore, installing it in a machine running NT Workstation, installing Seagate BackupExec Workstation on that machine, and then copying the appropriate registry keys, I didn't see any obvious way to get the information I needed, so I simply punted and used the Server version on theodore to do the backup. Also, the Tecmar 3900 DDS-3 tape drive in kiwi keeps track of how much use it's gotten since the last cleaning. I've never cleaned it, and when I started the backup it decided it was due for its first cleaning. That presented a problem, because I don't have a cleaning tape. Thinking about that, I realized that the Tecmar Travan NS/20 tape drive on theodore could probably use a cleaning too, although it's too polite to complain. That sent me off to Insight, NECx, and PC Connection in search of both DDS and Travan cleaning tapes. That was harder than it should have been. Everyone had DDS cleaning tapes. Like DDS tape cartridges, they're cheap, about $6 to $10 or so. I had my choice of major brands like HP, Sony, Tecmar, Seagate, and so on. Any of those would be fine. But when it came to Travan cleaning tapes, there was a distinct shortage of products. Having once seen a tape drive destroyed by a rogue cleaning tape, I really prefer to clean my QIC80 drives manually with alcohol, but the Tecmar Travan NS20 manual made a point of insisting that you should use a dry cleaning tape because using liquids might damage the drive and would void your warranty. So I definitely wanted a dry Travan cleaning tape. One problem is that Travan cleaning tapes, like Travan data cartridges, aren't cheap. I was able to find some at $30 per cleaning tape or so, which I wouldn't have minded paying. The problem was that they were available only in five-packs, and I certainly didn't want to spend $150 for five when I only needed one. Neither Insight nor NECx was able to sell me what I wanted--one each DDS and Travan cleaning tapes. I was finally able to find what I needed at PC Connection, but I decided to check around locally before I ordered from them. I used to buy regularly from PC Connection. Their prices were usually a bit higher than other reputable sources, but they delivered overnight for a flat $5 shipping charge (unless you ordered a monitor, printer, system, or other extremely heavy product). They used to promise that if you had an order in by 3:00 a.m. it would be delivered by 10:00 that day. I remember being over at our friends the Tuckers late one Friday night/Saturday morning. I ordered something from PC Connection at about 1:30 a.m. Saturday morning, which in fact turned up at our door less than eight hours later. It usually worked out that the total cost with shipping was about the same from PC Connection as from Insight or NECx, and PC Connection never caused me any problems. Neither did Insight or NECx, for that matter. At any rate, PC Connection changed shipping policies a while back. They now charge variable shipping depending on weight. Apparently, their minimum is $10, because that's what they stated as the shipping charge to ship two tapes to me. And they no longer offer standard Saturday delivery. A $10 shipping charge on a $40 order is pretty significant, so I decided to check around locally first. I'd just as soon pay $10 more to a local place. We're supposed to get another winter storm, so I suppose we'd better batten down the hatches. This one supposedly moves in this afternoon and continues through tomorrow night. Perhaps we'll get only an inch or two this time also, but then again we may not. It appears this one is likely to be an ice storm rather than a snow storm, so things may get a little bad around here. Ice storms tend to kill the power, and a bad one can take the power out for days at a time. Oh, well. We have food, a generator, and natural gas logs, so we won't suffer. * * * * *
You shouldn't have anything to worry about. What I was referring to was Socket 7 systems, not many of which are likely to be upgraded to W2K. For Pentium II/III and Celeron systems, you won't have a problem. The problem arises with Pentium systems, because many Socket 7 chipsets supported more memory than they could cache. For example, the latest Intel Socket 7 chipset, the 430TX, supports up to 256 MB of memory, but can cache only 64 MB. The immediate predecessors of the 430TX, the 430HX and 430VX, were "twin" chipsets targeted at different markets. The HX was the "high-end" chipset, with support for dual processors, and the ability to cache 256 MB. The VX was the "economy" chipset, and could cache only 64 MB. Most systems of that vintage used the VX, which was much cheaper. I have two of those sitting here: a Dell Dimension XPS M200s (200 MHz Pentium) and a Gateway Pentium/133. Both use the VX chipset, and both are effectively limited to using 64MB, although more can be physically installed. Intel was not the only one guilty of making chipsets with a 64 MB caching limit. Many competing chipsets from VIA, ALi, SiS, and so on had the same limits. And, in fairness to the makers. that wasn't really a limitation when the chipsets were designed. No one thought for a moment that anyone would ever install more than 64 MB, because memory was simply too expensive back then to make that feasible.
|
|
Search [tips] |
[Last Week] [Monday] [Tuesday] [Wednesday] [Thursday] [Friday] [Saturday] [Sunday] [Next Week] We have ice. There appears to be half an inch (1.25 cm) or so on the ground, but it's still coming down and is supposed to continue through tonight. Our forecast high today is 30F (-1C), so it's likely to be staying around for a while. Looks like the kids will be getting another day off from school tomorrow. I think Barbara said we're supposed to get more later in the week, too. This has been a very wintry January for Winston-Salem. I blame it on Global Warming. * * * * *
Thanks for the correction. I believe you're right that the HX can cache up to 512 MB rather than 256 MB as I stated. That's what happens when I give an answer off the top of my head instead of checking. As for the rest, I've not tried running Photoshop or any other memory hogs on such a configuration. With the availability of cheap Celeron systems, running something like Photoshop on a Pentium makes no sense in a corporate environment anyway. The point I was trying to make is that Windows 9X, Windows NT Workstation, and (presumably) Windows 2000 Professional are all happy in 64 MB. Going above 64 MB provides some return, certainly, but even on a system that can cache more than 64 MB, the improvement is relatively small compared to the jump from 32 MB to 64 MB. The real problem is that Windows loads top-down in memory, and that means that if you install more than 64 MB in such a system, you'll be running Windows 100% of the time in uncached memory. My experience is that the performance degradation from doing that is more than the marginal performance improvement from having more than 64 MB. I actually installed 128 MB in such a system one time, just to see what would happen. I ran quite a few benchmarks, and they all showed that the system was slower with 128 MB. It also "felt" slower, although that may have been simply because I was expecting it to be. This discussion is probably moot anyway, as there is little reason to upgrade any Pentium system to more than 64 MB in a corporate setting. Given the costs involved in any corporate PC upgrade, it'd make more sense to replace the old PC with an inexpensive Celeron/400 or similar system that comes pre-configured with Windows 2000. Attempting to upgrade an elderly fleet of Pentium systems to run Windows 2000 is likely to be an exercise in frustration. Unless the systems are identical, and perhaps even then, the staff time required will be so great as to subsume the marginal costs of the drop-in upgrade. * * * * *
That's fine for Windows 9X systems, but we're talking about Windows NT Workstation and Windows 2000 Professional systems, for which SiSoft Sandra has only very limited functionality, or so the president of the company tells me. The problem, of course, is that NT and W2K isolate the underlying hardware from any application, including Sandra. All Sandra can do is report back to you what the operating system tells it, which is of very limited utility. I downloaded the free version of Sandra a couple months ago, and then contacted them with questions about the full version, W2K, and so on. The problem is that Sandra is a Windows application, and so cannot be run from a DOS boot floppy. For that reason, I regard Sandra as a user-level diagnostic program. I prefer using an industrial-strength diagnostic program like CheckIt Pro. As far as keeping multiple programs open and expanding memory beyond 64 MB, see my response to Francisco García Maceda. I still think it's a bad idea.
[Last Week] [Monday] [Tuesday] [Wednesday] [Thursday] [Friday] [Saturday] [Sunday] [Next Week] |