Home Daynotes Home Week of 20 October 2003

Photograph of Robert Bruce Thompson Daynotes Journal

Week of 20 October 2003

Latest Update : Saturday, 25 October 2003 18:35 -0400

Click Here to Subscribe Buy PC Hardware in a Nutshell, 3rd Edition: [Amazon] [Barnes & Noble] [Bookpool] Visit Barbara's Journal Page

Monday, 20 October 2003

[Last Week] [Monday] [Tuesday] [Wednesday] [Thursday] [Friday] [Saturday] [Sunday] [Next Week]
[Daynotes Journal Forums] [HardwareGuys.com Forums] [TechnoMayhem.com Forums]
{Five Years Ago Today]

12:47 - Late post today. I had to take my 4X4 out to get inspected, have the oil changed, and so on. We'd put almost 4,000 miles on the truck since it was inspected this time last year. That's actually quite a bit, given that we drive it almost exclusively on astronomy observing trips. We were planning to go up to the Green Bank, West Virginia radio telescope observatory this weekend, but Kerry's condition and Barbara's heavy workload made us decide that we'd best stay home.

One of my readers suggested that I add a "Five Years Ago Today" link to each day's journal entry, so I did that. I actually started doing it last week, but forgot to mention it. At first, I thought it'd be time-consuming to update it for seven days each time I created a new week's journal page, but it's easy enough to edit the html directly.

I've actually started clicking on those links myself. Some of the stuff I wrote five years ago I don't even remember writing. Not surprising, considering that with my journal page, books, email, and so on, I've written literally millions of words since then.

Barbara's sister's HP OfficeJet 7110 still isn't working. I spent a few hours over there Saturday trying to get it work using the instructions that Ron Morse kindly sent me. Unfortunately, the most I could get the PC to recognize was that there was an HP 410 scanner connected. It refused to load the driver for that and never did recognize the rest of the hydra functions. I sat down with Frances and Al and recommended they take that HP back for a refund and buy a separate laser printer and scanner. They don't really need color output, and I explained the hideously high cost of maintaining ink jet printers. Their PC came with a Lucent WinModem installed, so adding a separate laser printer and scanner will allow them to print, scan, copy, and fax. That's everything they bought the hydra for.

Now I need advice about what printer and scanner to tell them to buy. I haven't bought a laser printer or scanner in quite a while. They're looking for something that totals around the $300 or so that they paid for the HP 7110 (or less). I told them that there are numerous personal laser printers that sell in the $200 range, and $100 buys you a very nice scanner nowadays. As far as the printer, I'm thinking HP, Brother, or Samsung. As to the scanner, they need something primarily for making photocopies to the laser printer, scanning photos, and outbound faxing. Ideally, the scanner software (or buttons on the front of the scanner) would make it easy to do a photocopy by pressing one button, or to send a fax by pressing a button and entering the recipient's fax number. Consumables costs for the printer aren't a huge issue, because they'll probably not go through more than a ream or two of paper a year.

 If you have any suggestions for particularly suitable (or unsuitable) models, please post them over on the messageboard.

Plextor has responded to the questions I sent them about Mount Rainier support (and changing the Book Type field) with the PX-708 DVD writers. I've edited the following to remove some extraneous stuff. My own comments are in bold italic. Plextor's are indented and in monospace:

The PX-708 does, indeed, support DVD+MRW.

I don't know where the confusion occurred since I was not in the conversation with our European support. However, in itself, Mt Rainier is a bit confusing, since it contains a number of components, for both CD and DVD.

Background formatting for one, which is a component of DVD+MRW and is supported by the PX-708. However, there is another component called "Defect Management". Philips has only just released the spec for DVD+MRW Defect Management, so few, if any, DVD drives support Defect Management for DVD media.

Thanks. I'll let my readers (and Mr. Thorarinsson) know.

I do have a few other questions, though:

1. Is adding support for defect management to the PX-708 a simple matter of upgrading the firmware once such an upgrade is available, or does defect management require hardware support that is not present in the PX-708?

KC: A firmware upgrade will accomplish this

2. Based on what you've said, is it fair to say that the PX-708 is currently Mount Rainier compliant for CD-MRW but not DVD+MRW?

KC: As of this date, that is the case with CD-MRW. With DVD+MRW, the drive supports Mount Rainier, but is not fully compliant because of the lack of defect management.

3. Several of my readers have asked about the lack of support in Plextor DVD writers for changing the Book Type field "compatibility bit". More than one has told me that they've installed foreign firmware on their Plextor DVD writers to add that capability. Can you tell me (a) why Plextor does not include that capability in its DVD writers, (b) if you plan to add that capability in a future firmware update, and (c) if not, why not?

KC: Plextor is honoring this request from the DVD Forum (-R people) as is Philips and others. For this reason, Plextor has not, and will not, add this support

As I understand it, changing the Book Type field simply causes the disc to lie and report itself as a DVD-ROM, which allows some DVD players and DVD drives to read discs that they would otherwise refuse to read. I understand that HP DVD writers do this by default. Is there some downside to offering users this capability, or even to making it the default?

KC: That is a fairly accurate description. I am aware of no downside, with the exception of possible patent conflicts.

So there you have it. The Plextor PX-708 DVD writer doesn't fully support Mount Rainier yet because Philips just released the Defect Mapping part of the specification. Plextor will no doubt issue updated firmware to add that support. As far as not allowing their drives to change the Book Type field, it's because the DVD Forum (which owns the rights to DVD) is being a pain in the butt about it. The DVD Forum, if you recall, is the same group that originally insisted that "DVD+RW" be renamed to remove the reference to "DVD". For a while, the DVD+RW folks changed the name to simply "+RW", but later changed it back.

Mark Huth has some interesting news.

-------- Original Message --------
Subject: xml
Date: Sat, 18 Oct 2003 18:49:46 -0700
From: Mark Huth
To: Robert Bruce Thompson

Just found out that Microsoft has again changed file formats in Office 2003, only this time they are using XML. Ordinary, plain Jane XML. Further, I'm told that they have published the file formats. I'm told that it isn't perfect, but pretty good. Only Outlook can't use the XML data as yet.

Mark Huth

"Space isn't remote at all. It's only an hour's drive away if your car could go straight upwards." Fred Hoyle

That'd be very nice, if true. I can't believe it's true, though. Microsoft owes their current dominance in both operating systems and productivity suites to their proprietary file formats. I can't believe they'll start using open, non-proprietary formats unless forced to do so by either the government or market forces. Call me cynical, but I think there must be a catch somewhere.

Finally, long-time reader Norman Yarvin tells me I made some errors in PC Hardware in a Nutshell:

-------- Original Message --------
Subject: PCHIAN errata
Date: Sat, 18 Oct 2003 20:55:49 -0400
From: Norman Yarvin <norman.yarvin at snet dot net>
To: Robert Bruce Thompson

I have several corrections for the third edition of "PC Hardware in a Nutshell". If you'd like me to post any or all of them on your message board, so that you can get comments on them from other readers, I will; but I figured sending them privately was best. (Also, please don't publish the above email address; if you want to publish one, use "<obfuscated by RBT above>". It also gets to me, but is more heavily filtered.)

p. 16-17, on ISA interrupts (edge-triggered) versus PCI interrupts (level-sensitive):

... "different PCI devices can assert different voltages on the same physical line"...

That's not how level-sensitive interrupts work. They use only two voltages, as in the rest of the system. Yes, if several devices share an interrupt line, the processor doesn't know which of them has produced the interrupt. But with level-sensitive interrupts, the devices on the interrupt line keep it high so long as any of them needs interrupt service. So the operating system can just keep trying to service them all, until the interrupt line goes low, whereupon it knows that all of them are happy.

With edge-triggered interrupts each device just emits a pulse when it interrupts; the processor detects the leading edge of that pulse. This gives the processor no easy way to determine that all devices are happy. (Keep in mind that multiple devices may interrupt at the same time -- this would be a problem for a system based on multiple voltages: which voltage would be present to indicate two devices interrupting at the same time?)

To a large extent this is theory which a PC hardware hacker need not know. But a practical consequence is that sharing interrupt lines slows things down a bit, since device drivers for devices on a shared line get run to see whether their device needs attention when it really doesn't. Still, this is only important when one is really pushing the limits of what the machine can do -- in particular, doing lots of I/O, and not just lots of I/O, but lots of little chunks of I/O, as occurs in multiple-track sound mixing/editing in contexts where low latency is required. (With I/O done in big chunks, the time spent schlepping the data around exceeds the interrupt overhead.)

By the way, Google gives me about equal numbers of hits for "level-sensitive interrupts" and "level-triggered interrupts". So the common usage is about even. But "level-sensitive" is the term I'm used to, and seems more descriptive.

p. 22: [regarding having space for memory-mapped I/O] "Modern processors use a flat (unsegmented) 32-bit address space, which allows them to access up to 4 GB of distinct addresses" ... but ... "few systems have anywhere near 4 GB of memory installed" ...

It's even better than that. On the x86, that is actually 4GB of usable address space *per process*. x86 machines can and do have more than 4GB of physical memory. (They could even have more than 4GB of memory per process, if programmers were willing to deal with 286-style segmentation -- which of course they aren't. x86 segmentation didn't, strictly speaking, go away; the segments just got much much bigger, so that only one of them has to be used.) Running out of physical address space for memory-mapped I/O is a non-issue; the 386's address space for physical memory is something like 42 or 48 bits.

p. 51: "PC components accumulate greasy brown residue, particularly if you smoke or if you heat with gas and oil."

Over the past three years I've been heating with oil, and my computer has accumulated nothing but dust. You'd have to be damn poor to vent the combustion gases of an oil-burning heating system into your house. If you're too poor to afford a bit of sheet-metal ducting, you're probably too poor to afford a PC. Even gas furnaces usually don't vent combustion gases into the house, although gas is clean enough that even if they did, it wouldn't cause PC components to "accumulate greasy brown residue". Sorry, but I think you smokers have to take the blame here.

p. 170: "The Pentium 4 uses 128-bit floating point registers..."

A nitpick here. Please just word that as "128-bit registers" or "128-bit registers which can hold multiple floating point or integer values". The term "128-bit floating point" is a term of art, which it's best not to invoke here (or seem to invoke -- I doubt it was your intention to try to invoke it), since it doesn't apply. (128-bit floating point, aka quadruple precision, is a very useful tool for people developing floating-point algorithms, but unfortunately no processors support it natively.)

p. 194: (on the Opteron) "Relocating the memory controller from the chipset ... to the processor core allows memory to be more tightly integrated with the processor for higher performance."

Well, it should, but in practice Intel has managed to squeeze almost the same memory latencies out of its latest chipsets as AMD has out of the Opteron. The Opteron is a lot better than AMD's older offerings, though. And Intel has also been better than AMD about improving latencies of L2 cache memory.

p. 230: "We don't pretend to understand this issue, but we've been told by memory experts that for systems with 512MB of RAM, using ECC versus nonparity memory is about an even trade-off in terms of extra cost and lost performance versus the likelihood of memory errors. For systems with 768+MB, we use ECC memory exclusively..."

Throw those memory experts out and get new ones. That rule is almost complete nonsense, which is presumably why you can't understand it. The value of data has almost nothing to do with how much data there is. A system used for video editing may require 2GB of memory. But the videos that are produced may be completely noncritical -- who cares about a single-bit error in one frame of home video? On the other hand, a system for logging geophysical data from a 50-million-dollar oil field seismic mapping project may only need 256MB of memory -- but that memory had damn well better be protected (with armed guards if need be).

It's true that the error rate is pretty much proportional to how much memory you install. But how much memory is installed varies from system to system by a factor of, say, 2 or 4 or 8, whereas the value of data varies by factors of hundreds or even millions.

p. 445: "SATA uses 500 millivolt (0.5V) peak-to-peak signaling, which results in much lower interference and crosstalk between connectors."

Though both of those facts are correct, the one isn't really the cause of the other. Serial ATA does have reduced interference and crosstalk, but that is because of the use of differential signaling over pairs of wires, with each pair individually shielded. Lower voltage doesn't really help in this regard; although a lower voltage reduces the amount of RFI that a cable broadcasts, it also increases that cable's susceptibility to incoming RFI by the same factor.

The main purpose of the lower voltage is to save power and/or increase speed. In this context, power use is proportional to the square of voltage. Hefty output drivers also take up chip space, and are slow. Reducing the voltage by a factor of 6 reduces the power by a factor of 36, which is a big improvement.

The number of data lines doesn't make much difference in the power-use calculation, since power use is proportional to frequency. For a given data rate, using fewer data lines means that you have to run them at a proportionally higher frequency, which means that the total power consumption stays the same. Using differential signaling, however, means that there are twice as many lines; so instead of a factor of 36 the improvment becomes a factor of 18.

By the way, I like the comparison table between bus standards on that page. It's the first good summary of speeds that I've seen.

Thanks. As I've said before, I'm no electrical engineer, let alone a processor designer, so it's good to have technical experts checking what I say. I did have an amateur radio license back when I was a teenager, and at one point I did have a first-class radiotelephone license, but those days are long behind me. I've put your message in my "corrections holding bin" and will check everything you've mentioned and make the appropriate corrections for the next edition.

13:31 - I don't have a good answer to this question. If anyone has a better answer, please email Rick and let him know.

-------- Original Message --------
Subject: Telephone wiring book recommendation
Date: Sat, 18 Oct 2003 13:58:09 -0700
From: Richard G. Samuels <rick at samuels dot com>
To: Robert Bruce Thompson

You have mentioned several times your experience with telephone system wiring. I just got drafted to do some work on the phone system at my daughter's school. After all, I'm a certified BOFH, so I must know something about phone systems. I've ordered a manual for their phone system, but I also have to deal with the wiring. The wiring all goes to punch down blocks, but nothing is labeled (of course). Can you recommend a book which might help me?

Unfortunately, I don't know of any, which is not to say that there isn't one. Most of what I learned was by on-the-job training and cut-and-try. That was backed up by numerous seminars I attended, most of which had large binders full of detailed information about structured cabling systems and so on. As far as I know, the only way to get those is by attending the seminars, most of which are pretty expensive. You can also get the EIA/TIA specification documents from Global Engineering. They cover a lot of pertinent material, but aren't tutorials in any sense.

There are a couple of web sites that cover this kind of stuff, but those I've seen aren't particularly trustworthy. Basically, if you're going to be doing any serious amount of work on a phone system, you'll need the following tools:

1. punchdown tool - they come with either fixed or interchangeable blades. You may need a tool for 66-blocks, 110-blocks (called 88), or both. 66 blocks are obsolescent, but are still sold and used. The 110 blocks (also called 88 blocks) are smaller and denser and are used in most new installations. You can use a screwdriver to seat wires in the claws of the blocks, but it's a pain in the butt and really not a suitable way to do any large amount of work.

2. toner and inductive amp - the toner puts a musical tone on a particular wire. You then use the inductive amp to locate that wire at the other end of the run or at an intermediate wiring closet. You can substitute a radio or other sound source for the toner and an earphone with a couple of clips for the inductive amp, but again it's a pain in the butt.

3. buttset - this is basically a telephone handset with a couple of extra features. You use it to verify dialtone and so on. You can substitute a regular telephone handset with clips instead of a modular connector, but again it's (you guessed it) a pain in the butt. Better to beg, borrow, or steal a buttset until you get the job done.

You may also need some supplementary tools like can wrenches, depending on your exact environment. None of the tools are outrageously expensive, but they cost enough that you probably don't want to buy them for one job. You can buy a cheap punchdown tool for $25 or so, although good models were $75 the last I checked. You can get an inexpensive toner & inductive amp for probably $50, although again a good set is going to cost a lot more. Buttsetts come in no-frills versions for maybe $75, although a full-featured model can easily cost $250, and some buttsets that have extra features cost $500 or more.

There are two color codes in common use. Jacks are usually wired with the old-style color codes, green and red for tip and ring on the first pair, and black and yellow for the second pair. Station cable never uses those colors unless it's quad (which is simply four wires with no twist). Quad has been obsolete for 30 years or more, but is still sold and used. Unfortunately.

Modern station cable is typically four-pair. The first pair is blue/white (white with a blue stripe goes to green on the jack; blue with a white stripe goes to red.) For the second pair, white with orange stripe goes to black on the jack, and orange with a white stripe goes to yellow. In this case, white is called the primary color, and the blue or orange is the secondary color.

You may also have to deal with 25-pair or larger trunk cables. Those are color coded using five primary colors and five secondary. Back in pre-Politically Correct days, phone company installers were taught to memorize these colors using the phrase "We rape beautiful young virgins, but only girls beyond sixteen". That's a mnemonic for the five primary colors, in order white, red, black, yellow, and violet, and the five secondary colors, in order, blue, orange, green, brown, and slate. On a 25-pair cable, the first pair is therefore white (primary #1) with blue (secondary #1), and the 25th pair is violet (primary #5) with slate (secondary #5).

To be honest, if you don't know what you're doing, I'd suggest you hire someone who does. You could make a real mess otherwise.



Tuesday, 21 October 2003

[Last Week] [Monday] [Tuesday] [Wednesday] [Thursday] [Friday] [Saturday] [Sunday] [Next Week]
[Daynotes Journal Forums] [HardwareGuys.com Forums] [TechnoMayhem.com Forums]
{Five Years Ago Today]

11:40 - Here's something interesting I noticed. I'm keeping track of my overnight mail. Total messages, total spams, percent caught by SpamAssassin, percent of the remainder caught by Mozilla Mail, and so on. For the entire 30 days of September, I had 8,535 overnight emails, of which 4,457 were spam and 4,330 were caught by SpamAssassin. For the first 20 days of October, I've had a total of 6,880 overnight emails, of which 4,446 were spam and 4,314 were caught by SpamAssassin.

In other words, I've received almost exactly the same number of spams in the first 20 days of October that I received in the entire 30 days of September. Fortunately, SpamAssassin is keeping up with them, catching 97.2% in September and 97.0% so far this month. Not only is the total number of spams increasing significantly, but the percentage of spam is going up as well. My main account averages 71.2% spam. That's five spams for every two real messages. And some days it's close to 90% spam. That's nine spams for every one real message. Something has to give soon.

One thing I get a lot of grief about is my advocacy of SCSI hard drives. I've gotten emails from people who outright called me a liar for saying that SCSI drives outperform ATA drives under load. The truth of that statement should be self-evident to anyone who stops to consider why servers use SCSI drives almost universally. It isn't because server admins enjoy paying higher prices for hardware. It's because SCSI simply blows the doors off ATA under load.

I think an awful lot of people see the world the way they want to see it instead of the way it really is. I don't understand the "I-don't-want-this-to-be-true-therefore-it-cannot-be-true" mindset, but there's no doubt that it is prevalent. In the case of ATA versus SCSI, a lot of people don't want to hear the truth because SCSI costs more than ATA, a lot more. But it's my job to tell the truth, at least as best I can determine what the truth is, so I'll keep on doing it. I will say that no one, ever, has installed SCSI on my advice and then come back to tell me he wished he'd installed ATA instead. The converse is not true, though. A lot of people have installed ATA despite my advice and then come back and told me they wished they'd gone with SCSI. Most of them ended up ripping out the ATA drives and installing SCSI, as they should have in the first place.

Don't get me wrong. SCSI is not a panacea. If you're running a single-user desktop system that doesn't make heavy demands on the hard drive, you're better off with ATA. It's cheaper, simpler, and, yes, faster. But that's true only on a lightly loaded system. If you start to hit the hard drive heavily, ATA soon bogs down. SCSI just keeps on ticking. Here's an interesting article that confirms that for those who doubt.

I tell a story in PC Hardware in a Nutshell about a guy I know who'd recently bought a high-end Pentium 4 system with an ATA hard drive. He was visiting one day and ended up sitting in front of one of my test-bed systems. The conversation went something like this:

He: "Does this thing have dual processors?"

Me: "No. Why?"

He: "Things just fly up on the screen. When I double-clicked the Word icon just now, Word loaded almost instantly. On my new system, there's a lag of several seconds."

Me: "Nope. Just one processor."

He: "What kind of processor?"

Me: "An 800 MHz Celeron."

He: <strangled noises, gasping, and clutching of chest>

Me: "Of course, it also has a 15,000 RPM Seagate Cheetah X15 hard drive on an Adaptec Ultra160 SCSI controller. Nyuck, nyuck, nyuck."

That system was kind of like one of those undercover police cars that looks like a candidate for the junkyard but has a supercharged 500 HP engine under the hood.

SCSI has more overhead than ATA. Under light load, the lower overhead of ATA means it can provide slightly faster performance, assuming equal rotation speeds, cache sizes, and so on. But when the load increases, the simplicity of ATA claims a penalty in reduced performance. ATA starts out fast, but quickly bogs down. SCSI starts out a bit slower, but doesn't bog.

Which reminds me of the tweaked 1979 Honda CB750F motorcycle I used to own. That was one of the last true superbikes. Several of us with fast bikes used to run them at Pittsburgh International Dragway, which had closed down. My 750F was great fun to run. Its best 0 to 60 time was 2.8 seconds, and that was in first gear (albeit slightly over redline). I've never been in an F-14 as it was catapulted off an aircraft carrier, but the acceleration must be similar. I always braced myself and hung on for dear life, because I was afraid that the bike was literally going to accelerate out from under me.

A guy I worked with had a 1,050cc Honda CBX, which had six cylinders to my four. His CBX looked like it had an automobile engine mounted sideways in the frame. If we both sat at the line and cracked our throttles together, I'd at least match him and sometimes beat him to 60. The problem was, my bike had no top end. It'd get to 135 MPH in fifth gear, and that was all it had in it. As I strained to maintain 135 MPH, he'd still be accelerating away from me.

I loved that bike, but it didn't last long. I may be the only person in history whose bike was stolen by his bank. I moved to Winston-Salem in 1980 in the winter. I wasn't about to ride my bike down in the snow, so I left it stored at a friend's house. My mother dreaded the arrival of warm weather, because she knew I'd be going back up to Pittsburgh to get my bike and ride it home. Several of her friends had had sons killed on motorcycles or left paralyzed or in comas, and my mother had a real fear that I'd be next. 

I didn't help matters any. When she said one time that she lived in dread that I'd be hurt in a motorcycle accident, I replied, "Don't worry mom. I won't be hurt. At the speed I ride, I'd be killed instantly." Which was true. I used to cruise up I-79 at 120+ MPH to visit my girlfriend in Butler, PA. I didn't worry about the state cops, because they couldn't catch me. They didn't have many helicopters back then.

So, one day I was sitting at my parents' house. Because I'd parked him in, my father had taken my Jeep to the grocery store. I had loans on both the Jeep CJ7 and the Honda CB750F from the same bank up in Pittsburgh. The phone rang, and it was the lady from the bank who'd done the loans for me. She said she had some bad news, that the bank had repossessed my Jeep. I told her my dad wasn't going to be happy when he came out of the grocery store to find the Jeep gone. That confused her, so she checked and said she'd been mistaken, that it was my motorcycle they'd repossessed.

Hmmm. I asked why they'd repossessed it, and she said it was because I hadn't made any payments on the loan for several months. I told her that was wrong, that I'd paid every month on time, and that I had the cancelled checks to prove it. She asked me if she could check on something and call back. A few minutes she called back and apologized profusely. As it turned out, they'd gotten some wires crossed somehow and credited my payments to some other account. I told her that was no problem, and that she could simply return the motorcycle. It was a problem, she said, because they'd already sold the bike. Arrrrghhh.

I was outraged that the bank had literally stolen my motorcycle, and for a moment couldn't figure out why my mother was smiling. She was so pleased it was gone that she didn't care what had happened.

As it turned out, the bank got away with it, too. I called an attorney up in Pennsylvania who was the father of a girl I used to play tennis with. He told me it wasn't worth bothering about. They sold the bike for the same amount that was due on the loan, and from what he said I'd have trouble proving any significant damages. My mother was so obviously relieved that I no longer had a bike that I decided for her peace of mind not to buy another. 

And to this day I've never ridden a motorcycle again. When I moved down here, I got the motorcycle endorsement on my driver's license, which I've maintained to this day. But I haven't gotten on a bike since late 1979.

Which is probably a fortunate thing.

More from Dr. Huth on XML in Office 2003.

-------- Original Message --------
Subject: RE: xml
Date: Mon, 20 Oct 2003 11:54:14 -0700
From: Mark Huth
To: Robert Bruce Thompson

Oh ye of little faith.


Natch, Office still has the proprietary formats as the basic format, but XML is very much there and very usable. Is there a catch...well it isn't the native format for the product, and some developers (most notably Gary Edwards (who works for OpenOffice)) have said it isn't a full implementation of XML. Further, not all versions have the XML implementation. I gather that the Enterprise and Professional versions do have it, I gather that standard doesn't. Product isn't shipping yet, and everyone is working off betas.

Further, there is some question about what constitutes the "full" XML implementation, natch. Some reviewers say it is a "full" implementation, others say no. I can't judge.

Our IS guy has been playing with it and is very impressed thus far.

We'll see what the shipping version brings.

Ah, yes. I was aware that the expensive versions supported standard XML. I thought you meant that that had been extended to all versions. Oh, well.

-------- Original Message --------
Subject: RE: xml
Date: Mon, 20 Oct 2003 12:53:21 -0700
From: Mark Huth
To: Robert Bruce Thompson


Little bit of exaggeration in your comments below. Microsoft owes their success not only to their proprietary formats, but also to the fact that they make an excellent productivity suite. We tried OOO last fall. Our transcriptionists rejected it out of hand. Too slow, too cumbersome, lacked features. Lasted 10 days and we removed it. They used to use WordPerfect, but asked to change to Word...feature driven 4 or 5 years ago. I don't spell check, format, template, configure, merge, edit, etc., but our transcriptionists do that every day, all day. Now one can argue that Open Office is much better now, but what exactly is the reason we should look at this again? Because the data format is proprietary? Because we don't like Microsoft? Because we both think that the world would be a better place if Microsoft had more competition? All good reasons, but when push comes to shove not good enough in my environment.

Perhaps we are abnormal and the world is looking for "free" software with open data formats (I've concerns about the long term viability of such software, but...), however we don't think that the cost of Office products is staggering. We've a volume license, I don't remember what we paid the last time we upgraded, but it was a few thousand dollars for Office 2000 for all of our machines. (Note that when we buy new machines we usually get a copy of the latest version of Office with the machine).

Bob, I don't want to be an apologist for Microsoft, but I think we are a pretty typical small office site and I get to see lots of other small office sites in my medical community and I've never heard anyone bitch that Office doesn't do the job, nor have I heard anyone complain that it is too expensive.

I've never disputed that Office is an excellent productivity suite. As I've said, I use it myself. My point was that Microsoft's dominance in productivity suites and thereby in operating systems, is largely due to the proprietary formats used by Office. I remember when Office first shipped. Back then, everyone bought WordPerfect 4.X ($495 retail, dealer cost $173.25) and Lotus 1-2-3 ($495 retail, dealer cost about $216 IIRC). People wanted inexpensive full-featured office software that was completely compatible with WordPerfect and Lotus 1-2-3, and that's the market Microsoft went after. They offered cheap bundles of Office to grab market share. Once they'd marginalized WordPerfect and Lotus 1-2-3, all of a sudden Office wasn't as cheap as it had been. Then Microsoft began intentionally introducing incompatibilities between their operating systems and applications and competing products (remember "Windows ain't done 'til Lotus won't run...?) Once they had a lock on the office productivity segment, they began introducing new and incompatible file formats with each version, which had the effect of forcing upgrades as well as driving their competition out of the market.

Sure, Office is a good suite, but it's not enough better than products like OpenOffice.org that it could sell for hundreds of dollars if it wasn't for the format lock-in issue. If Office weren't bundled and used open data formats, it'd have a small percentage of the market. How small would depend how much Microsoft sold it for. At $25 for the full package, they might hold a quarter of the market. At the current price, I'd be surprised if they could hold 1% of the market. Assuming full compatibility and interoperabillity, how much would you pay for Office?

-------- Original Message --------
Subject: RE: xml
Date: Mon, 20 Oct 2003 15:15:22 -0700
From: Mark Huth
To: Robert Bruce Thompson

Bob, one last thing on costs.

Just asked Ted (our IS guy) to give me a rough idea of what it would cost to buy all of our Microsoft software again, now. He just worked through the numbers. We've 35 users (for the purposes of Office usage). Including the server software and all of the Office clients we'd need to budget "between 9 and 10k". That would include all upgrades for 3 years under our current agreement with Microsoft. Now note, our costs won't ever be that high. That is in part, because we upgrade machines on about a 3 year cycle and they are usually spec'ed with the newest version of office, and in part because our actual cost to upgrade don't come close to that. We're still running NT server which is, what, 4 years old at this point. We've a mix of XP, 2000, and 98 on the desktops...with the last 3 machines running 98 scheduled to go away once Office 2003 comes out. It costs us far more for our firewalls (OpenBSD based), routers, and wired and wireless networking.

My point, well 10k isn't a big cost for us, heck you could quadruple it and we'd still not flinch. We run our entire business

Are you under the impression that the Office license that comes bundled with a new PC is sufficient? If so, you might be surprised. Under most Microsoft volume licensing plans (perhaps all by now), that license you get with a PC doesn't count. You have to buy the software a second time.



Wednesday, 22 October 2003

[Last Week] [Monday] [Tuesday] [Wednesday] [Thursday] [Friday] [Saturday] [Sunday] [Next Week]
[Daynotes Journal Forums] [HardwareGuys.com Forums] [TechnoMayhem.com Forums]
{Five Years Ago Today]

11:10 - Sorry for the scrolling problems those of you who run Mozilla experienced yesterday. I wrote that entry on my main office desktop system, which runs a 19" high-resolution monitor. On that display, the page looked fine. It wasn't until I viewed the page on the 1024X768 monitor on my den system that I realized that page now required horizontal scrolling. The problem was a long URL in one of Dr. Huth's messages. There's a known bug in Mozilla that causes the problem. If you were running IE, you didn't see the problem.

I have always said that one thing IE has going for it is that it renders problem pages sanely. No matter how bad the HTML is, IE usually renders the page in readable form. Opera is the worst at insisting on correct HTML. If Opera renders a page with bad HTML, instead of doing the best it can it simply butchers the page. Mozilla is somewhere in between, but fortunately a lot closer to IE than Opera. Opera's rendering problems with less-than-perfect HMTL were the main reason I pretty much stopped using it.

Here's an interesting article that speculates that Dell outsourcing their support to India is going to cost them customers. The same, alas, can be said for HP. I assume that IBM still uses support staff who speak English as their first language, but I'm not sure of even that. I do know that when people have problems with their computers they expect to be able to call someone who speaks their own language natively, and Indian support operations aren't cutting it.

If I need support, I expect to be connected to someone who speaks American English without an accent. If I spoke Spanish as my first language, I'd expect to have the option to be connected to someone who speaks Spanish as his first language. If I'm French Canadian, I expect to be connected to a Canadian Francophone. That's not unreasonable. In fact, it should go without saying.

In Soviet Russia, there was an old joke that summed up why nothing much ever got done: "They pretend to pay us, and we pretend to work." Well, I have news for Dell, HP, and the rest. If you pretend to support us, we'll pretend to buy your products. And what Dell, HP, and others are doing now is no more than a pretense of support. My sister-in-law's recent experience with HP's so-called support is yet more evidence of that.

So what's the solution? Build your own, if you can. If you can't, establish a good relationship with one of the so-called "white-box resellers" or "screwdriver shops". Those folks are local. They live and die on word of mouth, so they can't afford to provide poor support.

If you're buying one PC for yourself, make it clear to the white-box vendor that you want quality rather than the absolute lowest price. Make it clear that you'd rather have a top-notch power supply than save $10, that you want good name brand motherboard, memory, and other components. You'll pay more, both because top-notch components cost a bit more (although not as much as you might think) and because they may have to order things individually for you. But you'll end up with a more reliable system. And if you do need support, you'll be talking to someone local rather than some ESL speaker in India.

If you need a bunch of systems for your company, specify the exact components you want and put the systems out on bid to several white-box vendors. Include your requirements for support. You may not be able to get 24X7 support, but then you don't really get that from the big-name vendors either, at least unless you're a large company and are willing to pay through the nose. Specify maximum allowable response times, and consider buying a spare system or two to use as drop-in replacements (yes, I know, that never works, because the "spare" system always ends up being used, but it's worth the attempt.)



Thursday, 23 October 2003

[Last Week] [Monday] [Tuesday] [Wednesday] [Thursday] [Friday] [Saturday] [Sunday] [Next Week]
[Daynotes Journal Forums] [HardwareGuys.com Forums] [TechnoMayhem.com Forums]
{Five Years Ago Today]

9:06 - Here's bad news. The US Senate voted 97-0 yesterday to legalize spam. Oh, they didn't phrase it that way, but that's what it amounts to. The US Senate bought into the specious arguments made by the direct marketers that only "deceptive" unsolicited commercial email should count as spam. If Joe's Garage wants to send 10,000,000 spams to advertise its oil-change special, that's perfectly okay under this bill.

The one apparently good thing about the bill is that it will establish a do-not-spam list. Those who submit their email addresses to the do-not-spam list cannot be sent UCE by the "legitimate" spammers as the bill defines them. Of course, that's a two-edged sword, because the do-not-spam list will inevitably be used by "rogue" spammers as a wonderful source of good email addresses.

I only wish they would make provision for blocking entire domains rather than just individual email addresses, but that won't happen. It would be simple enough to check the domain name record to verify that the person who is submitting the entire domain for exclusion is the admin contact for the domain. They won't offer that capability, of course, because the first domains blocked would be ones like earthlink.net, aol.com, rr.com, and so on. The direct marketers wouldn't stand for that, because they know that allowing blocking entire domains would instantly eliminate their ability to spam legally to any significant number of addresses.

The real purpose of this bill has nothing to do with reducing spam. It's an attempt by the DMA and similar organizations to eliminate the competition from what they term spam. They don't consider their own messages spam, although any normal person does. They want to eliminate all the spams for Viagra and penis-enlargement and Nigerian deals to reduce the clutter so that their spam becomes visible again.

Although the bill proposes million-dollar penalties and jail time for sending what it defines as spam, the result is likely to be more spam, not less. Companies that have refrained from spamming will likely begin spamming once it is legalized. And the current big-time spammers are unlikely to be frightened off by the fines and jail terms mandated by the bill. They'll simply disappear like smoke, simply to return under new names in other countries. And they'll no longer have to worry about keeping their mailing lists up to date. The government will be doing that for them.




Friday, 24 October 2003

[Last Week] [Monday] [Tuesday] [Wednesday] [Thursday] [Friday] [Saturday] [Sunday] [Next Week]
[Daynotes Journal Forums] [HardwareGuys.com Forums] [TechnoMayhem.com Forums]
{Five Years Ago Today]

10:36 - I awakened early this morning, and Barbara encouraged me to go out and look at the night sky. At 6:15 a.m., it was magnificent, with Jupiter and Saturn blazing away, the pentagram of Auriga the Charioteer nearly overhead, and mighty Orion the Hunter well up in the south. Five of the ten brightest stars in the heavens were visible: Capella (#6) in Auriga, Rigel (#7) and Betelgeuse (#10) in Orion, Procyon (#8) in Canis Minor, and Sirius (#1) in Canis Major. If I'd had a clear southern horizon, I might have been able to spot a sixth, Canopus (#2) in Carina. If I'd been out a bit earlier I'd have also seen Vega (#5) as it set. Summer is the time for galaxies, but the winter sky has more than its share of the brightest stars and most interesting constellations and deep-sky objects.

One of the prices we pay for urbanization and light pollution is the loss of the night sky. There was a time, not all that long ago, when most children could identify at least some constellations and stars. Nowadays many children grow up never having seen the night sky, and are unable to identify even the Big Dipper. We astronomers complain bitterly about light pollution, but it seems that our objections fall mostly on deaf ears.

Things may be looking up, though. There is increasing evidence that light pollution has very real physiological effects on us, none of them good. Humans evolved in an environment that alternated brightness and darkness, but now we're exposed to brightness 24 hours a day. An increasing body of evidence suggests that this unnatural constant light has deleterious physical and psychological effects that are only beginning to be recognized.

The worst thing about light pollution is that most of it is entirely gratuitous. For example, it costs no more to install full cutoff streetlights, those that are shaded to direct all of the light they generate downward rather than in all directions. Light that is directed above the horizon does nothing except brighten the night sky. It contributes to glare, and uses electricity unnecessarily. The US wastes literally billions of dollars a year to pay for the electricity used to light the night sky. That benefits no one.

Every study ever done on the subject has concluded that bright street lighting without full cutoff shading is actually counter-productive. The glare and high contrast it produces makes it harder to see things, not easier. The brightness and glare causes people's pupils to constrict, which means they can see only objects near the lights. Those areas away from the lights become inky pools of blackness in which nothing can be seen. If the goal is to improve nighttime visibility, the answer is to use dimmer lights with full cutoff to produce a dimmer but more evenly-lit environment. And that would have the additional inestimable advantage of allowing us to see the night sky again.



Saturday, 25 October 2003

[Last Week] [Monday] [Tuesday] [Wednesday] [Thursday] [Friday] [Saturday] [Sunday] [Next Week]
[Daynotes Journal Forums] [HardwareGuys.com Forums] [TechnoMayhem.com Forums]
{Five Years Ago Today]

12:12 - Business Week magazine has declared the microprocessors war over, basically saying that AMD is toast. I hope they're wrong, but I fear they're right.

The original Athlon was a wake-up call for Intel. The Athlon outperformed Intel's then-flagship Pentium III processor clock-for-clock and, for the first time ever, provided an alternative processor than not only matched but actually exceeded the floating-point performance of Intel's best. The Athlon pushed Intel into releasing the Pentium 4 before its time. A lot of industry observers scoffed at the Pentium 4, which in its initial incarnation was actually slower than the Pentium III it replaced. But, as I said at the time, the Pentium 4 was very, very bad news for AMD, because the Pentium 4 had all kinds of headroom whereas the Athlon core was reaching its limits. That allowed Intel to regain the performance crown.

More importantly, because of the architectural differences, the Intel Pentium 4 ran at much higher clock speeds than the Athlon. Most of my readers know that clock speed doesn't mean anything when comparing different architectures, but in the all-important consumer market clock speed is king. AMD tried to counter Intel's clock speed advantage with their hokey PR system. That helped a bit, but as the Athlon fell further and further behind the Pentium 4, not just in clock speed but in real performance, AMD processors were again relegated to bargain systems. AMD could sell a lot of processors, but the all-important ASP (average selling price) kept falling lower and lower. Intel could make more profit selling one high-end Pentium 4 than AMD could make selling a dozen Athlons.

AMD's 64-bit processors, the Opteron and Athlon 64, might have saved them. They still might. But AMD's twin curses have always been execution and marketing. The Opteron and Athlon 64 processors might be the best processors in the world, but if AMD can't produce them in volume they aren't going to help the bottom line. Right now, the new AMD processors are scarce. AMD, of course, claims that's because the demand is huge and outstripped their predictions. That may even be true. But informed speculation has it that AMD is having severe yield problems, which means they can't produce enough of their new processors to meet even predicted demand. If the latter is true, and I think it is, AMD is in really deep trouble.

The other aspect is marketing, and AMD is doing their typical bad job of that. Of course, it's not all their fault. It's pretty hard to convince people to pay a premium for 64-bit processors when there's no mainstream 64-bit operating system shipping and 64-bit applications are as scarce as the proverbial hens' teeth. The phrase "a day late and a dollar short" comes to mind. AMD is hideously late to market with their new processors. They should have released them at least a year before they did, and 18 months would have been better. Instead, they're playing catch-up, and they're doing it with a severely depleted bank account.

About the only bright spot on the horizon for AMD is that Intel is having some teething pains of their own with their new Prescott-core (Pentium 5?) processors. If Intel had been able to execute as planned on Prescott, AMD would certainly have been toast. As it is, AMD has a window. A very small window, but a window nevertheless. Whether or not they can take advantage of it is the question. If they do, they have a fighting chance to remain a viable contender in the processor wars. If not, they're history.

How things change.

Not all that long ago, Barbara was convinced she'd never be able to find faint fuzzies with our telescope. A year or so ago, she was trying to find the Little Dumbbell Nebula, Messier object M76, a planetary nebula in the constellation Perseus. She worked on it for what must have been half an hour. Planetary nebulae in general and M76 in particular are small, relatively dim objects and Barbara thought that she might have gotten it into the eyepiece and just couldn't see it. So I walked over to the scope and spent all of 15 seconds getting it into the eyepiece (it's about halfway between the bright stars Ruchbah in Cassiopeia and Almaak in Andromeda, and less than a degree from the 4th magnitude star Phi Perseii) so that Barbara could see what it looked like. She said something like, "I'm never going to be able to find stuff like you do."

So, fast-forward a year. Barbara can now knock out Messier objects like an old pro, and we're both now working on the Herschel 400 list, which comprises a lot of objects that are much harder to find (and see) than the Messier objects. Many of the Herschel 400 objects are hard to see because they're dim. Others are hard to find because they're not close to any bright stars. Still others are hard to find and/or see because they're in the midst of very rich star fields in the Milky Way.

Barbara and I share our 10" Dob telescope. I had just bagged and logged NGC 7606 (Herschel H104-1), a galaxy in Aquarius, located near the triple star Psi Aquarii. Although it's the brightest galaxy in Aquarius, that's not saying much. Its magnitude is a relatively bright 10.8, but its surface brightness is about 14.4, which means it can be elusive.

After I logged the object, as usual I moved the scope so that Barbara could locate the object on her own. As I walked over to the table to record my observation, I said something like, "You might have a bit of trouble with that one." Barbara's response? "You found it. How hard could it be?"

How things change...

18:35 - I just sent the following message to subscribers:

This is no joke. Attempting to install Mandrake Linux 9.2 on a system that has an LG Electronics CD-ROM drive destroys the drive. Note that you needn't attempt to install the software from the drive. Even a network install renders the drive, as Mandrake puts it, "physically dead". For more details, see:


Although Mandrake mentions specifically CD-ROM drives, it is possible that other LG Electronics optical drives, including CD writes, may be affected by this problem. Note that LG Electronics is a major OEM of optical drives that are relabeled by other companies. In particular, Dell systems are known to use relabeled LG Electronics drives, as are some HP, IBM, and Compaq systems. Many house-brand and retail-boxed optical drives are also made by LG Electronics but have other names on them.

Until there is more data available, I recommend that you not attempt to install Mandrake Linux 9.2 on any system.




Sunday, 26 October 2003

[Last Week] [Monday] [Tuesday] [Wednesday] [Thursday] [Friday] [Saturday] [Sunday] [Next Week]
[Daynotes Journal Forums] [HardwareGuys.com Forums] [TechnoMayhem.com Forums]
{Five Years Ago Today]




Copyright 1998, 1999, 2000, 2001, 2002, 2003, 2004 by Robert Bruce Thompson. All Rights Reserved.