Week of 3/1/99

This month marks the first anniversary of this web site, and I'm getting ready to move it to a different web hosting company, probably pair Networks. I've been with BigBiz.com since the beginning. They did a good job originally, but I've had nothing but problems with them since the first of the year.

Yesterday was the final straw. It took me 1.5 hours to publish my changes. I don't mean that I tried several times over the course of an hour and a half and finally succeeded, mind you. I mean that I started the publishing process and it ran constantly for 1.5 hours solid before it finally finished publishing. Throughout this process, I was constantly expecting FrontPage to time out on me, but it never did.

I then retrieved the Daynotes page from the server to make sure the changes had in fact been published. It took literally five minutes to load the page, and my browser showed throughput of less than 100 bytes/second. This when I was able to call up pages from other servers with normal performance.

Well, I'll try to publish this now, although I don't have any great confidence that I'll succeed. Updates may be a little sporadic until I get my site moved to another service provider. In the interim, I appreciate you bearing with me.

And then I'd better get back to work on my current chapter...

Tuesday

Tuesday, March 2, 1999

I've heard from many of you to tell me that my pages are loading very slowly. In one instance, someone called up my current daynotes page from a bookmark, watched the blank screen for a minute or more, and finally decided to go take a shower. When he returned from his shower, the screen was still blank. He was about to close the window when the page finally finished loading. It had taken, by his estimate, at least 15 minutes. I myself timed the page load the other day, and it took 8.5 minutes. This is ridiculous.

I'm keeping these updates very short for a while because I'm having so much trouble getting them published up to my server. I'm also having some trouble getting email, so if you've mailed me and haven't gotten a response, please send it again.

I signed up yesterday morning with pair Networks to host my web site. At the end of the sign-up process they told me that they'd create my account within two to three business days. At that point, I'll have a web server I can publish to, but it will be accessible only by IP address. I'll send the Modify form to InterNIC to change the primary and secondary DNS servers, but that will probably take them at least several days to get done. Once that happens, the change will take a while to propagate through the Internet. The upshot is that it may be a couple weeks before you can access this web site by name on its new server. I'll do the best I can to keep the original server updated as well, so at some point you will simply hit the new server instead of the old one when you point your browser to http://www.ttgnet.com.

And it's not just me. My friend and future co-author, Paul Robichaux, hosts his web site at BigBiz. He's also been having the same kinds of problems that I've been having, and they also moved his web site from Server01 to Server05 some time ago. He told me yesterday that he was unable load his home page in his browser at all. He signed up with pair Networks about an hour before I did. I feel particularly bad about that, because Paul originally signed up with BigBiz on my recommendation.

This mess is costing us not just aggravation, but readers. I noticed that my stats are down, and I can only attribute that to the fact that people aren't willing to wait several minutes for a page to load. And I can't blame them. Please bear with us. We're doing everything we can to get the problem fixed.

Wednesday

Wednesday, March 3, 1999

Paul Robichaux mailed me yesterday afternoon to tell me that his account was already set up at pair Networks. I signed up with them about an hour after he did, so I was hoping that meant I'd get my new account information from them soon, but I haven't so far. Well, they said two to three business days, so I can't complain.

I've gotten mail from many people telling me that my site was responding slowly or not at all, but I've also gotten quite a few messages telling me that they weren't having any problems accessing it. One even said that my site was the fastest of any he visits. I think the problem is intermittent, because at times I've been able access my own site normally, and at other times I can't even get the home page to load. Same thing when I POP my mail. Sometimes it transfers quickly, and other times it takes several minutes to even make a connection or it times out. And I don't think it's a connectivity issue. I've done PINGs and TRACERTs each time, and the network is performing normally. The problem seems to be simply server load. What is particularly aggravating is that I've sent four email messages to BigBiz during this latest problem period, and I haven't received a response to even one of them. This over a period of a couple of weeks.

But the amount of traffic to this site is growing fast, and I can't continue to have it hosted by a service that has this kind of problems. One of the reasons I picked pair Networks is that they host some of the biggest sites on the Internet, including Tom's Hardware. They have something over 100 web servers, and connectivity out the wazoo. If they can handle sites like Tom's Hardware, which recently passed the 1,000,000 pages/day, it can sure handle my site at less than 1% of that load.

And speaking of Tom's Hardware, I decided to buy his book. I've read through the thing now, and I can't believe how bad it is. What I was expecting was a technically competent, insightful, no BS book about PC hardware. What I found was an error-plagued, superficial book that's not of much use to anyone, novice or expert. It really made me wonder why anyone pays any attention to Mr. Pabst. Here are just a few amazing facts I found in Tom's Hardware Guide:

30-pin SIMMs are 16 bits wide, and so must be installed in pairs in 386 and 486 systems. [30-pin SIMMs are 8 bits wide, and must be installed four per bank in 386/486 systems, and eight per bank in the few Pentium-class systems that use them.]
DAT stands for Digital Analog Tape [DAT stands for Digital Audio Tape, and anyway the proper terminology for a "DAT" tape drive is DDS, not DAT]
xDSL throughput varies according to how heavily other users on your xDSL line are transferring data [xDSL is a point-to-point service, and does not use shared media]
"A 7GB DAT cartridge costs only about $12 per unit, for an incredible value of 597.3MB per penny. That's half a gig for each penny you spend." [When I do the calculations, I get (7 GB*1,024 MB/GB)/1200 cents = 5.97 MB/penny. This isn't isolated, either. Anywhere Mr. Pabst does math, you'd best check it yourself.]
Most telephones use Category 2 cable [there is no such thing as Category 2 cable, and never has been]
No video is available during a flash BIOS recovery procedure [Mr. Pabst apparently doesn't realize that this is true only for PCI video cards, which is an excellent reason to keep an old ISA video card around.]
TCP/IP doesn't do "transfer checking" and is therefore less reliable than IPX/SPX. [Huh? From his garbled explanation, it appears that Mr. Pabst is referring to UDP rather than TCP. What Mr. Pabst doesn't know about network and transport layer protocols could fill a book.]

And it goes on and on. I don't normally trash competitors books, and every technical book I've ever seen has at least a few errors, but this one has so many mistakes that I felt I should say something. For now, if you want a good hardware book, buy Scott Mueller's Upgrading & Repairing PCs. Steer clear of Tom's Hardware Guide.

* * * * *

I've not been posting much mail the last several days, but I thought posting this message from Timothy Werth [timothy.werth@eds.com] was worth the small additional risk that my update will time-out because of the page size:

Don't know if you have seen this yet or not. I thought you would find it interesting, I certainly did.

-----Original Message-----

I am participating in an Internet campaign to stop a regulation which would require your bank to spy on you, and I'd like to invite you to join me. We now have less than 20 days to contact the FDIC and demand that it kill its proposed "Know Your Customer" rule. Please forward this message to any friends, family, co-workers, neighbors, or other people you know who may be interested, then go to http://www.defendyourprivacy.com/ and sign the petition. It will be submitted directly to the FDIC. Plus, a copy will be sent to your representative in the U.S. House and to both your U.S. Senators. The FDIC's Know Your Customer rule would force banks to "monitor" your checking and savings account and report any "unusual transactions" to the federal government. This frightening threat to your financial privacy would force your bank to:

* Discover your "source of funds"

* Determine your "normal and expected transactions"

* Report any "suspicious activity" to federal investigators

The government claims it is trying to thwart money launderers and drug dealers. But what this law will do is turn every bank teller into a government informer and everyone with a bank account into a criminal suspect.

In a free society, the government has no business asking where you get your money or how you spend it -- and politicians have no right to force your bank to monitor your account. But that's exactly what's going to happen, unless we can generate enough opposition before the FDIC's comment period expires on March 8. Outraged Americans have already flooded the FDIC with over 20,000 comments against the Know Your Customer regulation -- but the agency hasn't backed down yet.

Let's keep up the pressure. Please forward this e-mail to everyone you know who might be interested in helping, but please don't send it indiscriminately -- spam will only hurt our campaign. Then go to http://www.defendyourprivacy.com/ and sign the petition.

When I first hit this web site, I'll admit that I was concerned. They ask you to provide your name, snail-mail address, and email address. My first thought was that this might be a site run by a scum-sucking spammer to gather solid data on a lot of people. So I did some investigating. I hit InterNIC to find out who the domain was registered to, and found that it belongs to the Libertarian Party. You can trust those guys. You might not agree with all of their positions on issues, but they're not going to turn around and sell your information to a spammer. (I used to be the Finance Vice-Chairman of the Libertarian National Committee, so I know they're trustworthy.) As it turns out, if I'd just scrolled down to the bottom of their home page, I'd have seen there that this is an operation of the Libertarian Party.

I just added my voice to the petition, and I'm encouraging all my readers to do the same. Enough is enough. Although they make reference to only 20 days left, it appears to me that the comment period ends March 8th, so if you're going to do it, do it now.

* * * * *

And, from Chuck Waggoner [waggoner@gis.net], here's another very disturbing piece of information about violations of privacy.

Discouraging article in today's NYTimes on the Web.

It suggests that computer privacy is already non-existent, and may one day even become illegal.

Would like your comments.

Well, I can't say I'm surprised, although I wasn't aware that Microsoft applications were embedding my Ethernet MAC address into the documents they created. That's an outrage, of course, and the rationale given in the article is totally bogus, but I don't know what to do about it. I could stop using Microsoft applications, I suppose, but short of writing my own word processor I'm not sure how I'd know what some other application is doing.

I did find it ironic that before I could access the article I was required to supply various personal information, including my email address. I did as I usually do and gave my name as anonymous, my email as anonymous@anonymous.com, etc. Quaintly, the username anonymous had already been taken, but the NY Times web server suggested anonymous75 instead.

Privacy, electronic or otherwise, really has only two enemies. Unfortunately, they're government and business, which is a pretty formidable pair. Much though I abhor Scott McNealy's position, "You already have zero privacy -- get over it," I'm afraid he's right. All governments and many businesses are determined to violate our privacy, and I don't see any solution to the problem short of hanging them all from lampposts.

This latest example is bad, but it's by no means the worst. It's bad enough when organizations are up-front about violating my privacy, as is often the case with government and sometimes with businesses. It's worse still when they violate my privacy sneakily, as in this case. But the worst of all situations is when they pretend to make provisions for protecting my privacy and then violate it behind my back.

A good example of this last is the phone company and their Caller-ID. Most people think that per-line blocking or per-call blocking prevents their information from being sent to the number they're calling. It doesn't. What it does do is instruct the switch to toggle one bit in the string it sends. Toggling that bit instructs consumer-grade Caller-ID devices not to display information for that caller. But Caller-ID information is sent as a simple ASCII string, and, for anyone with a modicum of knowledge, it's easy enough to grab the raw data. So much for that supposedly blocked call.

If you can get anyone at the phone company to even admit this, they'll tell you it's done that way so that 911 emergency services can get the information even on a per-line blocked call. That's totally bogus, of course. It would be easy enough to program a switch to simply not send Caller-ID information for blocked calls, but doing that would provide actual privacy to people, and that's not acceptable. Blocked calls directed to 911 could be programmed to deliver the caller ID information, thereby putting that red herring to rest.

And the phone company never mentions the similar ANI (automatic number identification) technology. This is used by large mail-order retailers like LL Bean to know who's calling them before they even answer the phone. This information is delivered out-of-band via digital trunks. When that nice lady at LL Bean picks up the phone, she already has your record up on her screen. And that's whether or not you have call-blocking in effect.

All of this sucks, of course, but I just don't know that there's anything to be done. I may be tilting at windmills, but lost cause or not I'm going to keep protesting at every opportunity.

Thursday

Thursday, March 4, 1999

I know that I probably bore people talking about the variable weather here in North Carolina, but yesterday reached a new pinnacle of weirdness. In mid morning, our indoor/outdoor thermometer showed an outside temperature of 73F/23C. After lunch, we had winds gusting to 40 MPH and a strong thunderstorm. When I looked out the window at 3:48 p.m., it was snowing heavily. The low last night was about 20F/-7C.

* * * * *

My account information from pair Networks showed up yesterday evening. I immediately called my friend, Steve Tucker, who had just gone through getting his account set up with pair. With his help, I got my mail forwarding configured, which is the most important thing. As the nameserver changes propagate through DNS, some SMTP servers will be delivering mail to mail.ttgnet.com at the old BigBiz IP address, and others to the new IP address at pair Networks. Until the change propagates fully, I'll be getting some mail at each place, so I needed to get forwarding set up at pair so that everything didn't end up in my main mailbox.

Getting my web site moved over will be aggravating, through no fault of pair Networks. The problem is Microsoft FrontPage. As usual, its lack of configurability is going to cause problems. Right now, my <Root Web> is located on a server at BigBiz, and maps to www.ttgnet.com. I have a copy of this root web on my local hard drive. Ideally, I should be able to just publish that local copy to each location. But I don't think FrontPage will let me do that.

Actually, it will let me do that, but only by specifying the IP address for the new server (or by specifying its physical location as something like www109.pair.com/ttgnet, which I don't want to do. FrontPage is badly in need of some .ini files or something.

The other aggravating thing is that pair Networks sent me my main password and FrontPage password. They suggested I change the main password immediately, which I did. But there's no apparent way to change the FrontPage password. I could live with that, but for two interrelated problems: (1) pair Networks assigned my FrontPage password for me, and it's the usual long meaningless string of mix upper- and lower-case letters and numbers, and (2) FrontPage has no way that I can find to save the password locally. You actually have to type it in each time you open FrontPage to publish. Geez.

* * * * *

pair Networks has some kind of "Platinum" deal with InterNIC that supposedly gets faster service for requests from pair. Well, it appears to work as advertised, because I just got notification this morning that the changes to the domain record had been made. That meant that getting my web site moved over to pair Networks became increasingly urgent, because before long some users that hit www.ttgnet.com are are going to be directed to the web site on pair Networks. As of this morning, there was nothing there but a generic banner page, so I needed to get everything published to the root web on pair Networks.

Because I'm essentially setting up a new web site and would have to publish every single file up to pair, now was the time to do all the reorganization that needed done. My web site on BigBiz "jest growed." The site comprises hundreds of files and about 10 MB. Many, perhaps most, of those files were in the root directory. That needed fixed, so here's what I did:

First, I called up my local copy of the web in FrontPage 98. I then "published" it to another local copy in a different directory. After exiting FrontPage and restarting it, I called up the local copy of the new web. I made a bunch of organizational changes, e.g. moving all the weekly daynotes files out of the root folder and into the \rbt\daynotes folder. I also renamed the few files that had .htm extensions to .html. Once all that was done, I did a Recalculate Hyperlinks to make sure the local copy of the root web was internally consistent.

Once all that was done, I held my breath and published the local copy up to the new web site at pair, using the new IP address (216.92.40.142) rather than the site name, which still points to BigBiz. Publishing to pair versus BigBiz was like the difference between night and day. I published the full site--close to 100 folders, nearly 2,000 files, and about 8 MB--to pair in less time than it took me yesterday to publish exactly three files to BigBiz, one 4 KB file and two 32 KB files.

All these changes mean that bookmarks you've set may no longer be valid. For example, thisweek.html used to be in the root folder, and is now in the \rbt folder. I apologize for the inconvenience, but the reorganization really needed done, and now was the time to do it.

For now, at least, you can get to the home page of the new site by using the IP address, http://216.92.40.142/ or the underlying server name, http://www109.pair.com/ttgnet. Once the changes propagate through DNS, using the URL http://www.tttgnet.com should work fine once again because the CNAME record that aliases the machine name www.ttgnet.com to the underlying IP address will work. At least, I sincerely hope so. I may end up having to republish the whole site to http://www.ttgnet.com rather than http://216.92.40.142/ in order to fix internal FrontPage references. I hope not, but I'll do it if I have to.

I'm going to put up a notice on the home page on the BigBiz copy of the site to point people to the new site. I'll leave that up through mid-month, which should be plenty of time for the DNS changes to propagate. I was considering doing a redirection page on the old site, but I think posting a notice about what's going on is more useful.

* * * * *

Okay, here's the kind of thing I was talking about where an advertising-supported hardware web site gives (to me, at least) the appearance of not being impartial. Check out the Kryotech Cool K6-3 500 review on AnandTech. Anand praises this product with faint damns. And to me, this product is simply a stupid idea with absolutely no redeeming features. Basically, it consists of a refrigerator built into a PC case. The idea is to keep a K6-III CPU cold so that it can be overclocked.

But the thing costs $1,250 for the case with an ASUS motherboard and a K6-III CPU, which is a whole lot more than the motherboard, CPU, and a decent ordinary case would cost you. That would be okay, except that this whole contraption yields a speed boost that most people wouldn't even notice. I mean, this thing gives you 500 MHz performance, but you can go out and buy a plain AMD K6-III/450 and run it without all the extra cooling. Is a 50 MHz/11% speed increase worth all that extra money? If it allowed you to run the CPU at 1 GHz, that might be a different story. Some people would find it worthwhile to spend that extra money to more-or-less double the speed of the K6-3. But even then, for less money you could get approximately equivalent overall performance from an Intel processor without using all the kludges. What is the point of this product? I sure can't see one.

If I reviewed this product, I'd do so in one short paragraph. Something like "You would have to be a complete idiot to buy the Kryotech Cool K6-3 500. It's very expensive and provides only a barely noticeable performance boost relative to a system using a plain vanilla K6-III. And you can spend much less money and get a faster system in the bargain by buying an Intel-based system. For many applications, even a system using an Intel Celeron CPU will be faster than this so-called high performance system. Steer clear of the Kryotech Cool K6-3 500."

If anyone takes the time to read that review and look at the benchmarks, I'd appreciate you telling me whether you agree with me or not.

Friday

Friday, March 5, 1999

Well, the move to pair Networks seems to be going well, with only the minor problems that can always be expected with such a move. InterNIC sent me a notification yesterday morning that the ttgnet.com domain name record had been changed to reflect the move to pair Networks, and the change finally showed up on whois this morning. That change will gradually take effect as local DNS servers stop using the old ttgnet.com IP address from their cache and refresh it from the new nameservers at pair Networks.

And that brought home one of the problems. Most people think that SMTP works on a store-and-forward basis, sending your outgoing mail messages to intermediary servers until they finally reach the destination. That's not how it works. When you send an email message, your SMTP server opens a direct session with the destination SMTP server and transfers your message directly to it. And because my DNS information is in a state of flux, that means that some SMTP servers are sending mail destined for ttgnet.com to the BigBiz server, and others are now sending it to the pair Networks server.

That's how I found out that something wasn't right. Barbara uses the mail address barbara@ttgnet.com. At BigBiz, I had that address set up to autoforward to our actual ISP account, which is thompsrb@bellsouth.net. She POPs her mail from there. When I got my account information from pair Networks, the first thing I did was go in and create a .mailproto file to autoforward mail for Barbara and a couple of other people who have ttgnet.com mail addresses. I ran the program that converts the ASCII text file to a .forward file, and everything appeared to have worked properly. I was pretty sure I'd done everything right, because I used my friend Steve Tucker's ASCII file as a template, and his forwarding is working fine.

But this morning I found some of Barbara's list serve mail in my mailbox. That's a minor aggravation, but I do want to get it fixed before all her mail starts coming to me. I sent mail to support@pair.com and asked them if they could just fix it for me. We'll see what happens.

* * * * *

This from Shawn Wallbridge [swallbridge@home.com]:

I would have to agree with your take on the Kryotech Cool K6-3 500.

Here is my take.

Kryotech should be pushing their Alpha boxes to the High End NT users. When you are paying $5-6k for a machine, tacking on an extra $800 wouldn't even be noticed. When you have complete systems shipping for less than your kit you are after the wrong market. With the constant increases in performance I can't see the desktop market being a viable option. I am not to sure how often Compaq releases a new Alpha chip, but I would think that they would have a much slower release cycle. I realize that Intel basically killed any plans KryoTech had for the PII or PIII, but maybe the PPGA Celeron's will offer chance of cooling.

I paid $380(CAN) for my PII 300 SL2W8 that I run at 450. My other machine (Celeron 300A@450) cost less than $1200(CAN) to build. I can't see them selling too many kits. Especially now that the PIII 500 is out. Now a 1GHz Alpha would be nice :)

I don't understand why anyone in his right mind would spend the extra money on a refrigerator when he could do so many other things with it that would give him more bang for the buck. Before I spent money on a Kryotech cooler, I'd use it to buy more memory, a bigger/faster hard drive, a bigger monitor, or any number of other things that would provide more real benefit. And under NT, that money would be much better spent on a dual-processor motherboard and a second processor.

* * * * *

This from Adam Lieber [aslieber@Princeton.EDU]:

I agree that Anandtech's review of the Kryotech unit was amiss. In no way would I ever call that unit "revolutionary." I would have to say that many if not most of the articles there are, to be charitable, of limited use. This is because they do not take into account that the intended use of the product. One time, I went looking for a cheap motherboard that had sound and video integrated into it. I found one, but it got low marks and the review spent most of its space on the lack of expandability and flakiness when overclocked. I never did see any mention that the board would have been good for it's intended purpose- a cheap, stable no frills system.

I do frequent that site, however, because it seems to have the best variety of reviews that around. I don't rely on the advice at all though. It just informs me that a product exists, and then I do my own research on it. If I have to put up with the ads and the ramblings of a teenager (although I am one myself), so be it.

From the one review of yours that I have read, concerning the EIDE Raid product, I do hope that you choose to write more, because I have never read anything better. Keep up the good work.

Well, in one sense I can't complain about Anand's giving good reviews to products, because I also review only products that I like. That is to say, unless I'm pretty sure I'm going to like a product, I don't even request an eval unit. There's too much stuff out there to waste time reviewing products that I suspect will turn out to be poor or mediocre. That said, if I get a product and it turns out to be a dog, I won't hesitate to say so. Before I post a bad review, I will give the vendor a chance privately to tell me why I'm wrong. In the past, I've sometimes found that it was my own ineptness rather than a problem with the product itself that caused the problem.

I don't have any problem with the fact that Anand is a teen-ager. I know that old guys like me (I'm 45) sometimes take the position that these young whippersnappers have so little experience that they couldn't possibly have anything useful to say. But this industry changes so fast that no one has more than five years of experience. I just happen to have that five years of experience five times. And that is useful in the sense that it gives me perhaps a bit more perspective on things. I've seen it all before, usually more than once, so I tend not to get excited as readily as someone for whom everything is new.

And I do plan on doing many more reviews. In fact, I'm testing a bunch of stuff now. Finding the time to write formal reviews is the problem.

* * * * *

This from Robert Morgan [robert@morgan.pair.com]:

I have to agree that the Kryotech's should win an award for 'Pointless Use of Technology'. And is refrigeration even technology today? Sheesh, the award should be Pointless Use of A Refrigerator.

Anand's always been a very weak type of person. "I'm sorry the update is late", "I take the blame for the server being down again", "Bear with me", blah blah blah.

However, he doesn't recommend the system and is even doubtful about future versions. As far as Anand goes, that's a negative review. The performance graphs tell all: most of them have various Celerons as fast or faster than the Kryotech; failing that, the new PIII's are on top. It should go without saying that any overclocked system anyone puts together today will be beaten by an off-the-shelf system in a matter of months. The sad fact is that this overclocked system is already being beaten by off-the-shelf systems.

Yes, he does tend to gush, but then he is only 17 years old. Perhaps his constant mea culpa stuff is culturally influenced. And you're right that he doesn't recommend the system. I guess what struck me was that he seemed to be looking for nice things to say about a product that doesn't appear to have anything at all to recommend it. I wouldn't have reviewed the product in the first place. If for some reason I had, I'd certainly not have written an eight or ten page report about it. I mean, how long does it take to say, "This product is a loser. Don't buy it."

* * * * *

I think I'll go ahead and publish this now. One of the very nice things about the move to pair is that I can publish any time I want to. With BigBiz, it's always been tough to publish successfully any time other than early morning. Lately, it's been difficult to do it even then, and next to impossible any other time of day. When I posted the final update to the BigBiz servers yesterday afternoon, it literally took more than an hour and a half to post three updated pages. With pair, it takes a couple of minutes any time of day.

* * * * *

Late Afternoon: I haven't gotten much done today. I wasn't feeling very well, so I just went back and took a nap for a couple of hours. When I arrived back at my computer I found a bunch of Barbara's mail waiting for me, along with a response from pair Networks support suggesting that I edit my .procmailrc file manually to change all occurrences of ^FROM_DAEMON to ^FROM_MAILER. According to them, "procmail will not forward what it deems to be "system-generated" mail, because of the increased risk of mail loops that can flood your mailbox with thousands of messages and load down the server." I made the change, and it seems to work. At least Barbara has now gotten her first list serve message directly on her machine from pair, so that may have solved the problem.

The second message from pair Networks was a nasty shock. It was an invoice, saying that they'd charged my credit card for $234.97. I knew that it was supposed to be for $194.10, so I checked the saved .html page I had from the sign-up process. Sure enough, $194.10 was the correct amount. The problem seems to be that they applied only an 8% discount, which is their usual discount for pre-paying a full year, rather than the 24% discount that they offer as a monthly special for new sign-ups during March. I fired back a nastygram with the .html page attached, telling them to please issue me an immediate credit.

Then I mailed Paul Robichaux, who signed up with pair Networks the same day I did, also for a full year. He says his invoice shows the 24% discount, and is for the correct amount for the full year, but they have his account showing as paid through October instead of next March. I'm pretty pleased with pair Networks so far in a technical sense, but their billing department seems to be zero for two so far.

* * * * *

Early Evening: The domain changes appear to be propagating through DNS successfully. All of my mail and Barbara's for the last couple of hours has been coming through pair rather than BigBiz. I just tried hitting http://www.ttgnet.com from my own machine, and it brought up the new site. Up until dinner time, it had been bringing up the old one, so apparently BellSouth's DNS servers hit the root servers for an update. If you are using the numeric address to access this site and haven't tried using http://www.ttgnet.com lately, it'd probably be worth giving it another try.

My first hint that my local DNS was now pointing the the new server was when I used Outlook to retrieve my mail after dinner. I have two services set up, one pointing to the numeric identifier at pair Networks and the second pointing to mail.ttgnet.com. When Outlook tried to hit the second POP server, it blew up with an error message because it was trying to log on to mail.ttgnet.com at pair Networks using the account name and password from BigBiz. I reconfigured that one to point to the numeric address at BigBiz. Of course, it timed out, as everything at BigBiz has been doing for me lately.

* * * * *

And speaking of Outlook, Bo Leuf is looking for feedback from Outlook users on his site.

* * * * *

I sent the following message to BigBiz.com just now:

Please discontinue service immediately for my domain, ttgnet.com.

Since the beginning of the year, your service has been completely unacceptable. My mail has been delayed or not delivered at all. My readers have told me, and I have experienced myself, that small pages often take several minutes to load. Some time ago, you moved me from server1 to server5, saying that that would cure the problems. Instead, it has made the problems worse. It has been nearly impossible to publish my pages to my server.

Over the last two weeks, I have sent four email messages to you about these problems, and have not received even the courtesy of a response to any of them. I have been more than willing to allow you time to fix the problems, but I finally found it necessary to relocate my domain. This has cost me money. Worse still, it has cost me time that I shouldn't have had to waste. The worst thing is that it has cost me readers.

I recently received invoice #23410 from you for service from January 1 through March 31. Given the lack of service that you have been providing to me, I find it amazing that you have the nerve to bill me. Please issue an immediate credit for all amounts billed.

Saturday

Saturday, March 6, 1999

I see that Pournelle is back from his dinner with Bill Gates and his day-long visit to Microsoft. He's upset because the folks over on WebPagesThatSuck.com are trashing his site in a discussion thread. Mr. Heinlein commented about critics in several of his later books, and these comments apply here in spades. Creating is hard work. Editing is easier, but still hard work. Any idiot can criticize, and many of them do.

Pournelle makes a salient point. Few of those who have posted criticisms of his site appear to have web sites of their own. Most of those who do list sites have tacky personal sites on AOL and similar places that provide cheap or free personal web space.

One such site is that of Susan Miller, who criticizes Pournelle's site by saying "Looking at this site from a design perspective was painful. Poor alignment, poor use of colors and that boring font! It made me totally uninterested in the content." That's some criticism coming from someone who last updated her own site 7 February, and whose apparent idea of good content is a banal collection of family photographs and the usual pathetic collection of links.

And if the absence of interesting content isn't enough, her pages suck at least as badly as Pournelle's. She appears to have attended the Ransom Note School of Web Page Design. She doesn't like the "boring font" that Pournelle uses. That happens to be Times Roman, an elegant standard serif font known for high readability. If it has become cliched, that's because it works so well. Ms. Miller uses a sans-serif font on her own pages, apparently under the impression that sans-serif is somehow cleaner or more elegant. The problem, of course, is that using a sans-serif font for body text sacrifices readability, particularly with the long lines she uses.

What's truly pathetic is that Ms. Miller will probably get more hits on her site this week from people who read Pournelle's site and mine than she's gotten in the last year on her own merits.

All of that said, some of the posters do make valid points. Pournelle's site is never going to win any design awards. Nor is mine likely to. Both of us concentrate on providing frequently-updated content and short download times rather than on incorporating snazzy graphics and cool web tricks. But so what? It's better, in my opinion at least, to have decent content and a Spartan layout than to have zero calorie content on a Web Page McNuggets site.

Perhaps I should submit my own page to WebPagesThatSuck.com and let them waste some more of their time savaging it.

* * * * *

This from Shawn Wallbridge [swallbridge@home.com]:

My point was, when there is nothing else you can do to increase the performance, $800 is nothing. If I go out and buy a $6000 Alpha machine loaded with the biggest and best, and it is still not fast enough then $800 to make it 10% faster is well worth it.

With the next Alpha chip being Slot based I would think that KryoTech is probably a little worried about what market to target. Maybe that is why they are trying out the Gamer market.

Well, true. There are some applications--real-time commodity trading, for example--where even a marginal performance increase is worth a lot of money. But I just don't see any application where this kludge makes any sense. No one is holding a gun to someone's head to force him to use the K6-III CPU. As Anand's benchmark graphics show, even a Celeron-based system beats this product in many applications. If raw performance is what you're looking for, you don't use a K6-III to start with. If you're using an Alpha and need more performance, you simply step up to the next faster Alpha, or to an SMP machine.

If Kryotech is looking to the gamer market, they're in a lot of trouble, because the K6-III is not a particularly good gaming chip, even with the 3DNow! extensions. How long do you think it'll be before game manufacturers jump on the Intel Pentium III SIMD instructions? Not long, I suspect. 3DNow! support in game software may soon be an afterthought, if that. If Intel ships a Celeron with SSE, as I suspect they will, that'll just be another nail in the coffin of 3DNow!.

* * * * *

This from Chuck Waggoner [waggoner@gis.net]:

Just tried www.ttgnet.com, as you suggested, and I got the new updated pages without using the numeric address.

Thanks for letting me know. Given the nastygram I sent BigBiz yesterday, my old site will probably stop working pretty soon. Come to think of it, it had already stopped working for all intents and purposes.

Sunday

Sunday, March 7, 1999

We've gotten numerous messages saying that the DNS changes have propagated successfully, so if you're still accessing this page by IP address, you might want to try accessing it by name instead. Thanks to everyone who's stuck with us through this mess. The site stats are way down over the last couple of weeks, and particularly over the last few days. We're hoping that's a temporary aberration caused by the slow server response at BigBiz and the DNS mess during the changeover to pair Networks and doesn't mean that we've lost a lot of readers permanently.

* * * * *

My sources tell me that there have been some problems with getting 500 MHz Pentium III CPUs to work in some existing Slot 1 BX motherboards. It's not just a BIOS upgrade issue, either. Some Slot 1 motherboards, including recent models, cannot be upgraded to support the 500 MHz Pentium III. And you can't tell just by the model of the motherboard. For example, some Intel RC440BX motherboards support 450 and 500 MHz Pentium III CPUs, while others support only the 450 MHz version. An RC440BX with product code BERCMSTDK must be AA# 718163-208 or later; product code BERCASTDK must be AA# 723888-205 or later; product code BLKRC440BX must be AA# 724299-205 or later. And so on.

So if you're buying a Pentium III, make sure the motherboard supports it. If you're buying a new motherboard, make sure it supports the 500 MHz Pentium III, even if you aren't going to install a fast PIII initially. The faster Celerons due later this year will probably have the same requirements as the 500 MHz PIII.