photo-rbt.jpg (2942 bytes)

Email Robert

Daynotes Journal

Week of 6/7/99

Friday, April 18, 2003 07:56

A (mostly) daily journal of the trials, tribulations, and random observations of Robert Bruce Thompson, a writer of computer books.


TTG Home

Robert Home

Daynotes Home

Search TTG

Special Reports

Last Week

Next Week

Monday

Tuesday

Wednesday

Thursday

Friday

Saturday

Sunday

Top


[Monday]   [Tuesday]  [Wednesday]  [Thursday]   [Friday]  [Saturday]   [Sunday]


Monday, June 7, 1999

[Current Topics Page]

This week marks one full year of this journal. With the exception of a few days during the very early weeks, I believe I've made an entry for every day during this period. I was going to get rid of the "(mostly)" in the title, but I may in fact miss some days, so I'll leave it as is, at least for now.

* * * * *

The power failed about noon yesterday, and was off for an hour and a half. Thank goodness for APC UPSs. The big APC Smart-UPS 1100 did yeoman duty. It has three PCs, a 17" monitor, and a bunch of small stuff connected to it. All my important boxes are on APC UPSs, and all of them shut down normally except kerby, my main workstation.

When I tried to shutdown kerby, it took forever. After watching kerby try to shutdown for five minutes or so, I turned off the monitor. The hard disk indicator light kept flickering constantly, and I could hear the disk banging away. After 15 minutes or so, the hard disk indicator was flickering only sporadically, but I gave it a while longer. After five minutes or so of no disk activity, I powered up the monitor to see if it was ready to shutdown. Nope. As I was waiting for the monitor to come back up, another flurry of disk activity took place. I powered down the monitor and waited a while longer. Finally, after 45 minutes (!), I got another major flurry of disk activity followed by silence. I turned the monitor back on and it was displaying the "safe to shutdown" message.  Fortunately, kerby is attached to the Smart-UPS 1100. If it had been connected to a lesser UPS, it would have died from loss of power long before it finally completed its shutdown.

* * * * *

This from Paul Robichaux [paul@robichaux.net]. The URL contains a detailed description of how a guy built his own refrigeration system for his PC.

Subject: Isn't_that_special?

http://www.wizard.com/users/scfoster/public_html/

btw, happy birthday!

Cheers,
-Paul
--
Paul Robichaux | paul@robichaux.net |
<http://www.robichaux.net> | ICQ 25237660
Robichaux & Associates: programming, writing, teaching, consulting

Santayana defined a fanatic as someone who redoubles his efforts after losing sight of his goal, and it seems to me that this guy fits that definition. He's invested many hours of his time and spent $700 to run a dual Pentium III/500 system at 620 MHz. Big deal. A good rule of thumb is that it requires at least a 25% boost in processor speed for the difference to be subjectively noticeable. He's spent all this time and money to get a 24% boost. I don't understand the point. Surely a dual Pentium III/500 was already fast enough to do anything he wanted to do.

* * * * *

I see that Pournelle has announced on his site that I'm helping him move it, so he need no longer be nameless. FrontPage and the extensions are still driving us nuts. He started by publishing his entire site to the pair server. There were a bunch of upper-case directories and other elements that were driving us nuts because UNIX is case-sensitive and NT is not. I decided to import his web site to a local copy on my hard disk, fix the case problems and then publish it back up to the pair server. The first problem was that import doesn't work. Although the import appeared to succeed, there were numerous files on the pair server that didn't make it down to my local copy. In fact, there were entire directories missing. I fixed that by doing a file-by-file, directory-by-directory compare of what was on my local copy versus what was up on the server. After I fixed the case problems, I used FP to delete all the directories up on the server and then published my local copy up to the server. It almost works now. Almost.

I hate FrontPage. But, as Pournelle told me on the phone last night, the guy who got him started using FrontPage was named Bob Thompson. So, I have an obligation to make this all work. And I will.

 


TTG Home

Robert Home

Daynotes Home

Search TTG

Special Reports

Last Week

Next Week

Monday

Tuesday

Wednesday

Thursday

Friday

Saturday

Sunday

Top


Tuesday, June 8, 1999

[Current Topics Page]

Well, I have Office 2000. I've looked at it briefly on a system other than my main workstation, and it looks pretty good. One very nice feature is that Custom Install allows you to install O2K versions of the applications while keeping your old versions as well. The only exception is that you can't keep Outlook 97/98. It gets upgraded regardless. Which poses a slight problem. If I install O2K on my main workstation, I'm committing irrevocably to running Outlook 2000.

Fortunately, file formats aren't going to be a problem with this update. I understand that, with the exception of Access, all Office 97 file formats are used unchanged in Office 2000. That means I'll be able to run Office 2000 and still have people using Office 97 be able to read my Word files and so on without going through any export or convert hoops.

I spent ten or fifteen minutes looking at FrontPage 2000, and was very impressed. It appears to be a significant upgrade of FP98. I think I'd like to use it. The problem, obviously, is whether I should just dip my toe into this new pond or just strip down and dive in. I do have my Kamikaze headband with the Rising Sun on it (the one I always wear while doing stuff like this), so perhaps I should just strap it on and go for it.

* * * * *

This followup from Paul Robichaux [paul@robichaux.net]:

He's doing a lot of video rendering (allegedly, anyway), so a 24% increase in speed is valuable. OTOH I'd think a dual PIII-500 would be fast enough for anything, and that any further improvement is sort of like the guys who put NO2 kits on their Honda Accords.

Cheers,

Paul
--
Paul Robichaux | paul@robichaux.net |
<http://www.robichaux.net> | ICQ 25237660
Robichaux & Associates: programming, writing, teaching, consulting

Yes, but he doesn't get a 24% increase in performance. He gets a 24% increase in CPU speed. Those are very different things. Note the various benchmarks comparing a Pentium III/500 with the Pentium III/550. That's a 10% CPU speed increase, but the real-word performance increase usually shows up in the low- to mid-single digits, sometimes as little as 2% or 3%. There are so many other potential bottlenecks that going to extremes for a small percentage increase in CPU speed is a sucker bet.

* * * * *

I've been encountering server timeouts when publishing to my pair server. All of the changed files are published successfully, but during the final phase, "processing web updates for ...", the server times out after a minute or so. I'm not sure exactly what happens during this final phase, but I think it's when the text indices and other internal housekeeping information gets updated. I mailed pair Networks tech support about this late last week, and got the following response yesterday afternoon:

I can't find any circumstances on wawrra that would be causing this to happen. The number of users on the server can affect the load, but the server still has quite a low load and doesn't seem to be having any problems or slowdowns from this type of situation. I'd suspect the problem does lie somewhere else, though it's difficult to narrow it down. I've doubled the server timeout at this time, which may help you to some degree. Can you see if your FrontPage client is giving any more of an error than "timed out?" Also, just to check, can you run a traceroute to the server to see if you're possibly running into any network anomalies?

http://www.pair.com/matt/help/traceroutes.html

Regards,
Eric

I did do a tracert several times recently when this happened to me, and the results seemed completely normal. I also checked the server loads, which looked very low--something like 0.01, 0.03, 0.01--most of the times that the server timed out. Once, the loads were something like 0.13, 0.24, 0.35, but all other times they were quite low. I thought perhaps you had a new customer on the server that was doing a lot of heavy CGI processing. I'll be publishing again tomorrow morning, so I'll keep a close eye on things and mail you the tracert results if the problem recurs. Thanks for upping the CGI Timeout setting.

We'll see what happens when I try to publish today's notes...

* * * * *

The following from Chuck Waggoner [waggoner@gis.net]. As he says, it's long, but I'm publishing it in full because there are quite a few interesting points:

[feel free to edit or cut down, any way you like, if you publish this--Chuck]

The main drive on my Gateway P90 (yes, I do live in the dark ages) went flaky on Friday. Fortunately, the day before, after your rave on Maxstor, I saw an ad for an 8.4gb Maxtor drive being sold under the CompUSA brand for $99 after rebate. I couldn't resist.

That is amazing. I've been treating disk space as essentially free for quite a while now. At less than $12 per GB, that's more true every day...

What timing! That night, I got a message that my primary drive, an old Western Digital Caviar 31000, would no longer accept writes. Fortunately, I had moved the Windows swap file to the second drive, or I probably would have been completely dead-in-the-water.

To make a long story only modestly shorter, the representations on the box and in the instructions about the drive's ability to get around the 2.1gb BIOS size limitation on the Gateway P90 proved problematic. Instructions claimed that the drive would be useable on any machine without installing drivers, then promptly loaded EZBIOS--which as far as I can tell, is a driver activated before anything is booted. But Win98 would not install, giving up as unable to access a particular address, even before it asked for the CD Key number.

The Maxtor installation floppy provided an option to return the boot sector to some previous steps, I selected one that said, "before EZBIOS installation" and that worked. Mind you, EZBIOS is still there--no instructions tell how to uninstall that; but it works with access to all 8.4gb in 5 partitions, plus my other drive, so I'm not complaining.

There are a bunch of "limits" that arise with IDE and Windows. Some are caused by hardware and others by Windows itself. The best solution for all of the hardware limits (other than the absolute 137 GB IDE limit) is to install an updated BIOS that supports Extended Int13. I don't know for a fact that Gateway has such an update for your motherboard, but it wouldn't surprise me if they did. Your Pentium/90 is also old enough that it may require a BIOS patch for Y2K issues.

Having ample drive space, this time I took the advice you and Pournelle give, copied Win98 cab files (and eventually all the other software CD's) to the hard drive, and installed from the drive. That worked really fast and without a hitch. Previously, I had Win95 upgraded to Win98, but this clean Win98 install--probably combined with the faster Maxstor drive--makes the system at least twice as fast (which is still not that fast).

I continue with plans to build an NT dual, based on Tom Syroid's experiment, but have just not had the time to locate the Celerons yet.

Yes, the Maxtor drive is likely a great deal faster than what you were using, which would give you a noticeable performance boost all by itself. The clean install may also have a great deal to do with it. As far as the NT dual-CPU system, I note that some manufacturers are beginning to produce dual-Socket 370 motherboards, with built-in jumpering for dual operation. That would be an easier solution, as well as allowing you to avoid paying for slockets. As far as the Celerons, if you want 300/A ones it may already be too late. Intel ceased producing them some time ago, I believe, and they're probably already all gone. If nothing else, overclockers have probably seen to that. Not that it really matters much. You can buy 400 or 466 Celerons now, and they're likely to be as inexpensive when you're ready to build your system as Celeron 300s were recently.

Anyway, the main reason for writing is about IE5, to which I upgraded during the process. First of all, unlike others, I did not find that I could download a file to the directory of my choosing and install from there. When I clicked 'Save Target As', the only thing it saved was a shortcut to the web location. So I went with the only thing available: active download and install. Download took about 3 hours, and during install, I was offered no options whatever; it did the whole process by itself, uninterrupted by prompts. Even the readme file indicates that I would be given an opportunity during installation, to clean the drive of IE4 files, but that did not happen.

As I recall, I downloaded the iesetup.exe file, which was about half a meg. I then ran it locally. On the first screen that appeared after I accepted the license, there was a button labeled "Advanced" or something similar. Clicking that brought up a dialog with three or four checkboxes, one of which was set by default to do a direct install over the Internet rather than downloading the distribution files. If I hadn't cleared that box, I'd have been in the same position you are.

The ie5setup.exe file is customizable, so it's possible that you downloaded it from a source that disabled the option to download the distribution. It might be worth your while to download ie5setup.exe again, this time directly from Microsoft, and re-run it with the option to download-only enabled.

On to my peeve, which is this: I've used Subscriptions (Work Offline) since Day One with IE4, and became pretty experienced with it. Unfortunately, IE5 lacks a crucial element that--beyond making me wonder--leaves me positive that Microsoft programmers DON'T USE their own software creations.

In IE4, you could place Favorites links anywhere on the system that you liked, and I adopted your practice of building a catalog system of folders under 'Links'. When it came time to deal with the offline updates, by selecting the option Favorites/Manage Subscriptions, I could bring up a window listing all the links on the entire system that had been chosen for scheduled downloads, REGARDLESS of where the shortcut links had been placed.

To read the pages, I sorted that list by time of last update, and just clicked down the list to read the pages--like an index to a newspaper or magazine. Re-sort the list by name, and it was possible to modify update times, as necessary, right from that page.

But 'Manage Subscriptions' is GONE in IE5. The only way I can see to read pages is to: 1) remember where you have put all of your 'subscribed/download' links; and 2) search folder by folder for them, one by one, to open them. NOT EASY!!

And I had one link that somehow got itself scheduled to update at an odd time (I usually have them scheduled together at three or four defined times of the day, so the phone line isn't tied up much), and I had to go through the entire list of links, checking properties one-by-one, to find the errant offender.

If I am in error about that missing 'Manage Subscriptions' option--perhaps it's now called something else,--I sure would like to know. IE5 is faster and better in many other regards than IE4, but for automatic downloaders like me--this is a serious flaw! They have essentially eliminated the index page.

--Chuck Waggoner [waggoner@gis.net]

That's all news to me. I never used the subscription feature under IE4, except for a couple of halting experiments with it. I don't know of a Manage Subscriptions equivalent in IE5, but then I use IE5 synchronize only for a couple of sites, so keeping track of them is no problem. Perhaps someone else who uses this feature as heavily as you do can offer a suggestion.

* * * * *

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

Thanks for the tips.

I did try Gateway to see about a BIOS upgrade before installing the Maxtor drive. Gateway is no longer supporting the BIOS or motherboard my system has, and they say that means it will never be capable of more than 2.1gb. Actually, Gateway did refer me to another BIOS maker who sells an upgrade, but the numbers for my motherboard don't match exactly the reference numbers from my BIOS. I've read enough horror stories about dead BIOS from an incompatible flash, that I'm not tempted. They also want $100 for the upgrade, and these days, that's more than half the price of an up-to-date motherboard which already comes with current BIOS.

I see the new drive as stop-gap; I really think I need to move on up to an all-around newer system. I really can't complain about Gateway; this system has been a workhorse running 24/7 almost continuously since 1994--and it's still going!

I ran the Gateway Y2k test program, which informed me that the hardware is not Y2k compliant, but that the BIOS will correct that, so the system--at least in Gateway's view--is Y2k okay.

The download I took of IE5 was on the MS Windows update site that the Win98 installation automatically referred me to. I don't recall any options appearing at all after I accepted terms, but I did visit the regular MS IE download site, followed your instructions, and it's downloading the installation files right now.

Thanks again!

--Chuck Waggoner [waggoner@gis.net]

I'm rather surprised that Gateway does not have an upgraded BIOS for that system. Granted, 1994 is pretty old in computer terms, but Gateway is usually pretty good about posting BIOS updates. I'd have expected there to be a BIOS update much later than your original 1994 version, although perhaps a very new update would be a bit much to expect. I agree that buying a BIOS update is a sucker bet.

* * * * *

This from Dennis Loretz [loretzd@gte.net]

Situation: I would like to configure my system for dual boot between NT Workstation and Win95 OSRR2. My 8.4 gig harddrive is partitioned into two - 4 gig (FAT32) drives, with Win95 OSR2 as OS. I've backed up everything I need to onto my IDE CDRW (yes, I've read your articles and would recommend this configuration - easy setup and performance adequate for home use). The question - should I wipe the drive, reconfigure for 4 x 2gig partitions (all FAT16) and install Worstation on one partion, then Win95 on another partion? Is there trick/recommended process for setting up this configuration? Wiping the drive is not a big deal to me. I realize this will be a fairly time consuming process, but would like the option of learning Workstation and still having Win95 around for some of my kid's <g> games. Oh yeah, I have a FIC mb with an AMD K6-2 300 3D NOW processor, 128 mb SDRAM, Matrox Mystique 220 w/4mb RAM, SB 64, and generic 56k modem. Any recommendations would be much appreciated!!

Oh, one more thing. I recently (a few months ago) began reading your Daynotes. You have great website content and as a frequent reader I really enjoy your site. Thanks for all the hard work . . .

Thanks,

Dennis J. Loretz
Email: loretzd@gte.net
Web: http://home1.gte.net/loretzd/index.htm

Thanks for the kind words. As far as recommendations, here's what I would do:

1. Boot a Win95 startup floppy that has fdisk and format on it (as well as your CD-ROM drivers, which you'll need later to get to the Win95 distribution CD). Use it to blow away all existing partitioning information on the hard disk.

2. Create a 2 GB primary partition and mark it active. That'll be C: and Win95 will live there.

3. Create an extended partition that occupies the remainder of the available hard disk space.

4. Create three 2 GB logical volumes, D:, E:, and F: within the extended partition.

You have two choices for the volume where Windows NT will live. If you format it as FAT, both WinNT and Win95 will be able to see all four volumes. If you format it as NTFS, only WinNT will be able to see that volume, but NTFS is considerably more robust and secure. If you elect FAT for your NT volume, it doesn't matter which of the three volumes in the extended partition you install NT to. If you elect NTFS for the NT volume, you want to put it on the last (F:) volume, because if you put it on D: or E: the volume names will change depending on whether you have Win95 or WinNT running.

That is, if you install NT to a FAT volume, you'll have identical C:, D:, E: and F: volumes whether you've booted NT or Win95. But if you install NT on NTFS in the D: volume, for example, you'll have C:, D:, E:, and F: under NT, but only C:, D: and E: under Win95. Worse, what shows up as E: under WinNT will show up as D: under Win95, and what shows up as F: under NT will show up as E: under Win95. That way lies insanity. So, if you want to install NT to an NTFS volume, make that volume the last one available. You'll still have C:, D:, E: and F: under WinNT and only C:, D: and E: under Win95, but at least the same volume names will point to the same volume that way.

5. Restart the system and FORMAT C: /S to make it bootable. Format the D: and E: volumes as FAT as well. If you want to install NT to FAT, also format the F: volume. Otherwise, leave it unformatted.

6. Run Win95 setup and install it to the C: volume. Make absolutely sure that you install Win95 before you install WinNT. If you do it that way, WinNT will detect that Win95 is installed and automatically configure the system for dual-boot. If you install NT first and then Win95, you'll have to configure dual-boot yourself manually.

7. After you have Win95 installed and configured, shutdown the system and restart it with the WinNT boot floppies. Install NT to whatever volume you decided on, and restart the system. You'll get a boot menu that allows you to pick Win95 or WinNT at boot time.

 


TTG Home

Robert Home

Daynotes Home

Search TTG

Special Reports

Last Week

Next Week

Monday

Tuesday

Wednesday

Thursday

Friday

Saturday

Sunday

Top


Wednesday, June 9, 1999

[Current Topics Page]

Well, I got the same timeout problem yesterday morning when I published this, so the fact that pair Networks doubled the CGI Timeout period on my server seems not to have helped. It was obvious that they'd done so; it took twice as long to timeout this time, but timeout it did. Pournelle is having similar problems, but more extreme. When he mailed pair Networks tech support to ask what was going on and how it could be fixed, they responded:

[...] The situation you are seeing is actually due to FrontPage itself having problems. There are over 29 MB worth of files in this single domain's directory. FrontPage, being somewhat inefficient, loads *all* files for the web site at once when publishing. This becomes something of a memory hog and also makes you wait for a long time. Eventually this hits the web server's timeout value or gets killed by the system "reaper" script for being an extreme memory hog user.

I've doubled the server timeout value so this may help you some, but your only real long-term solutions to this problem would be:

  • break up the web site into several smaller sites. Generally FrontPage starts to choke when the site has more than 20 MB of content or so. This is not a limitation on our servers but rather one in FrontPage itself.
  • simply make your site smaller (smaller graphics, etc.)
  • use an alternate uploading method instead of FrontPage.

I have little doubt that the situation is exactly as they represent it. When I checked my site size versus Jerry's, I found that his was about 1,500 files and 20+ MB, while mine was about 1,700 files and 10 MB. We have roughly comparable numbers of files, but Jerry has many more large images than I do. Perhaps it's a matter of perspective, but I consider both of these sites to be small or small/medium in size. If FrontPage can't handle sites this size, it's a toy rather than a serious tool.

I'm inclined to see what happens with FrontPage 2000, but the fact is that these problems are all server-side, and I haven't heard anything about FP2000 extensions. Although FP98 has some aggravations on the client-side, I can live with those. I need a solution on the server-side, and I don't see that happening.

I think Jerry and I are agreed that we're not wedded to FrontPage, except to the extent that neither of us has time to re-do our entire site. We'd both drop FrontPage Extensions entirely, but we want the search function, which requires it. So what do we consider doing? Continue to use FP for page layout, but publish using WebPost? Change to one of the competing products from Symantec or others? What? If people give me some feedback on this question, I'm going to open a new heading under current topics for this (it won't be there until there's something to post).

* * * * *

This from Tom Syroid [tsyroid@home.com]:

Being as you have your headband on: <g> I think you'll find the water in the Office 2000 pond quite pleasant. Word 2000 is rock solid, and as you note, backward compatible to everything. For the first time in my life, I'm using it as a real productivity tool every day, and I would be lost without it. Outlook 2000 too is solid. Not too much has changed here except for the usual MS integration with IE5 (MUST be installed to run OL2K), some file management tools, and a few more bells and whistles. I can't tell you much more or you wouldn't have any need to buy our book this fall <g>.

I haven't used Access, so I can't comment, though I can confirm the new file format here. And while I've played with Excel, I haven't used it that much either except to build a few quick and dirty spreadsheets.

FrontPage is definitely improved, but don't let the pretty interface fool you. It's perhaps a bit more stable than '98, but it has it's share of annoying quirks. The biggest one for me is the lack of integration with the rest of the Office products. Word uses the common spell-checker shared by the rest of the Office suite; FehPage uses it's own. Word's autocorrect (a feature I use and love) is not in FP. And MS have done nothing to tighten up their implementation of HTML to be anywhere close to the emerging standard; expect the same proprietary tags and volumes of extra directories appended to your base web pages. I use FP, and there are days I like it, but MS has a long way to go with this product to bring it in line with the quality of, say, Word 2000.

That's disappointing, given that the real motivation for me to move to Office 2000 would be improvements in FrontPage. I'll probably still give it a shot, but I'm more-or-less satisfied with Word 97 and Outlook 98, so I'll probably test it on another workstation before I commit to running it on my main system.

* * * * *

This followup from Tom Syroid [tsyroid@home.com]:

PS: And the most glaring omission of all from FehPage -- the lack of any kind of AutoSave or AutoRecover (as per Word) when it does crash.

I wouldn't hesitate to recommend Word 2000 to you, though. I think you'll like the new features and stability, being the writer you are. And Outlook, like I say, is not dramatically different (unfortunately) from OL98 to have any concerns about as far as upgrading goes.

Now if they'd just clean up FP a bit... Oh well.

I'll probably go ahead and install it, but I'm always nervous doing anything like this on my main system. I spend about 99% of my time in Word, Outlook, and FrontPage, so I guess it'd be dumb not to take advantages of the improvements, incremental though they may be.

* * * * *

This from Dave Farquhar [farquhar@lcms.org]:

You might want to consider a motherboard replacement for the Gateway. Most older Gateways (except for the low-profile jobs) used a standard AT board, and socket 7 AT boards are cheap ($75 or so). Even if you recycle your CPU and memory, a modern socket 7 board will give you a speed increase over that old board, by giving you a newer, faster chipset, bigger L2 cache, and a UDMA-33 hard disk controller.

Dave Farquhar
Microcomputer Analyst, Lutheran Church-Missouri Synod
farquhar@lcms.org

Views expressed in this document are my own and, unless stated otherwise, in no way represent the opinion of my employer.

Good thought. I've used and can recommend the EPoX Socket 7 boards. They make several models in the AT form factor that use different chipsets, support different types of RAM, and so on. Another thing you might consider is buying a $40 WinChip2. That might easily double the integer performance of the Pentium/90, although the WinChip's floating point performance won't be as good as the Intel. Be aware that different motherboards have different voltage settings available, and that the WinChip is available in two different voltages, so it's important to get the right motherboard and the right WinChip.

* * * * *

This followup from Dave Farquhar [farquhar@lcms.org]:

That would work. AMD and Cyrix CPUs are pretty cheap these days too. I've seen Cyrix PR-300s and even 333s in the $40 range. Low-end AMD chips start at around $60. Cyrix is the best bang-for-the-buck integer performance, but the floating point performance is weak (not as bad as the WinChip, but still weak). The P-90 is probably better than either WinChip or Cyrix. AMD's floating point performance is better, but still not in the league of comparable Intel CPUs.

I'm pretty happy with my Cyrix PR-233 or whatever it is (it runs at 200 MHz). Performance isn't on par with the big boys, but I paid $50 for it six months ago. For what I do, available RAM is more important than CPU speed.

The last EPoX Socket 7 board I had was a 430FX-based board a few years ago. It was a workhorse. I really like Asus Socket 7 boards as well.

Good point. I don't know what I was thinking. WinChips are a good choice to upgrade an existing Socket 5 or Socket 7 system, but if you're buying a new system board it'd make sense to go with the AMD. I checked around briefly and found that the AMD K6-2/300 was selling for about $50, and it would be a significantly better processor than the WinChip for this application.

Of course, it also makes sense to consider a Celeron solution at that point. The Celeron is more processor than any of those we've mentioned, and buying a Socket 370 motherboard with a Celeron won't cost much more, if any, than buying a Socket 7 motherboard with something like an AMD K6-2. The only remaining issue is memory, because as far as I know all Socket 370 motherboards take only SDRAM DIMMs. What's currently installed on the Pentium/90 system may be too small or too slow to be worth worrying about salvaging. Or it may not. Isn't upgrading systems fun?

I've not worked much with ASUS motherboards, but I've heard good reports of them from various people.

 


TTG Home

Robert Home

Daynotes Home

Search TTG

Special Reports

Last Week

Next Week

Monday

Tuesday

Wednesday

Thursday

Friday

Saturday

Sunday

Top


Thursday, June 10, 1999

[Current Topics Page]

Lots of mail today, including some suggestions about FrontPage alternatives over in Current Topics.

* * * * *

We had a visitor last night from the Carolina Border Collie Rescue organization. Barbara contacted them and offered to do some volunteer work for them. We have two Border Collies. They make excellent pets if you understand their needs, but they are the worst imaginable pet for most people. They are working dogs with very high activity levels, and are extremely dominant. Their behavior and other characteristics are, in fact, closer to those of the wolf than are those of any other domestic breed. Probably fewer than 1% of potential dog owners are capable of handling them.

Unfortunately, they are the smartest breed of dog by far, and are extremely trainable if you know what you're doing. I say unfortunately because these characteristics make them ideal for use in television commercials. People read about how smart they are, see them on a TV commercial, and decide on that basis that that's the dog for them. What they find is that they've purchased a Tasmanian Devil puppy. They quickly decide that they can't handle the dog, and it ends up abandoned or in a shelter.

Responsible Border Collie breeders interview people who want to own a Border Collie, and refuse to sell a puppy to someone who is unlikely to be able to handle it. Unfortunately, as Border Collies have become more popular, breeders with fewer qualms have begun to offer them to anyone who wants one.

Carolina Border Collie Rescue tracks down abandoned animals and attempts to place them. Volunteers serve as foster homes until a dog can be adopted, handle email and telephone inquiries, transport dogs great distances, and so on. Barbara and Laura got on very well, and plan to start making home visits to people who have asked to adopt one of the rescue dogs.

CBCR is completely volunteer-staffed, and they also need help transporting the rescue dogs. They use a bucket brigade method. A dog in Winston-Salem may be adopted by someone in Texas or Oregon or British Columbia. A volunteer here may transport the dog to the Tennessee border, where they'll hand off the dog to another volunteer who will transport it another state or two towards its final destination. As you might expect, coming up with volunteers who have both the time and a suitable vehicle is not easy. Laura drives a Nissan Altima, and was delighted to find that Barbara drives an Isusu Trooper. Barbara has offered to help handle email and phone calls, as well as doing pre-placement visits, but I suspect she'll also end up doing some transport.

* * * * *

More from Dave Farquhar [farquhar@lcms.org] about upgrading PCs:

Lots of good points there. As for the challenges of upgrading, it could be far worse. Upgrading a PC is heavenly compared to upgrading a Macintosh.

Upgrading strategies come down to what you expect to be able to get and what you're willing to pay for it. There's little point in spending $500 to upgrade an older machine considering what $1,000 gets you in a new one. So I like to take a look at what the machine has in it, figure out what's still worth salvaging, then find a motherboard that lets me use it.

Let's just play with some numbers. I'll use Tyan motherboards since I happen to have their pricing right in front of me and I know they make plenty of AT boards. We have to go AT, rather than ATX, if the plan is to use the existing case. The Tyan S1590S board is compatible with all Socket 7 processors and all memory types. It's $94. With a K6-2/350, it's $165. (This vendor's pricing on CPUs seems a little high, but it does include a CPU fan and a 3-year warranty, so maybe it's worth it.) The Tyan S1830S Tsunami is an AT-based Slot 1 board at $122. With a Celeron/333, it's $206. The Slot 1 board will need new memory, of course.

Which is the better way to go depends on how much memory you have. The 1590S only has two SIMM sockets (as is the case with most modern boards) and three DIMM sockets. You'll probably be buying memory either way. Since I like competition and the AMD chip will do the job, that's the direction I'd be inclined to go, but the pricing is so close... And admittedly, the Slot 1 board does have better future upgradability. You won't see any Socket 7 chips beyond 500 MHz.

Decisions, decisions...

Like I said, it could be worse. It could be a Mac. Then you'd be fidgeting over whether the upgraded system will still work, whether MacOS X will be compatible with it (Apple definitely won't support it), and you'd be comparing bus speeds and cache speeds and all kinds of other messy stuff.

I'll stop at that.

Good points, all. And another thing to consider is the possible effect of mixing memory on system performance. Some motherboards allow you to mix SIMMs and DIMMs, but, depending on the SIMM type, you may not want to. For instance, I've seen one motherboard that allows you to mix FPM SIMMs and SDRAM DIMMs, but if you do so all memory performs at the FPM level. Given that SDRAM is available for around a buck a megabyte, it usually makes no sense to restrict your options with a misguided attempt to salvage old memory, which is often too small and too slow to be worth saving anyway.

I also agree that Slot 1 has much more future potential than Socket 7, but it's not as simple as it used to be. Intel will soon discontinue all Slot 1 Celerons, leaving the market bifurcated into Slot 1 for Pentium III and Socket 370 for Celeron. Of course, slockets will continue to be available, and will allow you to install a Socket 370 PPGA Celeron in a Slot 1 board, so perhaps Slot 1 does make the most sense in terms of keeping options open.

* * * * *

More from pair Networks tech support about the timeout problem:

This situation is somewhat elusive and difficult to diagnose. I can tell you that sometimes FrontPage users will run into problems with their sites using too much server memory and getting the FrontPage process "killed" off the server. This is not happening with your account. Your site is also not of a prohibitive size for FrontPage. Sites larger than 20 MB or so can have problems, but yours is only about 11 MB. In short, none of the normal causes of problems are there.

I've again increased the timeout on this server to 50% more (three times the original value), as this did seem to help you somewhat. The timeout on the server is now about as high as I can put it without causing other problems. (Who would want to wait 45 minutes to publish their site anyway?) Also, your traceroute looks just fine.

I share your concerns over the possible growing outdatedness of your site, but I'm not sure what to tell you at this point. About all I could recommend is that you possibly have the FrontPage extensions reinstalled, from www.pair.com/upgrade/, and try republishing after that is completed. I believe that the server configuration is correct, but the reinstallation would assure that to be the case. If this problem is still going on following this, I might recommend contacting MS tech support with regards to it.

Thanks for increasing the timeout. I'll let you know what happens the next time I publish. It almost seems that I crossed some sort of threshold a couple of weeks ago. If not the total size of the web, perhaps it was the number of files, the number of links, the phase of the moon, or some combination. Up until a couple of weeks ago, I was getting a timeout once or twice a month, and simply republishing immediately always solved the problem. Since late May, I think I've gotten the final "Published to..." prompt only once or twice. It's timed out every other time. I've also tried to update the indices by calling up the real web site (rather than my local copy) directly into FrontPage Explorer and doing a recalc links there. It times out just like the publish process does, presumably for the same reasons.

I hesitate to have the FP extensions re-installed, simply because I don't trust Microsoft or FrontPage not to bollix things up entirely. I'm also working closely with Jerry Pournelle, who recently relocated his site to your service and is having all sorts of similar problems. Dr. Pournelle is personally acquainted with the FrontPage product managers at Microsoft, so he may be able to get an answer and/or a workaround directly from the horse's mouth.

Incidentally, I'm considering upgrading to FrontPage 2000, and I wasn't able to find any material on your support pages about FP2000 or about the FP2000 extensions. I assume that simply upgrading my client to FP2000 won't do anything to solve the server-side problems I'm having, but I wonder if the FP2000 extensions might fix the problem. Do you plan to install FP2000 extensions any time soon? If so, you should check the Microsoft web site. There are all kinds of articles about problems with them running under UNIX/Apache, including broken installation scripts and worse. But if the FP2000 extensions solve this problem, it'd certainly be worthwhile doing whatever it took to get them installed. Surely we can't be the only users you have who are having these problems.

* * * * *

And, finally, a workaround for the bizarre behavior of IE4/IE5 when you use folders within the Links bar. For those who haven't been following this, IE behaves as follows: (1) open an IE session and load web page "A"; (b) open a second IE session (it doesn't matter how you open it) to display the default start page; (c) open a folder on the Links bar and click the link for web page "B" to display it. What happens is that IE displays web page "B" in the original instance of IE rather than in the second instance (which keeps displaying the default start page). Chuck Waggoner [waggoner@gis.net] came up with this work-around:

Here's what I just did to solve the problem with the 'visible' vs. 'non-visible' Links bug. I dragged the size of the Links toolbar frame to its smallest setting, so only the word 'Links' and the double right angle pointers show. This leaves no folders or links visible, but all are immediately available in a drop-down list by clicking the double angle pointers.

This also had the pleasant effect of widening the URL address frame. Formerly, it was a pain to see only a small part of URL's as visible.

--Chuck Waggoner [waggoner@gis.net]

Wow! It works. Finally, an answer. I was a bit confused at first, because my Links bar was a separate line below the Address bar. But I dragged it up to the same line and shortened Links down as you suggested. Now it works as it should. Thanks.

* * * * *

This from Bruce Denman [bdenman@FTC-I.NET]:

On my recent trip to Alaska I saw/used (briefly) an early Gateway Pentium upgraded with Mr. Bios and an IDT 233MHz Winchip. Can't speak to its overall speed unfortunately. Costwise, I suspect together they were far more than a standard m/b/cpu would have cost. It was a minidesktop so might have had a proprietary footprint m/b. I did not get that far inside the case. Twas tight.

Here are some thoughts/recommendations for the gentleman with the Gateway:

To Replace: Checkout/consider:

  • BIOS only -- Mr. Bios
  • CPU only -- IDT Winchip (but will still require BIOS upgrade)
  • CPU, M/B -- New Socket 7 m/b with AMD K6-2 or comparable Cyrix MII cpu. A new board with SIMM slots will minimize costs and fix BIOS/Y2k problem as well. see Note #1
  • CPU, M/B, Case Needed if original case is proprietary or if one wishes to move to ATX versions of m/b. see Note #2
  • CPU, M/B, Case, Memory -- RBT's suggestion of a Socket 370 m/b with a Celeron 333 (or better) is good; another alternative might be a SS7 m/b, AMD K6-2/333 (or better) and PC-100 SDRAM. At this point, if not earlier, it would be prudent to analyze whether or not its time to buy a complete new system. At some point; your other components may suddenly be lacking; slow/small hard drive; old CDROM, slow video card, etc. see Note #3

Note #1 I have worked with/installed and have been happy with ASUS, TYAN, Shuttle, and TMC boards The TMC AI5VG+ supersocket 7 m/b uses either two standard SIMMS or PC-66/PC-100 DIMMS. Have seen m/b and K6-2/350MHz for under $130 including shipping. I do not know of a Celeron m/b that uses SIMMS.

Note #2 The ATX form factor generally is considered the preferred design now. Most Celeron and P-II m/b are ATX. TMC's ATX SS7 m/b cost is about $10 more than the AT.

Note #3 Last week, I was quoted $79 for a Celeron 333. Pricewatch today indicates a namebrand m/b would be no more than $100, $35-50 for 32MB SDRAM, and about $50 for an ATX case. Prices do not include shipping.

Well, that looks confusing as all get out. time to leave.....

Later

Bruce in SC
bdenman@ftc-i.net
http://web.infoave.net/~bdenman
Bruce in SC, USA
bdenman@ftc-i.net

Thanks. I'll note for the readers that your comment about the IDT WinChip requiring a BIOS upgrade refers to the fact that the BIOS in the machine is not Y2K compliant. Ordinarily, you can drop a WinChip into any machine with any BIOS, and it'll just work. You may see weird things at bootup (e.g. a speed that bears no relation to the speed of the installed chip), but it will work.

* * * * *

This followup from Bruce Denman [bdenman@FTC-I.NET]:

Yes, The y2k issue was the reason for the bios upgrade. Long as you have the right version, an IDT Winchip should be drop in as I understand it.

The Gateway I saw belongs to my son's girl friend father; he had both the bios upgrade and an IDT chip installed. I will have to ask him again why he got the BIOS upgrade. Will email him and ask.

Later..

Bruce in SC
bdenman@ftc-i.net
http://web.infoave.net/~bdenman
Bruce in SC, USA
bdenman@ftc-i.net

* * * * *

This from Chris Carson [cc@bcsupernet.com]:

One small suggestion concerning:

<-------snip------------->

any suggestions about how to avoid or fix these problems. Obviously, one answer is to use all lower-case file and directory names from the start, but that's not helpful when working with an existing web that uses mixed case.

<--------------snip--------------------->

I'm sure pair.com would put a small perl script in your /cgi-bin directory to make everything lower-case if you asked them nicely.

You are starting to see why those of us who have to manage Web Servers are a little down on dear old MS.

I have trouble understanding why ftp is not an option for Jerry, but then again I really don't understand what puplishing is. I just stuff the html etc in the appropriate place ... there's more?

Good Luck

CC

--

Upgrade to Linux...the penguins are hungry!

Chris Carson aka "GreyDeth"
250-248-0142
http://carnagepro.com

I'm sure you're right that pair would have put a case-conversion script in. I probably would have asked them if they could help by doing something like that, but I got started on converting Jerry's site at 6:30 last Saturday evening. At that point, we were hoping to have everything done and the new site working by Monday morning. Although pair does have 24X7 "urgent" support, I didn't think this qualified as urgent. And, yes, I am certainly beginning to understand why people who manage web servers are annoyed with Microsoft.

As far as Jerry and ftp, the real problem isn't getting the files transferred. FrontPage does that just fine. The problem is that transferring the files is only a part of what the FrontPage publishing process does. After it completes the transfer, it does various housekeeping, including updating the text indices that it uses to support the Search function. That's the part that keeps blowing up. The only reason that Jerry wanted to migrate to a server running the FP Extensions was so that he could use Search, so this is a particularly aggravating problem.

* * * * *

This from D123FOX@aol.com:

Can you tell me if it's possible to find out if my phone is tapped, and how to do it?

No, it's not possible. If it's a legal tap, the phone company installs it at the main distribution frame in the central office, and you have no prayer of detecting it. If it's an illegal tap, any eavesdropper with the sense that God gave a chicken is going to use a high-impedance, inductive, or optical tap that will be undetectable. Those little boxes with the lights that supposedly tell you whether your line is tapped are completely bogus. The only thing they could possibly detect is a tap installed by someone foolish enough to put a low-impedance load (e.g. a normal speaker) across tip and ring of your line.

* * * * *

Afternoon: Well, I've just donned my Kamikaze headband and installed Office 2000 on my main workstation. So far, it appears to be working pretty well. This is my first edit in FrontPage 2000, so I'll post it shortly and see what happens.

I also have a Tecmar Travan NS20 20 GB tape drive sitting here on the floor, and a couple of Adaptec SCSI adapters sitting on top of kerby. I need to get a CD-R drive ordered, along with some CD-R disks so that I'll have everything I need to do the upgrade on kerby. Everything takes time, and that's what I'm short of.

 


TTG Home

Robert Home

Daynotes Home

Search TTG

Special Reports

Last Week

Next Week

Monday

Tuesday

Wednesday

Thursday

Friday

Saturday

Sunday

Top


Friday, June 11, 1999

[Current Topics Page]


Important

This warning from Werth, Timothy [timothy.werth@eds.com]:

http://www.symantec.com/avcenter/venc/data/worm.explore.zip.html

http://www.zdnet.com/zdnn/stories/news/0,4586,2273505,00.html?chkpt=hpqs014

Since you guys all get a lot of e-mail, and since Jerry and Bob got hit the last time I thought I would forward this to all of you. This time it is a worm instead of a virus and it is a lot nastier, capable of deleting files. It also uses Outlook to spread like the Melissa virus did. I know first hand that AT&T is getting hit w/this one and apparently Microsoft is as well. Have fun.

Tim Werth
(913) 491-2558 [8/559]
timothy.werth@eds.com


This is delightful. When I attempted to publish yesterday afternoon's update, FrontPage 2000 prompted me for the account name and password, which I entered correctly. It rejected that and looped back to prompt me for my account name and password over and over. I tried exiting and starting it again. No joy. Fortunately, I have FP98 installed on Barbara's system. I called up my web there, clicked Publish, and entered the account name and password. It published fine. Yep. Just verified it during this session. It won't accept the correct username and password. Watch out for FP2000.

* * * * *

I also ordered a CD-R/W drive yesterday afternoon. I dithered about what to buy. For reliability, I was inclined to buy a SCSI version, but part of my job is to suffer so that others won't have to. If SCSI CD-R(/W) drives are more reliable than IDE versions, I need to run the IDE version. So I bought a Smart & Friendly SAF798 4X2X24 SpeedWriter Plus Internal IDE drive. 

I got it from PC Connection. They're never the cheapest source, and are often significantly higher than NECx, but they ship overnight and I've never had a bad experience with them. They advertised the drive for $239. I always try to beat them down a bit, and got it for $234, which was competitive with what other good retailers sell it for. The absolute lowest price I found on shopper.com and pricewatch.com was something like $207, but I gave up dealing with bottom-feeders long ago. They invariably end up shipping late or not shipping at all, shipping you a repackaged unit, overcharging grossly for shipping, etc. Just not worth it. 

I also got a 100-pack of Smart & Friendly CD-R disks (SAF711) for $97, knocked down $2 from the advertised $99. I almost ended up buying a Fuji 100-pack for $139 that also included jewel cases. I have no use for those things. CDs I use end up in a disk wallet. Those I need to store end up in old Tyvek 5.25" diskette sleeves, safe in an old 5.25" diskette flip file. Jewel cases are an abomination.

So now I have everything I need for the upgrade I plan to do to kerby, my main workstation. I'm going to upgrade the CPU from a Pentium II/300 to a Pentium II/450 and the RAM from 128 MB to 256 MB. I'm going to install the CD-R drive, a second hard disk drive (a Maxtor DiamondMax Plus 10 GB 7200 RPM ATA drive), an Adaptec 2940U2W Ultra SCSI host adapter, and a Tecmar Travan NS20 tape drive. I've already upgraded to Office 2000. For a guy who really, really doesn't like to mess with his main workstation, I'm messing with my main workstation...

* * * * *

This from Michael L Earl [mlearl@statestreet.com]:

I don't have a solution, but an interesting point on your problem:

1. One of the most aggravating problems has been caused by the fact that Microsoft operating systems and applications are not case-sensitive. According to Microsoft, Windows NT is "not case-sensitive, but preserves case." Perhaps so, but that has not been my experience. These problems do not manifest as long as the web server is running NT, but attempting to migrate a web that uses mixed case file and directory names to a web server running UNIX is an exercise in frustration.

If one were being very picky, this is at least half the fault of the UNIX web servers; HTTP URLs, strictly by the standard, are case-insensitive. That is, a web server is supposed not to distinguish between www.foo.com/page.html and www.foo.com/Page.html, etc. (Hostnames, you may note, *are* case-insensitive, even on UNIX). The trouble is that most web servers just implement their URL store as a filesystem, and so inherit the case-sensitivity of their filesystem.

UNIX's are, of course, case-sensitive; NTFS and HPFS preserve case but are not case-sensitive (that is, two files with same name save capitalization can't be).

Windows apps are a whole 'nother issue, and seem to behave more-or-less arbitrarily with regard to preserving case, since it 'Doesn't matter.'.

One would imagine Apache could be patched to fix this (it puzzles me that it  hasn't been), or the ambitious could kludge it in other ways...

Well, yes and no. This is kind of a grey area, but if you check RFC1738, you'll find that it says that things "should" be case-insensitive. But "should" is very different from "must" in RFCs, so technically Apache fails to comply with RFC recommendations, but does comply with RFC requirements. In fact, of course, it would be very nice if Apache did ignore case, or if Microsoft in fact preserved case, both in their OSs and their applications.

* * * * *

This from Chris Carson [cc@bcsupernet.com]:

On another note I have just set up this:

http://www2.bcsupernet.com/test/

It occurs to me this is just about what Jerry needs to flog a few books.

It is a perl thingy too and should run wherever perl does ... handy stuff perl. If pair.com has perl as a module in that apache server Jerrys running on, it will be real quick too.

CC

--

Upgrade to Linux...the penguins are hungry!
Chris Carson aka "GreyDeth"
250-248-0142
http://carnagepro.com

That looks pretty slick. Unfortunately, the account type I have at pair doesn't allow me to run custom scripts, but Jerry's account type does. Incidentally, if I link to your $20,000,000 and $40,000,000 items, can I get a 10% associates fee like Amazon?

* * * * *

This from [MLewis965@aol.com]:

I admire your expertise and was wondering if you could give me any suggestions on how to connect two Win NT/98 LAN's that are about 60 miles apart together to create a WAN?

I would really appreciate your expert advice.

Mary

Unfortunately, that's an impossible question to answer because so much depends on your particular environment, the types and pricing of communications services are available in both areas, the amount of traffic that will be exchanged by the two LANs, etc. Broadly speaking, there are two likely alternatives: 

  • A private link via dedicated line or frame relay
  • A Virtual Private Network (VPN) - you connect each LAN to a local Internet Service Provider via a dedicated link and use VPN technology to tunnel secure data across the public Internet.

Deciding which will serve your needs better and what the costs will be is non-trivial. I'd suggest you locate a consultant who is experienced in this type of work, tell him what you need to do and how much you are willing to spend, and see what he can come up with. Designing and implementing a WAN, particularly one that bridges different telco service areas is a job for experts. And it won't come cheap. Expect to pay probably several thousand dollars in startup costs to design and implement the WAN. Competent WAN consultants don't come cheap. For example, if you hired me to design and implement your WAN, you'd pay me $1,500 per day, and designing and implementing the project might easily require a week or two of my time.

The first question you should really ask is whether you really need a full WAN. Simply providing both ends with decent full-time Internet access may solve most or all of your problems, particularly if you can use Microsoft's built-in PPTP VPN technology to access resources on the remote site securely.

* * * * *

This from cc [cc@bcsupernet.com]:

In response to Michael Earl on Unix case. I think you'll find Apache is case insensitive in the actual URL:

http://www.TTGNET.com/daynotes/0607RTDN.html

and

http://www.ttgnet.com/daynotes/0607RTDN.html

are the same.

Unix is however picky when we get to the file system.

How else could we do a;

/home/cc/HOME/Home/home/hOme/homE/.jpg/.tar.gzs/AwaY/

Ah yes ... maybe we are crazy, and I guess a problem for systems less complete.

I put up a Search Engine I boosted somewhere (full credits at bottom of page) at:

http://www.carnagepro.com/search.html

This searches the HOWTOs in my pub/Docs/HOWTOs/ dir (the caps are traditional). It will run anywhere there is perl which is any well equipped server. It is pretty tiny a .zip a little over 5k is at:

http://www.carnagepro.com/pub/search.zip

This is highly unoptimized, I do not have mod perl on this server (I'm only using perl for persistent stuff so my system perl is fine) so it's a bit slow.

<LowKeyNixRant>

The point is as Larry Wall, creator of perl, is fond of saying "there is more than one way to do it". The more choices you have the better off you are. I could build a good (not just nicked off the net) search engine in a couple of days in maybe 10 different ways. Anything from a compiled bit of C (so much lying around, hell just cut and paste up 90% of it), to a hopped up bash shell script, 

... well that's stretching a bit but you get the picture.

Once you get past the user-hostile *nixness of it all there is a great wealth of solutions.

</LowKeyNixRant>

A search engine needs to parse a bunch of text to find a string. FPs index deal is probably a simple data-base of pointers, a good way if you can index anytime, not wise over something as flaky as the net and useless in a dynamic environment. It would probably be good for intranet use in a static environment.

Jerry Pournelle is sitting beside (netopologicly speaking) a Cobalt Cube running it's version of Apache, yup a little late night portscanning. The NT box he's on runs II 4.0 and really should do Front Page Extensions no problem but the Cube could be made to do a great wealth of stuff ...

a terrible waste ... hi ho.

This is completely out of hand I'm babbling wildly ... thanks for your indulgence.

CC

--

Upgrade to Linux...the penguins are hungry!

Chris Carson aka "GreyDeth"
250-248-0142
http://carnagepro.com

Thanks.

* * * * *

This from [support@pair.com]:

It does seem like something broke or you crossed some sort of line from your description. Normally these things are pretty apparent and there's an entry in our logs of memory usage having been too high. The symptoms are normally the same as what you're seeing, too, which is why it's strange that the cause doesn't seem to be the same.

While it's understandable that reinstallation of the extensions could have adverse effects, this does normally seem to help things from what we've seen. The extensions supplied to us by MS aren't the most robust software, admittedly, but this is usually a step in the right direction. Usually.

There is not yet any FrontPage 2000 package of stable server extensions available yet at this time that we're aware of. We do not install beta or immature software so as to assure the integrity of our servers, and so we will wait for a stable release version before we will install any new extensions. In the meantime, FP 98 extensions will work with a FP 2000 client. While these will work, any new FP 2000 features will not yet be supported.

Regards,
Eric

* * * * *

This from David Silvis [HUPPNUT@aol.com]. We've been best friends since we were six years old. His email address refers to the fact that he collects Hupmobiles:

"This year will go down in history. For the first time a civilized nation has full gun registration. Our streets will be safer, our police more efficient, and the world will follow our lead into the future."

You forgot the source of that quote--Adolph Hitler.

Did not. I knew that you would know it.

 


TTG Home

Robert Home

Daynotes Home

Search TTG

Special Reports

Last Week

Next Week

Monday

Tuesday

Wednesday

Thursday

Friday

Saturday

Sunday

Top


Saturday, June 12, 1999

[Current Topics Page]

Well, the Smart & Friendly CD-R drive showed up yesterday morning. I donned my Kamikaze headband, stripped a bunch of excess equipment out of the test-bed system, and installed the drive. After trading my headband for my pirate's black eye patch and stuffed shoulder parrot, I put a CD in the CD-ROM drive, a blank CD-R disk in the S&F drive, fired up Easy CD Creator, accepted all the defaults and allowed it to "test and burn". After 13 minutes of test pass and another 13 of burn pass, EZ-CD told me the copy had been successful and ejected the disk. ARRRR!

Well, it *looked* like it worked. EZ-CD said the disk had been copied successfully. The trouble was, the disk wouldn't read in any CD drive, including the CD-R drive itself. Fortunately, I'd also ordered a 100 pack of bulk S&F CD-R blanks, so I tried some more. It didn't work at 4X, so I tried it at 2X. No joy. I tried it with the intermediate copy-image-to-hard-disk step. No joy. 

I called up S&F tech support. After 5 minutes or so on hold, I spoke to a technician who seemed to be in no hurry at all. He walked me through a test to make sure that the drive itself was working properly, which it was. So, at this point, all I can say is that I can't seem to burn a copy of a source disk that is itself a CD-R. ARRRR, indeed.

I talked to Paul Robichaux in the middle of all this. He asked if I had my Kamikaze headband on, and I told him that I had had it on while doing the install. He thought I should take a picture of myself wearing it and post it here. I told him that I'd thought about doing that, but decided not to for two reasons. First, the headband has a Rising Sun centered with (I presume) Japanese characters on either side of it. I don't read Japanese, and have no idea what they say. Probably something like, "the guy wearing this headband is an idiot."

But the real reason I don't is a bit more serious. As I told Paul, I always feel funny even mentioning the headband. The men who wore the real ones were incredibly brave. Trivializing their sacrifices, or those of the men who died on the ships they attacked, makes me feel uncomfortable.

* * * * *

I've decided to start playing around with sub-webs. If you highlight a folder in FP2000 and right-click, one of the options is Convert to Web. That option is not present in FP98. Someone told me that the only thing one has to do to create a sub-web is to put a file named index.html into a sub-folder. I just created a subfolder named \rbt\subweb and created index.html in that subfolder. When I publish this (with Barbara's FP98; FP2000 still won't let me publish) I'll mark the option that tells FP not to publish subwebs and see if the folder and page show up on my server at pair. This link should be dead if things work properly.

* * * * *

This from Paul Robichaux [paul@robichaux.net]:

"This year will go down in history. For the first time a civilized nation has full gun registration. Our streets will be safer, our police more efficient, and the world will follow our lead into the future."
You forgot the source of that quote--Adolph Hitler.

While listening to Rush Limbaugh[1] one day, I heard him admit to propagating this meme. Some web site (he didn't say which) that debunks quotes researched the claim and found that Hitler didn't say it. The big tipoff was that the full version, with citation, usually cites the date of his speech as _before_ the Nazis took power. There's also no similar quote in any of the transcripts of his speeches. 

Cheers,
-Paul

[1] Hey, when there's nothing good on the FM side of the dial it's AM radio time. What can I say.
--
Paul Robichaux | paul@robichaux.net | <http://www.robichaux.net> | ICQ 25237660
Robichaux & Associates: programming, writing, teaching, consulting 

You may be be right that Hitler did not say this. I have not myself read it in any of his writings, but the only thing I have ever read by him was Mein Kampf and small extracts of speeches contained in history books and so on.

But I'm afraid that, although Rush Limbaugh may have popularized it (I've never listened to him, so I don't know), he is not responsible for it, at least not since he has become famous as a right-wing radio announcer. I first read this quote at least thirty years ago, and had no reason at the time to doubt its attribution. I don't recall the original source. It may or may not have been an NRA publication.

I still have no reason to doubt its attribution. The sentiment certainly fits, as do the facts. Hitler spoke often and at length, and frequently spoke extemporaneously, so it is beyond my abilities to verify that this quote was not made by him. Since the the web site that "debunks" this quote is not named, I have no way to verify anything. Those who run that site may have an anti-gun agenda of their own, or they may simply be incompetent. Or they may be right. But I see no reason to believe they are right without further information.

* * * * *

This from Steve R. Hastings [steveha@mdlcorp.com]:

I recently had to move a web site from NT to UNIX. The problem isn't that UNIX is bad; the problem is that Front Page is bad, because Front Page will write links that have a different case than the filenames. Front Page gets away with this on NT, but not on UNIX. 

Anyway, an attached .ZIP file has my solution: a pair of UNIX shell scripts. They rename all the files so they are in all lower-case letters, and edit all links in all web pages so they are in all lower-case letters.

To use them, unpack the files and put them in your home directory, in a subdirectory called "flat". (This is because the scripts hard-code the path "$HOME/flat/whatever"; you can put the files anywhere if you change the path coded in the scripts.)

The "flatlinks.sh" script repeatedly runs the "vim" editor with a script, and it goes through each .htm or .html file looking for links, and forcing the filename part of the link to lower-case. (You could change the scripts to work with "vi" if you don't have "vim". vi doesn't have the "\+" operator; you can replace "x\+" with "xx*", where x is any regular expression that matches a character. Email me for details if you have any questions.) Also, if you need to flatten links in any files not ending in .htm*, you can edit the flatlinks.sh script to look for those files.

The "flat.sh" script repeatedly runs an "awk" script I wrote, which forces a file to lower-case if it has any letters that are not lower-case.

Both scripts use the "find" command to walk through a whole directory and all subdirectories. They were written, and tested, on a Red Hat Linux 5.2 system, but they should work almost anywhere.

P.S. I wrote these scripts at work, on company time. Thank you to MDL Corporation and my boss for permitting me to release these scripts to the world. Naturally neither MDL nor me can be held liable if you have any trouble with these; you're on your own, so be careful. Have multiple backup copies of your site before you play around with these scripts!

P.P.S. Jerry, I literally wrote and tested those scripts in under half an hour, and they processed a rather large web site with no errors. This is an example of why people love UNIX and Linux so much: once you learn your way around, you can do really cool things, really easily.

--

Steve R. Hastings "Vita est"
steve@hastings.org
http://www.blarg.net/~steveha

Thanks. I knew it was the fault of FrontPage rather than UNIX. I think I got Pournelle's site pretty well converted using the brute force method--renaming folders within FrontPage from mixed-case to all lower-case. That process should also fix the links accordingly. There are a few files remaining that use mixed case names, but not many. The major problem seemed to be with directory names rather than with filenames, anyway. I've normally use all lower-case (a hold-over from my programming days), so I don't have that problem here. I do have some upper-case filenames (e.g. each of my weekly journal files is "<date>RTDN.html) but that didn't cause any problems with my UNIX servers.

As far as the power and flexibility of UNIX, I don't think that Jerry or I disagrees with that at all. We *know* that UNIX is much more powerful and more flexible. We just don't have time to learn it. More's the pity.

* * * * *

This followup from Steve R. Hastings [steveha@mdlcorp.com]:

I do have some upper-case filenames (e.g. each of my weekly journal files is "<date>RTDN.html) but that didn't cause any problems with my UNIX servers.

There won't be a problem unless Front Page made a link to "<date>rtdn.html", or "<date>RtDn.HTML", or some other case than "<date>RDTH.html". You can have any case you want in filenames, under both UNIX and NT. The difference is that under NT, the links don't have to match the case exactly, but under UNIX they *do* have to match exactly. 

Apologies if you already understood that. (I'm not assuming that you did, or did not, understand that already.)

As far as the power and flexibility of UNIX, I don't think that Jerry or I disagrees with that at all. We *know* that UNIX is much more powerful and more flexible. We just don't have time to learn it. More's the pity.

Please forgive my little bit of advocacy. :-) I didn't mean to imply that either you or Jerry think bad thoughts about UNIX; I just wanted to emphasize that the credit for me being able to do powerful things in half an hour goes to UNIX, not to me.

--

Steve R. Hastings "Vita est"
steve@hastings.org
http://www.blarg.net/~steveha 

I did understand it, but it certainly bears repeating. Actually, I *never* use anything except lower-case. My original 9999RTDN.html files were actually named 9999rtdn.html. One day, FrontPage all by itself decided to rename every file in that directory to upper-case, so I ended up with all files named "RTDN". For some reason, FrontPage left the extension lower-case. In the process of gratuitously renaming my files, it also updated all the links to point to the upper-case names, so I decided to just leave them as they were.

I've probably mentioned this before, but I hate FrontPage. Every month or two, it arbitrarily decides to go out and edit/save every htm* file in my web site. I could live with that (although it means I have to publish hundreds of files each time it happens) except that it also updates the date/time bot in those pages. I've noted something about that in bold red at the bottom of my Daynotes home page.

No need to apologize for saying nice things about UNIX. I like UNIX. I just can't afford the time to become competent with it. If I could use Linux to do everyday things as easily as I use Windows NT, I'd switch to Linux in a heartbeat. I think I may get a copy of Caldera OpenLinux and build a computer to run it on.

* * * * *

This from Steve Worley:

I read your journal each day.. it's kind of fun to see the hassles you go through and your final solutions. Always informative!

I remember your annoyances with Pair Networks [especially billing] when you first moved your web page. However, it appears that Jerry has also now moved to Pair, probably at your recommendation. Does this mean that your opinion has changed? I'm very close to moving to Pair myself, and your (now mature) experience with Pair will probably be my final deciding factor. Their rates seem quite reasonable compared to my current ISP!

Thanks for any thoughts!

-Steve Worley

Everything is a trade-off, and picking a web hosting company is like picking a spouse. Jerry indeed is moving to pair Networks because I'm there, but I'll tell you the same thing I told Jerry: pair is not perfect. They have absolutely superb connectivity--better than some countries--and they know how to run web servers. But don't expect a lot of hand-holding. Not only does pair Networks not have toll-free tech support (as many web hosting companies do), they don't have telephone support of any type. If you have a problem, you send email to them and get an answer in a day or two (or three). If you have an urgent problem, you send email to urgent@pair.com and probably get a response within an hour or two. pair is UNIX-based and doesn't provide training wheels. Many other hosting companies, for example, provide web-based tools that allow you to configure mail autoforwards and so on. With pair, you use telnet and UNIX-based command line tools to configure your autoforwards. 

Overall, I'm pretty happy with pair, although I still hate their billing department. But what I want in a hosting company may not be what you want. If you want hand-holding, they're a terrible choice. If you want a professionally-run web hosting company that expects you to do a lot of things for yourself that other web-hosting companies do for you, with the additional flexibility that that implies, pair is a great choice. For example, my last hosting company (BigBiz.com) provided usage reports generated by Analog. I didn't have any choice about how the reports were formatted or what information appeared on them. pair Networks dumps the raw data files into a subdirectory of your home directory and lets you do what you want with them. My friend Steve Tucker has set up a cron job that processes the stat files with Analog and mails reports to him. I prefer to massage my data locally, so I simply download the raw stats files and run Analog locally. That's a choice that pair gives you that isn't even available with many hosting services. But the flip side is that many users will never see any of their stats because pair doesn't have a canned setup to generate them.

 


TTG Home

Robert Home

Daynotes Home

Search TTG

Special Reports

Last Week

Next Week

Monday

Tuesday

Wednesday

Thursday

Friday

Saturday

Sunday

Top


Sunday, June 13, 1999

[Current Topics Page]

Here's an amazing true fact. Did you know that when you copy a CD to a CD-R, the result is *never* an exact copy of the source? I had been under the impression that the process of burning a CD-R was a straight digital copy operation like using diskcopy to copy a diskette. Not so. In fact, it more resembles an analog copy like photocopying a photocopy or duping a VCR tape, with all the generational degradation that implies. Each generation introduces additional errors. 

These errors occur in the low level data structures that actually store the data on the CD. The visible data--a file or music track--are actually derived from the underlying low-level data. Errors in the visible data are not immediately evident, because the drive electronics actually construct that visible data on the fly from the underlying low-level data. Error correction allows succeeding generations to continue to deliver perfect digital data, up to a point. But when the error correction stuff is swamped by generational copy errors, the disk itself or some files on it may be unreadable.

My conclusion is that a CD-R source disk may be marginal but within limits for direct reading. But using that CD-R as a source disk with the inevitable errors that arise during the copy operation may render the target CD-R unusable. The only real solution is to use a bit-by-bit compare or CRC compare utility to compare the source against the destination. Rule of thumb: copy only from original CDs.

Obviously, all this has serious implications for people who depend on CD-R for backing up or archiving data.

* * * * *

Laura Kangas and Barbara just left to pick up three rescue puppies in Greensboro and deliver them to Raleigh. I was planning to go with them, but I just have too much work to do. I assumed that the three puppies were Border Collies, and was surprised to find instead that two are Beagles and one is an Australian Shepherd.

 

 

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