Week of 11/9/98

We're still cruising on the new project. It was supposed to be a 50 to 100 page technical paper in 12 sections. We've completed the first drafts on the first three sections so far, and they're 16 pages, 68 pages, and 28 pages. We've done a quarter of the sections, and we're already 12 pages past the maximum size we agreed to write. No wonder this isn't going as fast as we'd hoped.

In addition to everything else going on, we're preparing for Barbara to go into business for herself. She's been a librarian for more than 20 years now, and has decided to resign effective 12/31/98 and begin working as an independent research consultant. There are about a million details to deal with--everything from buying individual health insurance coverage for us now that we'll no longer be covered by her policy at work to getting telephone numbers arranged to getting business cards printed. The next few months will be an exciting time.

Tuesday, November 10, 1998

Got another section done yesterday on the new project. Things are starting to move along a bit better now. We're almost half way through the first draft. We should finish the first draft in another week or so. After that, there'll be another week or so of incorporating feedback, So, that takes us pretty much through Thanksgiving. I'd better get back to work on it now. I hope to knock out another section today, although that's probably optimistic.

Wednesday, November 11, 1998

The march continues. I got most of another section done yesterday on the new project, and have my co-author's part of yet another section waiting in my inbox. One thing I didn't think about when I took on this project was the effect it would have on my web page. There's a double whammy here: not only do I not have time to write much, but I don't have time to work on the stuff that's the raw material for what I write about. I got to thinking the other day that I'm likely to lose quite a few of my regular readers. I mean, I wouldn't continue to visit a place that didn't have much new content, so I can hardly expect my readers to do so. For those of you who continue to visit my page, thanks. For those who are getting bored, I promise that there'll be lots of new hardware-oriented stuff coming up once I get through this project. Thanks for bearing with me.

One bright spot. My newest book, Windows NT TCP/IP Network Administration, is apparently selling quite well. It's number seven on the Computer Literacy bestsellers list for new arrivals, and seems to be doing well elsewhere too. It's been sitting in the 400 range on Amazon.com for a couple of days, and during the last couple weeks has been as high as 159 and has probably averaged about 1,000. That's good for any book, and particularly good for one that has only been available since mid-October. If you want to buy a copy, you can get it directly from Amazon.com by clicking here. That helps me twice. Once because another copy of my book sells, and again because I get a small kickback from Amazon from direct purchases from this link. I don't get rich on the kickbacks--they totaled $16.72 last quarter--but they do help to pay the costs of keeping this site up.

Thursday, November 12, 1998

More of the same yesterday. Sometimes I feel like we're not making any progress at all. We work very hard and yet things keep changing and the end of the project seems to be further away than ever. Something has to give. I'm now 16 days into what was supposed to be a 21 day project, and the end is not in sight. I have previous commitments to consider. Going into this project, I made it clear that three weeks was going to be difficult to fit into my schedule. The way things look now, this thing could run two months or more, and there's no way that I can do that.

One very bright spot yesterday. David Rogelberg, my agent, called to say that we have an offer on the book that Jerry Pournelle and I proposed. I told David that the terms the publisher offered were acceptable to me, and that I hoped they were acceptable to Jerry. I was a little bit concerned, because the economics of computer books and fiction are very different. In fiction, the author typically gets a higher royalty percentage, and that royalty is based on the list price of the book. In computer books, the royalty percentage is lower, and it's calculated on the net. That is, a $30 computer book actually costs the distributor something like $15, and it's that $15 that royalties are calculated on. So, at first glance, it looks to a fiction author as though royalties are about one third what they are on fiction. And so they are, but that disregards one very important factor.

A typical fiction hardback has the shelf life of a bag of potato chips. There are exceptions for the big-name authors, of course, but basically a new fiction hardback hits the shelves, is actively sold for a month or so, and then gives way to the next batch of new books. It may continue to be sold less actively for three months, or even six, but sooner rather than later it ends up remaindered, which means the author gets no royalties for it. Computer books, on the other hand, typically remain as active titles for at least a year and sometimes two. They are also often revised and sold as a new edition. In some cases, they can go on selling through several editions over several years. In short, a fiction book earns a lot per copy for a very short time, while a computer book earns a little per copy for a much longer time.

At any rate, as I expected, Jerry was underwhelmed by the terms, but he really wants to do a computer book, so it looks like we'll be getting started on it before long.

Friday, November 13, 1998

Well, the special project has collapsed, and it's my fault. More properly, I'm the one that kicked the struts out from under the whole teetering structure. I went into this project with the understanding that I was to co-author a technical white paper that was to run from 50 to 100 pages, and that the project was to run three weeks start to finish. I made it clear to everyone going in that committing three weeks to a project was very difficult, given my current commitments, but that I would commit to three weeks and no more. After getting 16 days and about 150 pages into a 21 day, 100 page project, the client was still making major changes and additions to the outline at the A-head level.

There was no way this project was going to be anywhere near done at the 21-day mark--it was beginning to look more like a two month project--and I figured I'd do better to raise the issue now than wait until we got to day 21. I was hoping that would allow them to bring in a replacement author to finish the project. They decided instead to cancel the project, and told us we wouldn't be paid for the work that we'd already done. Not a wise move on their part, but given the way they've handled this project we had no reason to expect them to behave wisely now. The whitepaper was a competitive analysis, and I suppose we could turn around and sell it (with suitable modifications) to the competitor.

So, I've been working like a madman for the last 15 days for absolutely no return. Even at middling rates for technical writing, the hours I've put in are worth something like $10,000, which I'll end up eating, as will my co-author and agent. I actually kind of hope the client uses the material we did for them without paying us for it. It'd be kind of fun to watch our agent nail them to the wall.

The good news is, this means I can get back to work immediately on the book I was working on, which I really enjoy doing, and on the book I'm going to do with Jerry Pournelle, assuming we can come to an agreement on terms with the publisher. More good news is that I'll now have time to help my wife get her business started. Still more good news is that I'll have more time to work on this web site.

Afternoon: Well, it's 3:30 on Friday afternoon, which means it's time to backup the network again. I'm very nervous about depending on Windows NT 5.0 to do it. I like the new backup applet, but I just don't trust a beta far enough to use it for backing up. Since the tape drive is in that box, which currently dual boots NT5 as bastet and Win98 as anubis, and since Win98 won't recognize the tape drive, I guess that means it's time to blow away NT5 Beta 2 and reinstall NT4, my backup software, and so forth. At least that'll give me a chance to install Service Pack 4 to see what it adds, if anything.

That's interesting. I stuck the Windows NT CD in the drive and it autoran. It spent several minutes copying all the distribution files up to the hard disk and then forced a reboot. When the OS Loader screen came up, the default choice was Windows NT 4.0 Setup, as expected. There was a 5 second timeout on it, but every time it timed out, it just looped back to the OS loader screen. Perhaps because NT 5 is installed. At any rate, I stuck in the boot floppies and restarted. That should work.

Yep. The installation seems to be proceeding normally. I got to the point where Setup asks which partition to install NT on. The 1 GB C: volume has Windows 98 on it, and the 2 GB D: volume has NT5. The D: volume showed up as NTFS (rather than NTFS5). I told Setup to blow away the 2 GB NTFS partition, which it did successfully. It's formatting that space as NTFS right now. Okay, the format completed normally, and the normal forced reboot just occurred. Now I'll have to wait for several minutes while it copies all the files.

As I'm sitting here waiting for the files to copy, I'm looking at the InfoWorld that just showed up. It's dated Monday, 11/9, but just showed up today. That's been going on for several years now, and I'm not sure why. They used to actually arrive on the Monday cover date, but nowadays, Wednesday is about the earliest I can hope for. At any rate, there's a front page article about Windows problems. Reading the continuation on Page 23, I notice that Microsoft is building in a new mechanism that I don't much care for. Apparently, the Office 2000 beta will run only 50 times before it blows up. The only way to unlock it at that point is to contact Microsoft and register. According to Microsoft, this is a way to help customers register so that they can take advantage of updates and other Office resources. Yeah, right. I won't be using Office 2000, that's for sure. Perhaps it's time to request an eval copy of Corel WordPerfect for Linux and think about changing platforms.

Okay, we're in the gathering information phase of installing the network. I've provided my name, company name, and init key. I've named the computer bastet. Now it wants to know if I want this to be a PDC, BDC, or Stand-Alone Server. I think I'll opt for the last.

Windows NT 4.0 installed fine, and I now have a brand new server. The next step is to install Service Pack 4, which I have on CD. Sticking the CD in the drive autoruns an HTML file in IE. After reading through the release notes, I started the install. It went fine, except that it notified when when it was complete that I needed to install some Y2K components, specifically for IE 4.01 and some data access components. Okay, that should be easy enough. They're on the CD, with links to click to install them. I try doing that, but no dice. The setup program tells me that it was unable to update all files and that I should close all applications and try again. The only application I have open is the one displaying the SP4 HTML page, but what the hell. I closed that, opened Windows NT Explorer and browsed the CD. I found Y2KSETUP in the \i386\updates folder and ran it. Even after closing Explorer and letting Y2KSETUP run on its own, I still get the same message. I have no applications open, so I don't know what it's talking about.

At any rate, it tells me that SP4 is installed successfully. After the usual reboot, it's time to install Seagate BackupExec. Oh-oh. I better check Control Panel Tape Devices first. Yup. The Connor TR4 drive was detected, but NT never loaded the driver for it. I loaded the driver from the original CD, and applied SP4 one more time. Now to install Seagate BackupExec.

I have a version 6.0 eval copy that they sent me a couple of years ago, back when BackupExec was still Arcada. I'd downloaded an update to Seagate BackupExec 6.11 Build 184 (or was it 204? They have both) and it worked fine. So, I decided what the heck. I'll see if they have an even later update. They didn't, for V6.x, but they did have 7.2 updates. I don't know if my eval serial number will work with 7.2. It worked going from Arcada 6.0 to Seagate 6.1, but who knows. I figured it was worthwhile downloading the update to try it, so I started the download. The trouble is, the file sizes are listed without any commas. The 7.2 update shows as 45338556, which I parsed as being 4.5 MB. A little big, I thought, but worth the trouble to download. When I started the download, I noticed that the bar wasn't moving very fast. I finally went back and looked at the file size carefully. It turns out to be 45 MB rather than 4.5 MB. That seems excessive for a utility program.

Well, I'll let it run. I'll be a little bit disappointed if my existing serial number doesn't work, though. This is an live eval, not-for-resale serial number, so there's a better chance it'll work than there is that a regular serial number would. We'll see. I'd really like to use 7.2. I should probably call Seagate and ask for an eval if this doesn't work.

Saturday, November 14, 1998

I let the download complete, and ended up with a 45 MB exe file. By this time, it was about 9:00 p.m., but I decided to go ahead and see what I could accomplish. When I ran the 45 MB file, it uncompressed to yield individual installation sets for the entire BackupExec 7.2 product and all its options.

I installed the main backup product only. When got to the point where it prompted me to enter my serial number, I was a bit concerned. In the past, the existing serial number had always been filled in for me. I re-entered my existing serial number, which fit the matrix on screen. When I clicked OK, Setup told me that the serial number had already been registered. I wasn't sure what that meant, but I decided to hope for the best and proceed. Worst case, it was going to install a 60-day eval copy.

Sure enough, what I ended up with was a 60 day eval. I'm not crazy about using a time-limited eval to do a real network backup with, but the only other choice would be to uninstall everything, reinstall BackupExec 6.0 and reapply all the patches to 6.11. So, I decided to go ahead and use the 7.2 eval, figuring I could sweet-talk Seagate out of a live eval later. I configured it to back up all my network volumes and the local drives, got it running and went to bed.

This morning I came in to find it had completed the backup and verify successfully. It was very slow, however. With 6.11, I was used to getting throughput of about 25 MB/min on network volumes and about 38 MB/min on local ones. This version gave me throughput overall of about 21 MB/min. That may be my fault, because I didn't spend much time investigating options. Compression was on, though.

The more I think about this, the more I think I may just revert to the 6.11 version. It does everything I need to do. It's faster than this latest version. I have a permanent license for it. BackupExec 6.11 is more than Good Enough. That's what I'm going to do--remove 7.2 and revert to 6.11.

Okay, I've done that and am getting ready to do my second network backup in the last 12 hours. I better get this published.

Afternoon: It's about 2:00 p.m., and BackupExec 6.11 just finished the verify pass. The backup pass went as expected. Backing up data from a network volume yielded about 21 MB/min throughput, compared to about 38 MB/min from a local volume. I'm still trying to figure out what the problem is. Obviously, it must be either BackupExec, the network or Windows NT itself.

I don't think it's BackupExec or the network. If BackupExec were incredibly inefficient at backing up network volumes, I think I'd have heard the howls of outrage by now. As far as the network, it's a 10BaseT Ethernet, which should yield close on 1 MB/sec actual throughput. My hub has status lights to provide a real-time indication of how heavily loaded the Ethernet is. There are lights for 1%, 2%, 3%, 6%, 12%, 25%, 50%, and 80%. During a backup, these "pulse" from 1% up to 25% and sometimes 50%. If BackupExec were sucking down data across the network at capacity, all of the lights should remain on pretty much constantly. Instead, they pulse from 1% to 50% at about once per second, spending about half their time at each loading level. That means I'm effectively getting about 25% utilization of the Ethernet overall, which pretty much explains why I'm only getting 22 MB/min throughput.

There's no reason why the network should be behaving this way. Come to think of it, the throughput was identical back when I was running 10Base2, so the hub and UTP can't be the problem. This has to be a problem with Windows NT not being able to deliver data fast enough. That's pretty pathetic--not being able to saturate a 10 Mb/sec Ethernet. Perhaps I'll test this assumption by copying large files back and forth over the network, first from NT to NT and then from Linux to Linux. That should prove interesting. If the NT-to-NT and the Linux-to-Linux copies are about equal speedwise, I'll know it's BackupExec causing the problem. Somehow, I don't think that'll turn out to be the case.

That brings up another issue with BackupExec. When I installed it, it asked for an account that it would assign the Log on as a service right. BackupExec Setup installs three services, two of which are set to start automatically. Several months ago, I did some trial and error troubleshooting to figure out why the server that BackupExec was running on took forever to shutdown. I mean, I'm talking about 15 minutes or more between the time you issued the shutdown command and the time it finally actually shut down.

I'd run into this problem before with Exchange Server, where it turned out that at least one of the services installed with ES was causing the problem, so I figured it was probably due to a poorly designed service in this case too. BackupExec services were the only likely suspects, so I disabled all of them and did a shutdown. It took about 1 minute total. I then restarted the services and did another shutdown. Sure enough, it was back to taking 10 or 15 minutes to shutdown. So, I reconfigured the BackupExec services for manual startup and made it a matter of routine to start the BackupExec services only when I was ready to start the backup and to stop them immediately after the backup was complete. I just now went into Control Panel Services to do the same thing on this machine.

And that brings up still another interesting issue. I'd always noticed that the verify pass took a much shorter time than the backup pass. Today, I happened to look over at the hub while BackupExec was doing the verify pass. There was no activity at all, even though BackupExec was in the process of verifying a backup of a network volume. That seemed odd, so I continued to watch the hub for several minutes. BackupExec verified more than 300 MB during that period, and the indicator lights on the hub remained dark during that period.

Okay, I thought, there are two possibilities here. It's certain that BackupExec is not doing an actual compare between the data on the tape and the data on the drive. Perhaps that's why they call it a "verify" instead of a "compare." Possibility number one is that it's literally just doing a verify--reading the entire tape to make sure that each bit is readable. I hope that's not all it's doing. Possibility number two is that BackupExec calculates a CRC value for each file as it backs it up, and stores that CRC with the file. By then reading the file from tape, recalculating the CRC from the tape data, and then comparing the two CRCs, it could in fact determine whether the data stored on tape was the same as that read from disk. Let's hope this is what they're doing.

At any rate, the backup and verify is complete. Per my usual, I'll go stick this latest tape in Barbara's purse and retrieve last week's version. With Barbara quitting her day job as of 12/31/98 and starting to work for herself at home, this field expedient method of rotating backups off-site isn't going to work much longer. I'll have to come up with another solution.

And now, on a completely different topic, I'm considering overhauling my web site appearance. FrontPage lets you use "themes" which are a group of standard backgrounds, lines, buttons, etc. that give a site a consistent appearance. When I first started using FP, Barbara and I looked at the available themes. Of the two or three dozen themes that came with FP, we liked only two. This one, called "Expedition" and one other. The only problem with using Themes and FrontPage components is that they kind of lock you in--the flip side, I suppose, of ensuring consistency.

At any rate, Jerry Pournelle has been going on about wanting to change the appearance of his web site to make it look more "professional." I cobbled together a quick-and-dirty home page for him just as an example. The more I look at that home page, the more I like it. I think I may start using that style for this site instead of the current Expedition theme. Among other advantages, it's very light on graphics, so it should load pretty quickly.

I'd appreciate your thoughts about whether I should shift to this new style or keep the existing one. You can see the example I dummied up by clicking here. [Yes, I know this link doesn't work. I decided just to redo the whole site instead of showing an example. RBT] Note that the links don't work and so forth. This is just an example. Please let me know what you think.

Late Afternoon: Well, as usual, once I decide I want to do something, I go ahead and do it. I decided I liked this simple, text-oriented format so much better than the FrontPage theme that I went ahead and started replacing older pages with the newer style. As usual, I depend on Barbara's opinion quite heavily. She's the one that originally picked the Expedition theme, and she likes this better. So do I, so I'm changing over to it.

That's me up there in the top left corner. I don't know what I was looking so unhappy about when that picture was taken, but I had it handy so I used it. Actually, that picture originated with some shots I took for my Images page. DigitalThink, for whom I wrote a series of Windows NT MCSE training courses, took the original photograph and did the blue background fade on it. I figured they wouldn't mind me using it here since (a) it's a picture of me and (b) I took the shot originally. They wouldn't let me use the shot of me wearing a Viking helmet, more's the pity...

Sunday, November 15, 1998

Okay. I got half a dozen responses concerning the new page format. All of them said it looked a lot better, and most also said the pages download faster. So I'm now in the process of overhauling the whole site, page by page. This'll take a while, and I probably won't bother changing all the older, less active pages, at least not right away, but I'll change stuff as I have time to get to it.

Today, I help Barbara get her office into shape. We still have some shelves left to install, and I have quite a bit to do on her main PC.

I spent most of today reorganizing the web site, adding text here and there, and helping Barbara get her office straightened out. Tomorrow, back to the O'Reilly book, the one I really want to be working on.

Coming Soon (I hope)

Experiments with Universal Serial Bus (USB)