In Part 1 of this article (June '93) I described how the AC subscriber list got started and how disaster was narrowly averted when our original database setup was suddenly discovered to be inadequate. In this issue I'll wrap up the story of the world's most battle-hardened database.

The January Panic
    I had originally installed the AC subscription database on a Seagate fixed-media ST-128N 20-meg SCSI hard drive. I had this drive set up in two partitions of 10 megabytes each, with the first partition being my boot drive and the second one containing all the TurboFile program modules and the magazines's database files. After the October Crisis things settled down a bit, and I grew comfortable using TurboFile and the hard drive.
    Around the middle of January 1993 I went to print up the mailing labels for the February AC, and something nasty happened. As the Seagate was booting up in SpartaDOS 3.2d, a demon bit my computer. It only executed part of my STARTUP batch file, then dumped me into the Sparta command processor (it should have loaded the Sparta menu). Well, OK, maybe it was just a glitch. I powered down, then re-booted. Uh-oh, it happened again. It was October '92 all over agan as that sick feeling of terror seized me. I fought the urge to panic and tried to think logically.
    I reasoned the drive must have somehow lost track of the location of the two command files my batch file was supposed to execute as part of the bootup sequence. The easiest way to scope things out was to load them by hand. I typed MENU and got an error: not a good sign. I knew the file MENU.COM was on the drive; it had booted properly dozens of times before. So I typed DIR to have a look at the directory from the Sparta command processor. I was totally unprepared for what happened next.
    Instead of the usual filname/extender/file size/time/date format, all I got was row upon row of the numeral "1". That did it. The last vestiges of reason went out the window as I succumbed to full-blown panic. My hard drive was a goner. And with it, the AC subscriber list. You see, this was in the days before I got Religion: I felt so invincible with my hard drive I hadn't made backups of the database for a couple months. Some 350 paid subscribers had just gone up in smoke, and sifting through the rubble of the wrecked database to reconstruct the missing people would be several months' work. And just as the February issue was due for mailing! For quite some time I stared in stupefied astonishment at the rows of 1's on my screen. I finally decided this was some bad dream, turned off the machine and went to bed.

Another Close Call
    Next day I called ACs Technical Consultant, Bob Puff. I had hardly spoken a whole sentence about the bootup problem when he breaks in, real cheerful-like, "Did you get a directory full of 1's?" I got excited. "Yeah, yeah! Is my drive damaged?" "Naaaah", says Bob, "your drive is fine". Whew! Big relief! Now we're getting somewhere. I casually mention the magazine's precious database is also on the ailing drive, can I save it? "Oh, no" says Bob in a nonchalant, almost cavalier tone, "You'll have to low-level format the whole drive and rebuild your partitions and stuff, so you'll lose all the data you had on it." My heart sank; that was not what I wanted to hear!
    Then Bob launched into a tekkie discourse about how SpartaDOS can mess up your FAT and other vaporous weirditudes reported by people who use Sparta with Seagate embedded SCSI drives. The longer he talked, however, the more I became convinced the problem was confined to the first partition, and that the second partition containing the AC subscriber list was probably unscathed.
    I was right. I used the BB menu to boot from a floppy instead of the damaged hard drive partition, and from there I was able to get into the TurboFile data on the hard drive's second partition. (I always leave lots of empty sectors between my partitions as a buffer; maybe this helped.) You can bet I rescued those files as fast as my sweaty palms could type! I then reformatted the damaged partition with SpartaDOS and dumped all my boot files back onto it. The magazine's precious database files had never been in any real danger, but I didn't know that.
    The entire exercise was a religious experience, leading to my present Monday night rituals. This Religion consists of a single Commandment: Thou Shalt Back Up!

Good-Bye Seagate, Hello SyQuest!
    The January Panic really unnerved me, for it had demonstrated just how vulnerable the database files really were. I had a second Seagate ST-128N lying around and decided it was time to press it into service as a backup. This turned out to be extremely inconvenient, as I had to physically exchange drives due to lack of space, the whole thing got to be a real pain. I needed a better way.
    I saw some guys at work walking around with these SyQuest cartridges, and it immediately piqued my curiosity. I started pestering them with questions. Were these drives SCSI? (Yup.) Were they fast? ("Faster than a speeding Seagate!") How much memory capacity? (Your choice: 44 or 88 megabytes.) Was the cartridge thing reliable? ("Haven't had one fail on me yet.") I started getting that trembly feeling like I get when I'm on the verge of some great discovery: Since I still had a lot of doubt- especially as to compatibility with the Black Box- I decided to purchase one locally. That way, if the thing didn't work, at least I'd be able to return it with a minimum of hassle. The only local store that sold them was CompUSA, and I almost went into hyperventilation when I saw the price (the prices have since fallen quite a bit, though as hard drives go the SyQuest is still expensive). But I paid for it with a credit card, so it wasn't real money anyway (at least, not until the end of the month!).

You Get What You Pay For
    Smartest move I ever made. In five months of hard use the SyQuest has proven completely reliable. Its SCSI implementation appears very robust, and it lives happily in digital harmony with the Black Box. I pooled resources with friends at work, and between us we placed a bulk order (at an extra discount) for 44-meg cartridges at $62 each including shipping. Sure, the drive mech is expensive, but then each cartridge is the equivalent of a whole new drive, and a cartridge is a lot cheaper than a fixed-media drive. You could fill a bookshelf with them, and in the space of six linear inches you could probably store every single piece of software ever written for the Atari 8-bit (!). But that's just razzle-dazzle.
    Removable media was the perfect solution to my immediate problem. I now always have at least one hard backup to the magazine database. Making the backup copy is a snap. Storing it is easy (it goes on the bookshelf astride my computer desk). If a cartridge fails, or the computer demons launch another attack, my defenses are prepared: I just whip out my spare cart and I'm ready for business again.
    What if the SyQuest mechanism itself fails? No problem. When a fixed-media drive fails and you send it away for repair, all your data is wiped out when they return it to you. But with a removable media drive, you keep your data and just send off the mechanism to be fixed. I didn't like even that idea much, so after hunting in the pages of Computer Shopper I purchased a spare mechanism by mail, much cheaper than the first one. Today, if there's a failure of either media or mechanism, it won't impair the operation of this magazine at all. ACs most precious resource is now about as bullet-proof as can be. Not only because of the SyQuest hardware, but also because of my newfound Religion. The only way we'll ever lose this irreplaceable asset is if some idiot targets the village of Frazer as Ground Zero for World War III.

Labels: The End Product
    I began this tale of the AC database with a suggestion for you to have a glance at your mailing label. That little label makes all the difference in the world when it comes to the delivery of your magazine and your software disk. Without it, nothing moves. Nobody gets to read about the Classic Atari computers. And every single label is generated by me.
    A label print run begins with firing up the database and doing several sort runs. The output of a TurboFile sort run is stored in a pointer file which is suspiciously small. That's because the sort file doesn't actually store the data to be printed, but only vector data that another program module uses to retrieve data from the main database in sorted order. For the June '93 issue I did three sort runs: one run to compile a list of airmail labels, another for North American deliveries with American addresses sorted in ascending Zip Code order, and a third sort that compiled labels for all the subscribers to ACs Software Disk. The vector data files are stored on disk. Then I go into TurboFile's print routine, follow the prompts, and start dumping data to my ancient Atari 1025 printer which has been loaded up with labels.

Prehistoric Relic
    Wait a minute. A 1025? Anyone who ever had one knows the 1025 is a truly primitive printer. We're talking 7-pin dot-matrix here, no logic-seeking, no true descenders, text only (no graphics capability). It's slower than a tortoise and noisy enough to wake the dead. And I'm using this prehistoric relic alongside all this other hi-tech hardware? I must be crazy, right?
    Crazy like a fox. When it comes to the printing of labels, I don' care a whit for speed, graphics, or true descenders. What I do look for is a printer with a tough, reliable mechanism. Running lots of labels can wreck a printer. But the mechanism in the 1025 is indestructible. I can commit the most hideous atrocities against it, it doesn't care. A typical atrocity is manually adjusting the platen while the printer is running, in order to align the labels. (Try that on an Epson, you'll blow the Epson's platen motor drive transistors- not nice!) As to graphics, I've seen other people print fancy mailing labels with little pictures on them. Where are all those labels now? In the trash! Fooey. Descenders? You don't need them when your label is printed in all caps (as ACs labels are).
    What about speed? I printed about 900 labels total for the June magazine and disk. The print runs took several hours each, and I could care less. Once the printer is aligned and the labels are properly centered, I know from experience that no slippage or hangups occur in that rocksolid 1025/Okidata mechanism. While the 1025 is slowly grinding out labels, I'm doing stuff. Hey, summer's here, I got grass to cut. Change the oil and filters in the car. Take my shirts to the laundry, do a little shopping, back to the laundry to flirt with the girl at the counter. Get the picture? You can bet your grandmother's teeth I'm not standing around watching the printer crank labels.

Blank Label Blues
    Blank label stock is becoming a problem. The 1025 has a fixed platen width, requiring a carrier 9-1/2" wide. That size is getting really hard to find. Further, the labels must be two-across with the carrier perforated down the middle. At the moment the only place that sells such labels is Radio Shack- at a ripoff price, of course. Even worse, as soon as Radio Shack discovers that Ben Poehland is desperately dependent upon some product they're selling, they discontinue it. Happens to me every time. I have this love/hate relationship with Radio Shack, much like I had with Atari Corporation before I lost interest in the Sunnyvale Follies.
    Unicorn Publications recently tried to help out with a shipment of 10,000 labels they got at a great discount from Viking Office Supplies. But when I attempted to use them to print labels for the April issue, I found the carrier was 1/4-inch too short for the 1025. I hadn't measured the platen width properly; having to return them was a real bummer, and Viking didn't sell the 9-1/2" size. I tucked my tail between my legs and returned to the Rip-off Shack for more labels, snarling all the way.
    The perforation down the middle is real important. There are 22 labels per carrier page, two columns of 11 labels each. I calculate roughly how many pages of labels I'll need, tear off that many pages, and stuff the lead page into the 1025. TurboFile only prints one record (label) at a time, so the first run through prints half the labels all in one column on the left side of the continuous carrier. The printer runs out, and the program stalls halfway through. Having completed my shopping and flirting, I turn the whole mess around, installing the label carrier from the opposite end. I restart the TurboFile print routine, and it prints out the other half of the batch on the opposite side of the continuous carrier. I tear it in half down the middle via the perforations, and voila! I have two immense strings of single-column labels all ready to go.

Final Delivery
    As with the processing of new subscriptions, all this industry would be for naught if the labels don't get delivered. I place the magazine labels- in two batches separated with rubber bands- into a 2nd Day Air pouch addressed to Unicorn Publications. Our Publisher passes them directly to the printmailer who uses them for the mass mailing of your magazine. The disk mailing labels go in a similar pouch addressed to Alex Pignato, our Disk Distribution Manager. These pouches are placed along with all the rest of the outgoing magazine mail to be delivered to the post office on my way to work the next morning. Which is to say, they get delivered to the post office in my shopping bags. You know, the kind with rope handles, like bag ladies have... .

Postscript: Whither TurboFile?
    Sadly, MicroMiser Software abandoned the Classic Atari market around the same time Atari Corp. and ICD did (January '92). And with it went some of the most powerful business software ever written for our machines. Not only are MicroMiser's products no longer available, they refuse to support the products they've already sold. Just one more tragic episode in the larger panorama of a declining market.
    For AC, it's worse than that. Because despite its overall excellence TurboFile has an Achilles Heel: a truly evil little bug that causes it to overwrite its own data files under certain conditions. I discovered this the hard way when several paid subscribers vanished into the aether in December 1992 (I've since restored the missing subscribers). Being now aware of the conditions that trigger the bug, I avoid it by personal vigilance. But the potential for human error is still very real, and the integrity of the database is at risk.
    Naively, I reported the problem to MicroMiser, thinking they'd fix it (I did after all shell out $50 for this thing). Their response ran something like, "Huh, we sold 100 copies of that program and no one ever complained. Anyway, tough toodles, we're not in that business any more." In the background I heard a flapping sound, as of an 8-bitter twisting in the wind. It was me.
    There is hope. In May of this year I commenced negotiations with MicroMiser for the acquisition of rights to TurboFile. Never having done anything like this before, I'm finding the process far more frustrating and tedious than I ever imagined. As of this writing (early July), things have progressed to a point where I'm now confident of ultimate success by early autumn. Aside from the immediate objective of shoring up the foundations of this magazine, eventually I hope to see TurboFile once more offered in the 8-bit market after the code has been spruced up.