This was originally posted to the GG-l mailing list. Production procedures familiar to readers of my posts then are just mentioned in passing, so this version will have annotations to fill in the blanks for what the list readers already knew.

The file timestamp is January 6th, 1994, 7:36 pm

The Arabian Nightmare

Before we start, I should clarify the production process. Artwork would be scanned, shrunk, and inserted into the correct colored border in Photoshop. The borders were just the colorful background. There’s no type at this point.

Meanwhile, the designers and editor are working on a datafile that contains each card’s code number, title, casting cost, type, flavor text, and copyright notice with artist’s name. This gets fed into a custom typesetting system using AmigaTeX which positions all the elements, sets them in the right font (and resets the stuff in the box until it fits), replaces some letters with mana symbols as needed, adds drop shadows to the type on the border, and spits out a card-sized EPS file.

The artwork TIFF files and the type EPS files are laid out on a grid in Freehand (version 3 I believe). The Freehand file just used links to the TIFFs and EPSs, so we could make a change to one of those, and the next time we opened the Freehand file, that change would automatically update.

Then we had to make color separation files. What are color separations? We had to split out the colors from the file into four separate layers: cyan, magenta, yellow, and black. The printing press would have four separate printing plates in these colors, and when they’re all printed on the same sheet of paper, your eye will see a full range of color again. Just like the ink in modern inkjet printers.

While normally people just take their Freehand or PageMaker file into a service bureau, our jobs were so astoundingly huge that no service bureau had the computer time available to do the separations. Also, Freehand could not at that time make separations. No desktop publishing program could. We had a separate program (Aldus PrePrint) for that.

To put this in perspective, the standard method of getting files to service bureaus at this time was the 44meg SyQuest disk. Rarely did a job come in that wouldn’t easily fit on one. The absolutely biggest hard drive available was the brand-new 1 gigabyte hard drive. Our service bureau didn’t even own one. We, on the other hand, had three. The core Magic set of 300 cards occupied about 850 megabytes of drive space. For the typesetting and imagesetting technology of the time, Magic was a monster.

We had to have at least two because all the artwork TIFFs and FreeHand files took up that much space, and the PostScript color separation files ready to feed to the imagesetting equipment at the service bureau would also take up that much space. Our poor state-of-the-art Quadra 840av (40MHz 68040) would take about 3 hours to separate a quadrant (that is, one quarter of one sheet of Magic cards). 3 hours x 3 sheets (common, uncommon, rare) x 4 quadrants per sheet = 36 CPU-hours to create separation files. Also, since the 840av was the main production machine, and we had lots of other products as well, when "make some Magic" time came around, I worked graveyard. I came in around 7pm when everybody else was leaving, and spent all night baby-sitting the computer while it created separation files. So it would take three to four days (well, nights) to make color seps.

Once a hard drive was full of color separation files, it would go to the service bureau, and they’d feed the files to their imagesetter to make film negatives. It typically took the imagesetter about 30 minutes to image one file. 3 sheets x 4 quadrants x 4 files (CMYK) per quadrant = 48 pieces of film x 30 minutes = 24 hours of imagesetter time. We always had a hard time finding a service bureau that’d work with us, because they usually couldn’t commit their biggest imagesetting machine to spending two to three working days just doing our jobs.

When all the negative film sheets were done, they were shipped off to the printer in Belgium. The printer would lay them over photosensitive printing plates in order to etch the patterns on the film onto the plates. Then they could put the plates on the press, ink them up, and let ‘er rip.


“Hey,” I said to myself, “The Regathering fiasco took about a week. Arabian is small, it should go even faster, especially now that the PrePrint error is fixed. Piece of cake.” Of such hubris is torture born. Did I even hear that small voice, speaking to me of how there had been 1,001 Nights, and I might have the honor of experiencing 1,001 Nightmares, all in a single week? No, of course not.

The Regathering blew up mid-week, and I took a day or two to recover, heading into the office Friday evening to putter with the machines. I’d actually laid out some of the Arabian cards during the Regathering work. For those of you unfamiliar with the process, the image files and type .eps files are laid out on a grid before separating, to make life much easier for Carta Mundi. We don’t have an image setter that can print film for an entire printing plate, so we break it into quadrants. Placing card images on the final layout is called imposition.

Prescient Peter was already somewhat frustrated that I hadn’t gotten the datafile for arabian from Beverly on Thursday as originally scheduled. She was sick enough that effective editing was pretty much out of the question during the week, but she recovered enough to get me a file late on Friday. The datafile wasn’t set up correctly yet, so I rearranged the contents and fed it through the typesetting software, archived it, and headed to the office to finish laying it all out. The datafile was missing the artist’s names, and it hadn’t been proofed, but these steps could come later. I’d also printed out a no-images version of the file for Beverly to proof.

The rest of the evening was spent completing the imposition, and test-printing some proof sheets with art in place, a process that turned out to be painfully slow. It takes 16 sheets of paper to show all the cards on a printing plate, and each sheet took 11–15 minutes to print. Arabian has two plates. I wasn’t able to get a list of which artists had done what work from Jesper, since he, like most of the office, was in Philadelphia this weekend, so I reluctantly gathered the art to take back with me, so I could type in artists from their original work. Also, two of the images were in the wrong color frames, and two were missing.

Saturday night was spent fixing typos on the cards that Beverly had caught in proofing, and adding artists to the file. The most amusing thing that happened was the casting cost for Aladdin’s Lamp. A code of 1GG is automatically translated into a 1 in a grey circle, and two green mana symbols. The casting cost of the Lamp was 10, and naturally, this became a 1 in a circle and a 0 in a circle. :) More frustrating was that many artists hadn’t written the name of the work on their art, and I had to try to remember what it had looked like on screen, or deduce by elimination. Two or three artists didn’t even put their own name on the work (like on the unpainted edge, or on the back side), but I recognized the signature from earlier cards. The “A” signature was the only one that I had no idea who it belonged to.

Sunday night was spent cleaning the house and whatnot, since I couldn’t finish typesetting w/o the artists, and Richard hadn’t answered his mail to approve the changes, presumably due to being in Philly.

Jesper got back Monday, and told me who was missing, and that he’d have the missing image files ready for me Tuesday evening. He couldn’t do it that night, since I had the artwork. I took the art back down with me that night, and spent the entire night printing out black and white proof sheets, so that everybody could check for quirks and problems, and Jesper could make sure that we had the right artists named for the cards. This was also a chance to make sure the right type had been placed with each image.

Here’s where things started to go haywire. I had headed home, figuring to go to bed around 8am [Tuesday], so I could be up by 4pm to head down and start making the final film files for color separation. Jesper had a number of questions, as did Vic, and the result was I didn’t get to sleep until well after 3pm. Vic in particular was very concerned about the fact that there were no nifty quotes on any of the cards, and creatures that simply said “first strike,” or “flying,” looked very odd indeed. Such quotes should have been researched weeks ago, and I’d assumed that we just weren’t having any for some reason, but I did allow as how since Beverly wasn’t working on anything frantic at the moment, she did have the afternoon to find interesting quotes before I took the files away for the last time.

When I woke up, I still didn’t have the final datafile waiting for me. Calling Beverly, I found out that Richard wanted/needed to make changes in the cards, apparently based on some playtesting comments. I pointed out to her that any significant changes at this point couldn’t possibly be proofread: I was making the final film that night, and really wanted to start right now. The typesetting would take three hours, then I had to get to the office. Best case had me starting the computers up around 1 am [Wednesday], which meant having files to our service bureau by 8 or 9 am. A day and a half for the imagesetting was cutting it awfully close.

The datafile, as it happened, didn’t reach me until after 2am. That was when I found that there were a couple of blanks in the artists’ credits. The most mysterious was that I didn’t have anybody listed for the Green Efreet. I phoned Jesper’s machine, telling him to call me the instant he got the message, and set my Amiga to typesetting. This, as it turned out, was a portent of something bigger, but I didn’t realize it at the time.

He called back just after it finished, and we picked Ken Murphy Jr. as the artist. He wasn’t sure, because he didn’t have a list at home, but thought that sounded right. I reset that last card, and took off for the office.

It was when I was placing those last cards, around 4am, that I found out what had happened. One of the missing cards was the red efreet, and when I placed it, there was the same image as the green efreet. The artwork was labeled “red;” I hadn’t recorded an artist for the green because we simply didn’t have a piece of artwork depicting it. This card appeared on two of the eight quadrants. I started processing the other quadrants first, and called Jesper around 9am to explain the dilemma with the last two. Through the use of some of his own artwork and magic tricks in Photoshop, he managed to create a Green Efreet and get it to me around noon.

While this was going on, the computer was running even faster than expected, and I took the completed files into the service bureau around 11am [Wednesday]. It’s about a 30 minute trip, if the traffic’s good. This included files for the Arabian Nights Display Box, and the display boxes for the 2nd printing of the Gathering (aka “Unlimited”). These jobs had been added at the last minute, and I was hoping that the service bureau would have time for this much film, since I hadn’t planned on those extra 20 pieces of film in my schedule. After driving from Kent to Fremont, then back to Kent where the Quadra was still thinking hard about color separations, I drove back to Fremont to drop off the last two quadrants around 4pm, then headed back to Wizards for a group birthday party. There was a cluster of October birthdays (including mine), and we were running one party for all three at once.

The service bureau called a couple of times during the party, always causing some trepidation (“Did some file fail?”) but they were always merely trivial enquiries. I’m happy, since I’m scheduled to leave the next day around noon for Incon (a science fiction convention about 250 miles away), and I’m so ready for a teeny little, relaxing, small, quiet, unproductive weekend that I could scream.

The party got over after midnight. Note that I’d been up since 10pm the previous night, and with less than 8 hours of sleep before that. Needless to say, I was pretty strange at this point. Phil and Kaja Foglio had ridden down with me, or rather, Phil had driven us down in my car, so he took us back to his place. I was going to drive home myself, thus minimizing the time I spent behind a wheel.

Now, during all this high-pressure production stuff, I’d had to pretty much ignore my personal errands, and one of the things I’d been putting off until Arabian was over was getting my car license renewed. As luck would have it, tonight was the night some cop noticed, and pulled us over. He was, in the end, rather nice about the whole thing, and I apologized for Phil for putting him in the position of almost getting a ticket (it’s the driver’s fault, no matter who’s car it is), and we all though it would be a good idea if I spent the night at their house. On a mattress on the floor of their art room, but in my condition, concrete would have been comfortable.

As it happened, this was about the same time that our service bureau found that, of the 12 files I brought in, only 2 were going to run: quadrant 1a, and the card back. (Especially ironic, since we ended up not even using the card back, as explained at the bottom of this page.)

Phil and Kaja shared the house with James and Carol, also a married couple, and also good friends of mine. Carol was the one to take me home the next morning [Thursday] after we’d stopped by the service bureau on the way, so she was the one got stuck there while I found out that apparently my work of the previous day was almost a total wash. When it became clear I wasn’t going home any time soon, she took off, and I called the office, asking them to bring the Quadra from the office to the service bureau, so I could try to figure out what was wrong, and get the files to the imagesetter with the minimum possible delay. I spent most of the afternoon wasting yards of photographic film. Here were some of the symptoms: if all the .eps files (containing the type) were on a page, it would fail. I had failure down to 6 .eps’s until I changed the page size. Running just rows worked except for the last row of six .eps files. Running cards 1, 2, 3 worked, and running 4, 5, 6 worked. Running all six didn’t, nor did running all six when placed on the page in the order 4, 5, 6, 1, 2, 3. Removing the outside files worked (2, 3, 4, 5), and running just the two outside files worked (1, 6), but all six together still failed.

I finally made a guess that it was the mana symbols, which, although until now had given me no troubles, had been made by a pretty evil drawing program that I knew was just plain incompetent at making PostScript files. The switch to PrePrint 1.6 apparently let the evilness out, to confound the machine. I talked another friend into giving me a ride home to reset all the .eps files with some mana symbols that had been translated through FreeHand, hoping that the Freehandian versions would make the evilness go away. I was supposed to ride over with the rest of the gang in a car to that convention. I arrange with Cathy that I’ll join them at Incon Friday by flying in around noon.

I regenerate all the Arabian Nights EPS files in a few hours. Unlike the TIFF files, the EPS files are tiny, about 6k each, so a whole set fits on a floppy disk. Four or so hours later, I come back to the service bureau to find that the Quadra’s main system hard drive has failed.


Of course, not only is this what’s needed to make the &$#*! thing run, but it also contains our only copy of the Arabian Nights TIFF files, all 250 megabytes of them.


To get the machine to do anything at all required a startup disk. A Quadra startup disk. A Quadra 840av disk. A Quadra 840av CD-ROM disk. It categorically refused to start with anything else, specifically giving a message saying that other startup disks just weren’t good enough. We tried every startup disk that the service bureau had, and they had a lot of different Macintoshes. The only 840av startup disk I knew of was back at the office in Kent. So, somebody got to drive up with the stupid thing. Jesper had brought the computer itself; this time the pilgrims were Vic and Lisa. We spent the rest of the night working on the machine. When we finally found a way to copy the files off the machine, it went even deeper into catatonia, refusing to acknowledge the existence of any device BUT drive 0, the fried hard drive, resulting in an eternal happy mac face blinking on our screen.

To free it involved opening the case, unplugging the internal drive’s power connector, booting the machine, designating an external drive as a system drive, plugging the internal drive back in (while everything is on, mind you), and issuing a mount command to make this newly-appeared drive available to the system. Kiddies, Don’t Try This At Home. We did it under supervision of the tech. manager of the service bureau, while talking to Apple Technical Support. (Apple Tech Support was very nervous about suggesting this, but their only other viable suggestion was shipping the drive off to a drive recovery firm, and we needed that data RIGHT NOW!) It’s now 5am on Friday. Vic and Lisa stagger home, and I check to see if I can run any files, only to find that 1c, 1d, and all the uncommon imposition files (2a, 2b, 2c, 2d) are corrupt, and have to be redone. I called a taxi and went home, praying that most of the tiff image files would be all right. Incon, to me, is now lying dead in a pool of drive lubricant.

It’s now Friday evening of the day the film was supposed to be in Belgium. I wake up at 4pm, and run to the Licensing office to get a temporary permit for my car for the weekend. Our service bureau closes on weekends in the evenings, so I move the machine to Phil, Kaja, James, and Carol’s house, and assign Carol to rebuild the imposed quadrant files. I’ve been making silly little mistakes for the last four days due to an overload of brain sludge from thinking too hard under too much pressure, and I shouldn’t do detail work any more than a drunk should drive. She finishes recreating the FreeHand imposition files, and I spent time until about 2am [Saturday] generating new color separation files, before collapsing from exhaustion. Four of the remaining seven quadrants are separated at this time.

They say it’s an ill wind that blows no good. In this case, the teeny tiny bit of good is related to land cards. The sheets that Richard turned in had one of each kind of basic land on the layout sheets, along with acceptable substitutions if we needed them. Since I though that getting land would be a drag, and we didn’t have new art for them anyway, I I had made the substitutions, and mentioned it to Richard via EMail. He told me why they had been put in in the first place, and said he’d get some more feedback. That was the last I heard, and it looked like a request to me, so I dutifully put the land back in. My job, after all, as a production person is to realize the product as envisioned by the designer. Wednesday evening at the birthday party, I found out that everybody else hated the lands too, and that Richard had agreed to bag the lands. At which point I got to explain to all and sundry that it was about 9 hours too late to tell me this: the color separation files are already at the service bureau, and it would take two days to reseparate and retypeset the plates now.

So, three days later, when we’re reconstructing the quadrants, I removed the land once again. The exception is the single mountain on plate 1a, the only plate that printed properly the first time. Carta Mundi has been asked to strip the correct card in when they get the films.

Saturday morning I take all these files that are done in to the service bureau, and tell them we need them by Saturday afternoon for FedEx. Then I head back to keep separating the ones that are left. I spend the afternoon running back and forth processing a quadrant at my friends’ house, putting it on a 44meg SyQuest data cartridge (one quadrant would just barely squeeze onto the cartridge), starting the next one, going to the service bureau, having them copy that file to a drive, running back to the house, copying the now-completed file that had been running onto the cartridge, starting another one, running to going to the service bureau, having them copy and run that data file, and so on.

Way back on Thursday I’d also made some changes to try to solve the completely unrelated problems that the display box files had had, and according to the records, all three boxes had been run. However, we can only find output for the booster pack boxes, so I also bring in another set of the files for the Starter Display Box for the Gathering, 2nd printing.

Guess what? It doesn’t run. As it happens, FedEx’s deadline is 5pm on Saturday, at SeaTac Airport, a 30-minute drive, and at 4:15, the 2nd to last quadrant of film is just finishing, and hasn’t been developed yet, or had a blueline proof made. Clearly, some of the film wasn’t going to make it.

Fortunately, a couple weeks ago the secretary at the service bureau had given me this brochure about a courier service. If you’re not familiar with this, they arrange for people to split the cost of an airline ticket to various places, if they take a package with them as part of their checked or carry-on luggage. The courier service indicated that, using a connecting flight through Chicago, we could have our film in Belgium Monday if we were ready by 11pm that night, and when did I want pickup? I said 7, since we were almost done.

This was before I found out that the Starter Box wasn’t going to run that night. A font that we’d used on the other boxes without incident suddenly and inexplicably didn’t taste good to the imagesetter, and it refused to use it. After banging my head against the wall, we pack the rest of the film, and send it off.

Naturally, when I rerun the box file on Monday and take it in, it worked perfectly, and went out that evening.

A Note on the Expansion Symbol

The Arabian Nights booster box features the Magic card back, but all purple-y, like the box itself. That was what the backs of Arabian Nights cards were going to look like. If you’d mixed them into your deck, you’d know you were about to draw one. We’d found that this really was not a problem when we’d been playtesting. After all, playtest cards mixed into a real Magic deck were glaringly obvious, but none of the playtesters had any issues with it.

But when Magic players heard the news, they went ballistic. Mind you, this was the same rabid throng that had told us what utter losers we were to try and scam the public by releasing a game where you didn’t get all the cards when you bought it, but only some of them. Yea, right. Nevertheless, Peter was really worried, and felt we’d better switch the card backs back to the brown of the basic set.

(The only reason, by the way, this was considered was because Carta Mundi could just use the printing plates they already had. If we’d had to make new film, it would never have been changed.)

But changing the card back was also very problematic. City in a Bottle required that a player could identify the Arabian Nights cards, and that was supposed to be done by the color of the back. How could we mark the cards? As you can tell from the previous story, there was absolutely no way whatsoever that we’d be making new film for the fronts!

The printer told us that they could “step and repeat” over the black printing plate to put a symbol on the front of the cards. It had to be black, and it had to be identical on all of them. We figured a scimitar would work, and Jesper and I picked out a spot where we could squeeze it onto the cards.

But how to get this image to Carta Mundi, in Belgium? We’re already a few days late in getting the cards printed. Our fans are screaming, and any further delay means the printer won’t have time to print the cards now. They have other jobs that are scheduled to start. Our job would have to wait until the next open slot in their schedule, which was a couple weeks out.

“We’ll fax it to them.” I announce.

Naturally, the rest of the team looked at me like I was insane. Everybody knows how crappy and jagged faxes look. How can we possibly fax this teensy little scimitar to them in any useful form?

Blow it up first. We faxed over a copy of the simitar that’s eight inches long. An eight inch scimitar at 200dpi (that’s “fine” quality on a fax), is a 3200dpi scimitar when it’s shrunk down to 0.5 inches long.

And that’s what we did.