The Shareware Model and How Doom Spread

The received version of Doom’s story is about technology. Carmack’s renderer, the BSP trees, the speed. All true, all repeated to death, and all of it explains why Doom was impressive rather than why it was everywhere — and being everywhere was the achievement. Impressive things fail to spread constantly.

What put Doom on every machine in the developed world was a business model that id inherited from a man selling text-mode dungeon crawlers by post, and that model reached back into the design and rearranged it. Episode one of Doom is shaped the way it is because of how it had to travel.

Shareware was an honesty system

The idea predates games and comes out of the early PC utility scene: ship the software free, ask people to pay if they find it useful, and rely on conscience. Bob Wallace’s PC-Write and Jim Knopf’s PC-File were doing this in the early eighties, and Wallace is generally credited with the word “shareware” itself.

As an honesty system it is obviously flawed, and for a word processor it more or less worked anyway, because the people using PC-Write at work had a budget and a functioning sense of obligation. It does not scale to teenagers.

Scott Miller’s contribution, at Apogee from around 1987, was to notice that you could keep the free distribution and remove the reliance on conscience by splitting the product. Give away a genuine, complete, self-contained chunk. Sell the rest. Nobody has to be honest — they have to want more, which is a much more reliable human property.

Kingdom of Kroz was the prototype. Three episodes, first one free, the other two available if you sent Miller money. It worked well enough that Apogee built a publishing business on it, and well enough that Miller had to invent an audience to talk to at the start — the story that he wrote fan mail to his own company under assumed names is one he’s told himself, and it’s the most 1987 thing in the industry.

What episode one had to be

Here is where the model stops being marketing and starts being design, and this is the part that gets skipped.

If your free chunk is a demo, you can make it a tutorial. Modern demos are usually the first thirty minutes with the good stuff withheld, because the demo is an advertisement for a purchase that will happen in a shop where the full game sits on the same shelf.

A shareware episode cannot do that. It’s travelling on its own, by copy, with no shop behind it. The person playing it got it from a friend’s floppy or a BBS at 2400 baud, has no idea who made it, and has to be moved sufficiently by the free part to find a telephone, call Texas, and read out a credit card number to a stranger. That is an enormous amount of friction, and the only thing that can push someone through it is a free experience that was genuinely, completely satisfying and stopped somewhere unbearable.

So the constraints on episode one are brutal and specific:

It has to be complete. A full arc, its own boss, its own resolution — an album side that stands on its own.

It has to be the best material you have. You are spending your strongest levels on the part nobody pays for, which feels insane and is correct, because the free part is the sales pitch and the paid part is the reward for a decision already made.

It has to end on a wound. Complete, and then a door closes.

Doom’s Knee-Deep in the Dead is nine levels that teach the entire game, escalate cleanly, and finish with two Barons of Hell in a room with the lights on — a boss fight the player is barely equipped for. Then you’re dropped somewhere worse and the episode is over. The level design in that first episode still teaches better than most tutorials, and part of the reason is that it was carrying a commercial burden no tutorial has ever carried.

And it has to be small. A couple of megabytes, because the distribution channel is a modem and a floppy disk and a person’s patience. Every asset in Doom is fighting for space in a file that has to fit down a phone line in 1993 and survive being copied by hand across a school.

Copying is the channel

The other inversion is that shareware treats duplication as free advertising.

The rest of the industry spent the eighties and nineties fighting copying with code wheels, dongles, manual look-ups and increasingly deranged disk protection. Shareware’s answer was to make the copyable part the marketing and put the transaction somewhere copying can’t reach — a phone number and the three episodes that never went out for free.

That’s an elegant piece of engineering, and it explains id’s complete indifference to piracy of the shareware episode. There was nothing to pirate. The thing was already free. Every kid photocopying a disk was doing unpaid distribution work, and a percentage of the people at the end of those chains rang Texas.

Then id kept the door open in a second way, by separating the engine from its data. Doom loads its content from a WAD file, which Carmack structured deliberately so the game’s guts could be replaced, and within months there were level editors and thousands of homemade maps in circulation. A shareware game that keeps generating new free content stays alive on the network indefinitely. The channel doesn’t go quiet — it’s the same instinct that turned the Quake engine into an industry three years later.

The model was a publishing industry

Apogee’s real product was the pipeline rather than any one game, and by the early nineties there was an entire American software business running on it that had almost no European equivalent.

id’s first commercial hit came out of it: Commander Keen: Invasion of the Vorticons went out through Apogee in December 1990, three episodes, first one free, written by people still technically employed elsewhere. Wolfenstein 3D followed through Apogee in May 1992 on the same structure. Tim Sweeney built Epic MegaGames on the identical model from 1991, starting with ZZT, and Epic is still standing thirty-odd years later off the back of it.

Britain never had this, and the reason is that Britain had already solved the same problem with sellotape. A country with a dense magazine culture and a national newsagent network could put a free game in a hundred thousand hands physically — the covertape did the job a modem was doing in Texas, with an editor picking the software instead of a bulletin board. Two channels, same economics, different infrastructure. The American one turned out to have more room to grow, because a phone line has no print run.

10 December 1993

The launch is the model’s high-water mark and it’s a good story because it broke.

id put the shareware episode up on an FTP server at the University of Wisconsin–Parkside, by arrangement with an administrator there, at around midnight. By every account from the people involved, there were thousands of people already sitting on the connection waiting, and the machine and the university’s network went down under them almost immediately. They had to keep juggling servers to get the file out at all.

Nobody had planned for the demand because nobody had a way to measure it. There was no storefront telling you the wishlist number. The game had been trailed on BBSs and in magazines and the entire pipeline was word of mouth, so the first number anyone got was a crashed server.

Within a year Doom was reportedly banned on corporate and university networks — Intel and Carnegie Mellon are the ones usually named — because deathmatch traffic was saturating LANs. That’s a company having to write policy about a game nobody in the building bought. The model had put the software everywhere before the institutions noticed it existed. By 1995 Microsoft was shipping a Windows 95 promo video with Bill Gates spliced into a Doom level, which is what capitulation looks like.

Where it went

id abandoned the model at the moment it succeeded. Doom II shipped in October 1994 as a straight retail box through GT Interactive, because once your name is the marketing you no longer need a free episode doing the selling, and boxes make more money per unit.

That’s the trap in the whole design. Shareware is a tool for the unknown, and it works by giving away your best work to buy attention. Succeed and you stop needing it, and the generosity that got you there becomes a cost you can cut.

The mechanism keeps coming back, though, because the underlying problem never went away. Free-to-play is shareware with the paywall moved and the honesty removed. Early access is shareware run backwards, with the customer paying to be the QA department. The demo returned, went away again, and returned. Every version is a fight over the same question: how much do you give a stranger, for nothing, to earn the right to ask them for something?

id’s answer in 1993 was a third of the game, the best third, ending on a cliff, small enough to fit down a wire. It remains the most confident answer anyone’s given.