Date: 1993-11-19
From: John
Subject: First Smoothing Snapshot

My Dear Mr. Duk:

After several frustrating nights of bug-hunting I have finally succeeded in producing a "smoothed" coastline using the micro continent generation method. I put "smoothed" in quotes because the resulting coastline is anything but smooth.

The attached screenshot shows the main map window from my OracleCard stack; I recommend that you import it into a graphics program so that you can magnify the details. In order to keep the processing time down to hours instead of days, I only smoothed a section of the map 28 coordinates wide by 20 coordinates high. Of the 560 squares in this region a whopping 313 were coastal. Total pixels involved in the shoreline region were 16,566.

For this first attempt I used a limited initial set of land sparks. I choose one spark at random along each edge of a land-based seashore coordinate, a total of 187 sparks. The percentage of sea pixels eventually rose to almost 60%. If you look closely you can see the initial sparks in red.

The coastline produced is interesting, but flawed in several respects. The inward rushing of the sea was more vigourous than I had anticipated; as a result, the initial land sparks tend to form lacy arms that stick out from the rest of the coast. Much of the sea extends uninterrupted to the edge of the shoreline area so that many straight line segments and, even worse, intact 90° corners are visible. And when the land sparks do take hold, too often they push all the way to the edge, revealing more straight line segments. The overall effect is boxy and unnatural.

Although not yet acceptable, the technique certainly produces dramatic coastlines and I think, with some adjustment of the initial spark selection, can be much improved. My next experiment will be to return to my first instinct and select entire edges of land-based seashore coordinates. (I will probably omit the pixels near the corners of exposed pixel blocks except for one pixel on the corner itself; that one pixel guarantees more rounded corners.) I'll send the results of that experiment as soon as they are available.

Yours in haste,