Prior art as a design method
Wed, Jun 18, 2003; by Dave Winer.
Anyone who has worked with me knows how much I value prior art.
Here's how it goes. We're designing a feature, getting ready to implement it. At some point in the design process we ask "Has anyone else done this?" and if so, we consider doing it that way.
Sometimes there are two or three ways to do something, but usually, if something has been proven to work with users, or on current hardware, or somehow has connected with reality and worked, there's usually one best way to do whatever it is. Whoever did it first probably had to iterate, to try one approach and fail, then try another, or see a different way later and re-do it. By respecting prior art you can save all that time.
But there's another even better reason to respect prior art. If we do it the same way -- instead of two ways to do something, there's only one. That means that any software that worked with the other guy's product works with ours. It means that users who know how to use the other product also know how to use ours. It's respectful in the true sense of the word. We listened to you, we thought you were right, so we did it your way.
You can see in a narrative from Evan Williams that he was surprised when the UserLand design algorithm kicked out the obvious answer -- clone the Blogger API even though we already had ManilaRPC which was broader. Only one way to do something is much better than "I have a better way." There wasn't much of an installed base on ManilaRPC outside UserLand. We were aiming bigger. (And we continued to support the original interface and still do to this day. I'm using it to write this story.)
I became a believer in prior art when I made my first end-user product, a simple outliner for Unix. I followed the command structure and style of the Unix line editor (god I forget its name). Later, when I was making Apple II and IBM PC software, I followed the user interface popularized by Mitch Kapor in VisiPlot and then Lotus 1-2-3.
Macintosh was the crowning achievment in design-by-prior-art. They actually had a book of rules called the User Interface Guidelines that told you how you had to do it. Were the criteria subjective? You bet. Did it work? Yes it did. Was it worth the pain? Of course. It made the machine work better for users, and developers. A user who previously could only use one or two Apple II or IBM PC software products now could use five or more Mac products because they didn't needlessly reinvent things that didn't need re-inventing. And of course the Mac itself, an improved clone of the Xerox Star is a great demo of prior art.
PS: Another way to say prior art is "only steal from the best."