Erik,

I worked for a firm that was always reverse engineering product from major corporation (XEROX/CANON/Panasonic...). This was because they would not provide information regarding their device behaviour.

The "blind" approach you describe is certainly one way and engenders the least theoretical risk. However, in the US, there was much litigation in this arena in the late 20th Century and Craig's approach is the current working definition.

An interesting Variant of this is the SCO lawsuit vs. Linux et al. Where it was charged that SCO source turned up on Linux and source listings where provided to prove this. It turned out in some cases it was Linux that had turned up in SCO.

regards,
p