winscape 2252

The board shuffle discussed above, is what happens in the field. Debugging happens in a vendor lab. But many vendors of cheap peripherals may not have adequate lab facilities to do proper debugging either. Since most of the logic is in a few large chips anyhow, the makers of boards and devices are at the mercy of chip makers anyhow, who may not have provided enough information to debug at that level either.

A lot of the real hardware debugging happens in simulation at the chip design houses. The interfaces on the chips are designed to work with the published specs for the busses and adjacent chips, and the simulation exercises the various types of handshakes and validates that responses are driven to the chip's outputs within the allowed timing windows. But if the designer "forgets" to simulate some of the more esoteric interactions that may happen on the bus when different chips from different vendors interacts under obscure combinations of operational cycles, it results in errors that are very hard to debug - for anyone!!

When you think through the implications of this, you should be very very impressed with the people that write highly reliable device drivers for high-end peripherals on Linux, often without support from the makers of the peripheral controllers they work on. And unsurprised at how crappy Windows systems continue to be even though XP is so much better than previous releases (somewhat offset by how much more crap has to run on a Windows box nowadays).

For a real depressing story about home Windows boxes, google for "Sony rootkit digital rights".

Lars Poulsen