And of course, when IBM did try to do a next-generation PC architecture with MCA and did apply lots of systems engineering and thought, the result was hideous. We produced one MCA_based peripheral, but the experience was so horrible and the market so puny it was very easy to walk away from. Even basic tasks like being able to install and test prototype device drivers were battles - the combination of IBM's tight control over what you could and couldn't do, and the shifting sands of OS/2, was just no fun at all. Quick, dirty and open won in the end.