[stevenAthompson]

I would argue that all models are inherently incomplete because they are models (IE - they are the map not the territory). Rather than worrying about completeness, it's better to ask if the model is useful, and if anything would change about the requirement for tradeoffs in security if we used a more complete model?

I would answer that the triad IS useful in this scenario and further that if we used an alternative model (The 7-C's maybe?) we would still find inherently contradictory requirements for almost every security scenario. In fact, we would just MORE more of those trade-offs, further proving that security can never be "perfect."

For example, I can think of several fundamentals the triad doesn't cover directly. Privacy and non-repudiation spring to mind as concepts that don't neatly fit into the CIA triad, but they are the antithesis of each other!

Perfect privacy would require that nobody (including data-owners) can identify the user, and perfect non-repudiation would require that no access be granted without 100% proof of the current user. Again, you are forced to choose and this means that some aspect will always be less than perfect.