Explanation of "Case-sensitive"

[Dan DeVries]

I'm suprised this hasn't come up.

In (K&R) p179 it states that external identifiers may be constrained
by the loader.  It gives 2 examples (IBM,Honeywell) where the loader
only uses one case.  Therefore, the argument about increased naming
space is not true for external identifiers (assuming portable code is
desirable).  The loader may map Tree and tree to the same identifier.

I have yet to here a good argument for cases being distinct, and have
seen several that say it's a bad idea.  So, once again:

Why is C case sensitive?  And who can we blame for this mistake?