I have long maintained that the single best use of a million dollars to advance math and science would be to hire a small team of developers to make a LaTeX that doesn't suck. The amount of grad student, postdoc, and professor time wasted on dealing with it is staggering.

Can anyone give me an order-of-magnitude estimate for how much it would cost to make an "enterprise-quality" version of LaTeX? Something with the same powers, but say, the ability to use columns without reading web forums, with packages that don't conflict, and with human-readable errors.