[sterlind]

Not an expert at all, but you can think of lattices as evenly-spaced grid points in a vector space. Given a set of basis vectors b0..bn, and arbitrary integers a0..an, a0b0 + ... + anbn are points on the lattice b.

You can have a "good basis" where the norms for b are low, or an equivalent "bad basis" with the same lattice points but with high norms. That's one hard problem (lattice reduction), but there are polynomial-time approximations.

The shortest vector problem, iirc, is to find the vector with the smallest norm in the best possible basis of that lattice.