IVP and global errors
What we really want to know is not the error in
each single step but how the solution behaves
The solution needs to be stable: small changes
in the initial conditions should induce small
If �a� is positive the solution is unstable