You can't divide vectors in general. If you happen to know that (P-P0) and (P1-P0) are proportional, you can compute their dot product and then divide by the square of the length of (P1-P0).
P = P0 + t * (P1 - P0)
P - P0 = t * (P1 - P0)
dot_product(P - P0, P1 - P0) = dot_product(t * (P1 - P0), P1 - P0)
dot_product(P - P0, P1 - P0) = t * dot_product(P1 - P0, P1 - P0)
t = dot_product(P - P0, P1 - P0) / dot_product(P1 - P0, P1 - P0)
t = dot_product(P - P0, P1 - P0) / length_squared(P1 - P0)

According to the help (it seems to be in the help page for mrdivide), MatLab is solving a least squares problem, which in this case can be done the way I described. So, yes.

[quote name='akaitora' timestamp='1312766167' post='4845990']
So recently I came across a problem where I needed to solve for "t" from the line equation

P = P0 + t(P1 - P0)

that is actually an equation system. You look at each component separately:
P.x = P0.x + t*(P1.x-P0.x)
P.y = P0.y + t*(P1.y-P0.y)

you can now solve both of these equations separately (using the formula you already derived, just applied to each component). Now in the general case this system doesn't have a solution since you will get two different solutions for t. A solutions only exist in the case where P lies on the line defined by P0 and P1. In that case you get the same t from both equations.[/quote]

Let me finish that thought. When you have a system of equations with more equations than variables, you generally cannot solve it exactly, but you could still find a value of t that works well as a compromise for all the individual coordinates. You do this by minimizing the distance between P and P0+t*(P1-P0), which is a least squares problem. If you work out the problem (you can do this by orthogonal projection or by differentiating the square of the distance with respect to t), you'll arrive at the solution I posted earlier.