Why is the matrix model correct?

We consider the compactification of M theory on a lightlike circle as a limit of a compactification on a small spatial circle boosted by a large amount. Assuming that the compactification on a small spatial circle is weakly coupled type-IIA theory, we derive Susskind's conjecture that M theory compactified on a lightlike circle is given by the finite N version of the matrix model of Banks, Fischler, Shenker, and Susskind. This point of view provides a uniform derivation of the matrix model for M theory compactified on a transverse torus T-p for p = 0, ...,5 and clarifies the difficulties for larger values of p. [S0031-9007(97)04677-2].