We present photometry and analysis of the microlensing alert MACHO 96-LMC-2. The ~3% photometry provided by the Global Microlensing Alert Network follow--up effort reveals a periodic modulation in the lightcurve. We attribute this to binarity of the lensed source. Microlensing fits to a rotating binary source magnified by a single lens converge on two minima, separated by delta chi^2 ~ 1. The most significant fit X1 predicts a primary which contributes ~100% of the light, a dark secondary, and an orbital period (T) of 9.2 days. The second fit X2 yields a binary source with two stars of roughly equal mass and luminosity, and T = 21.2 days. The lensed object appears to lie on the upper LMC main sequence. We estimate the mass of the primary component of the binary system, M ~2 M_sun. For the preferred model X1, we explore the range of dark companions by assuming 0.1 M_sun and 1.4 M_sun objects in models X1a and X1b, respectively. We find lens velocities projected to the LMC in these models of v^hat_X1a = 18.3 +/- 3.1 km/s and v^hat_X1b = 188 +/- 32 k/ms. In both these cases, a likelihood analysis suggests an LMC lens is preferred over a Galactic halo lens, although only marginally so in model X1b. We also find v^hat_X2 = 39.6 +/- 6.1 k/ms, where the likelihood for the lens location is strongly dominated by the LMC disk. In all cases, the lens mass is consistent with that of an M-dwarf. The LMC self-lensing rate contributed by 96-LMC-2 is consistent with model self-lensing rates. (Abridged)