It is often said that the lack of a well established theory explaining the near constancy of M0V is one of the shortcomings of the GCLF method for measuring distances. However, even a method that has a solid theoretical basis can have hidden pitfalls when put into practice (e.g. the difference between population I and population II Cepheids confused the early distance estimates to M31). The final arbitrator of the success of a method is the empirical data, not the theory. Just because we do not understand the reason for the near universality of M0V does not mean there is not a physical mechanism that is responsible.
There are two basic approaches that have been used to try to understand why GCs have a characteristic mass, and therefore a characteristic luminosity since the mass-to-light ratio for GCs are generally similar. The first is to identify some aspect of the initial conditions that preferentially select a certain mass scale for formation (e.g. Peebles & Dicke 1968, Fall & Rees 1985, 1988, Kang et al. 1990, Murray & Lin 1992, West 1993, and Kumai, Basu & Fujimoto 1993). One of the difficulties for many of these models is the near constancy of M0V over a wide range of metallicities.
The second approach is to assume that proto GC's form over a wide range of scales, but destruction mechanisms destroy clusters outside some preferred range. For example, Fall and Rees (1977) argue that destruction from dynamical friction for high mass clusters, disk shocking for large clusters, and evaporation for small clusters results in a "survival triangle" responsible for the near universality of M0V. Similar studies have been done by Aguilar, Hut, & Ostriker (1988), Chernoff and Shapiro (1990), Capaccioli, Piotto, & Stiavelli (1993), Fukushige & Heggie (1995), and Murali & Weinberg (1996).
Observations of young clusters in the ongoing merger NGC 4038 / 4039 by Whitmore & Schweizer (1995) have provided some empirical support for this basic approach. They find that the luminosity function increases as a power law (L)dL L-1.8 dL. This is similar to the luminosity function found in open clusters (Elson and Fall 1985), as well as giant molecular clouds (Sanders, Scoville, & Solomon 1985), supporting the suggestion by Jog & Solomon (1992) that globular clusters are formed in giant molecular clouds, or in the cores of postulated supergiant molecular clouds (Harris & Pudritz 1994, McLaughlin & Pudritz 1996; but see also Schweizer et al. 1996).
This might explain the bright end of the GCLF, with the faint end being whittled away by one or more destruction mechanisms. The near universality of M0V in the presence of a wide variety of Hubble types, environments, galactocentric positions, and metallicities suggests that the destruction mechanism is more inherent to the clusters themselves (e.g. evaporation of stars by two-body relaxation within the cluster) than to their environment. It should be possible to study the process of cluster destruction in detail by observing merger remnants of different ages (e.g. Whitmore et al. 1996).
Blakeslee & Tonry (1996) develop a framework designed to determine whether creation or destruction mechanisms are more sensitive to environmental effects. At present the existing database is not able to distinguish these two possibilities from other effects such as a dependence on Hubble type or metallicity (see 3.2.4). With the rapidly increasing number of high-quality GCLFs it should be possible to test these and other theories in the near future.