Phenomenologically, two classes of GRBs (long/soft vs. short/hard) are identified based on their γ-ray properties. The boundary between the two classes is vague. Multi-wavelength observations lead to identification of two types of GRB progenitors: one related to massive stars (Type II), and another related to compact stars (Type I). Evidence suggests that the majority of long GRBs belong to Type II, while at least the majority of nearby short GRBs belong to Type I. Nonetheless, counter examples do exist. Both long-duration Type I and short-duration Type II GRBs have been observed. In this talk, I review the complications in GRB classification and efforts in diagnosing GRB progenitors based on multiple observational criteria. In particular, I raise the caution to readily accept that all short/hard GRBs detected by BATSE are due to compact star mergers. Finally, I propose to introduce “amplitude” as the third dimension (besides “duration” and “hardness”) to quantify burst properties, and point out that the “tip-of-the-iceberg” effect may introduce confusion in defining the physical category of GRBs, especially for low-amplitude, high-redshift GRBs.