GRBs are short, non-thermal bursts of low energy -rays. It is quite difficult to summarize their basic features. This difficulty stems from the enormous variety displayed by the bursts. I will review here some features that I believe hold the key to this enigma. I refer the reader to the proceedings of the Huntsville GRB meetings [42, 43, 44, 45] and to other recent reviews for a more detailed discussion [37, 38, 39, 40, 41].
A "typical" GRB (if there is such a thing) lasts about 10sec. However, observed durations vary by six orders of magnitude, from several milliseconds  to several thousand seconds . About 3% of the bursts are preceded by a precursor with a lower peak intensity than the main burst . Other bursts were followed by low energy X-ray tails . Several bursts observed by the GINGA detector showed significant apparently thermal, X-ray emission before and after the main part of the higher energy emission [64, 65]. These are probably pre-discovery detections of the X-ray afterglow observed now by BeppoSAX and other X-ray detectors.
The definition of duration is, of course, not unique. BATSE's team characterizes it using T90 (T50) the time needed to accumulate from 5% to 95% (from 25% to 75%) of the counts in the 50 keV - 300 keV band. The shortest BATSE burst had a duration of 5ms with structure on scale of 0.2 ms . The longest so far, GRB940217, displayed GeV activity one and a half hours the main burst . The bursts GRB961027a, GRB961027b, GRB961029a and GRB961029b occurred from the same region in the sky within two days  if this "gang of four" is considered as a single very long burst then the longest duration so far is two days! These observations may indicate that some sources display a continued activity (at a variable level) over a period of days . It is also possible that the observed afterglow is an indication of a continued activity .
The distribution of burst durations is bimodal. BATSE confirmed earlier hints  that the burst duration distribution can be divided into two sub-groups according to T90: long bursts with T90 > 2sec and short bursts with T90 < 2sec [71, 72, 73, 74, 75, 76]. The ratio of observed long bursts to observed short bursts is three to one. This does not necessarily mean that there are fewer short bursts. BATSE's triggering mechanism makes it less sensitive to short bursts than to long ones. Consequently short bursts are detected to smaller distances [74, 77, 78, 79] and we observed a smaller number of short bursts.