md5() algorithm, probabilities

kind of off-topic... Does anybody know what´s the probability producing two equal strings with md5() in relation to used string length? The md5 docu told me that its for strings with arbitrary length, since md5 is 128 Bits only there must be a probability in relation to used string length I think. The point is that I´ve used md5 on string <100 chars and now I´ll use it on a combination of primary key data to generate a URL transmittable unique identifier and I *think* it should be unique (p<0,00001 will be sufficient I hope). Anybody experience on this? regards -- o----------0-¬---------O-·---¬----o---®-----o o O ° . | http://www.kiffen.de | pRoteçt y0ur bRaín |0 O ° ¤ ° · 0°·³°²'²³-¹'³´³°^°³~³²³°'³²²¨³²^³¹³²°²³`³º³°Þ ° o © ° . · | psychedelic experience | gott@kiffen.de | O ° o ° o-¬--o--0-----©-·--O-----o-----0-¤----------o 0 ° · ° . ¤ ·