From what I've heard, syllable-timed languages have syllables of equal length throughout each breath-group (i.e. bit of spoken discourse said in one breath), and stress-timed languages have breath-groups of equal length. By this account, the former entails variation in the length of each breath-group and the latter entails variation in syllable length.
However, I could have botched these definitions. Also, I've heard rumors that the timing scheme in a given language can change with rate of speech. I've also heard a rumor that this distinction isn't quite valid.
Therefore I'm asking for a brief and authoritative statement about the difference between stress-timing and syllable timing, and whether this distinction is considered valid by linguists.