The paper argues that the groups of sounds in Ancient Greek dialects assumed to be long diphthongs could have been originally heterosyllabic. They either evolved to the long vowels (due to the sandhi phenomena and the generalization of this variant of the variable in the word positions other than final) or contracted to short diphthongs directly.
The second change is also postulated for archaic times (language of the epos, Mycenaean). The long diphthongs were then probably absent from the Proto-Indo-European, too.