The view that PIE case endings, nominative *-s and genitive *-ós, are somehow related by a magical ablaut and stemming from an ergative case came about from the fact that, based on the wealth of data from world languages that we now have, nominative cases which mark the subject of a sentence are supposed to be unmarked cases. The ergative theory seemed to help solve this problem for some since it provided a specific pathway for this unusual marking to develop. Yet it now has a strong disproof based on the very topological issues it was based on. Witness page 56 of Archaic Syntax in Indo-European - The spread of transitivity (2000) where the theory is artfully destroyed in a pair of brief sentences:
"Yet cross-linguistic analysis has pointed out that ergative marking affects first of all inanimates, and only later animates. The 'ergative' marking patterns of Proto-Indo-European therefore do not fit the noun hierarchy as proposed by Silverstein (1976) and therefore no longer support the traditional ergative hypothesis for Proto-Indo-European."That's what I call cerebral zing. As some often do when they find themselves on the wrong side of Athena's heartless sword, they pick up the fragments of their precious theory, caress them, dote on them, even try to glue the shattered remains back together again in order to breath life back into them by whatever unnatural means necessary. Let's move on. This ergative theory is going nowhere and we're left with only what I've said all along: The PIE nominative comes from an encliticized version of *so, formerly an independent, uninflected definite article (later absorbed into the paradigm of *to- 'that' only to mark the animate nominative). Read Christa König who writes in Case in Africa (2008), p.180:
"A marked-nominative case can go back to a former preceding definite element, resulting in stress and vowel change in the head noun, and the nominative is marked by vowel change and vowel reduction. Evidence for this pathway comes from Berber languages."She then goes on to explain a pattern that is precisely what I predict for Pre-PIE: "In some Berber languages, case is only encoded with definite nouns, in others with all nouns." A similar pattern is observable in Ancient Etruscan, a language whose nominative and accusative cases are unmarked for nouns but marked in definite postclitics, implying that an overt distinction between subject and object is only found in definite nouns. And let's face it, isn't it a lot less taxing on the brain to derive nominative *-s from pre-syncopated Mid IE animate definite article *sa (> PIE *so) than to try and turn the entire Indo-European case system on its head just to unravel the mysteries of this one suffix? Rest your weary heads, my stoic brethren, and heed Occam's Razor.