Praudhamanorama p144

p144

Why [is it said] "ending in a non-affix"? In those very cases, let it not apply to the two [bases] distinguished by a case ending. And it should not be said that by treating it as an final element (antavadbhāva), the status of nominal stem would certainly apply in kāṇḍe, kuḍye, and so forth, and consequently there would be shortening by "hrasvo napuṃsake..." and the origination of su; because the desired result is accomplished through the view of [operation at] the time of application. Even in the view of [operation at] the place of enunciation (yathoddeśapakṣa), a single substitute for a case ending is not treated like an end element, as shown by the explicit mention "saptamyadhikaraṇe ca" and so forth. Or let the mention of prātipadika in the rule "hrasvo napuṃsake..." be a general indicator for the stated point.
As for what they say—that the exclusion of an affix need not be made, because it is already achieved by the exclusion of that which ends in an affix, since by the treatment of a single entity as if it were distinct, it too ends in an affix—that is incorrect; because when an affix is mentioned, that which begins with the element after which it was enjoined is grasped, and an isolated [affix] is not such. Otherwise, by your line of reasoning, sup and tiṅ too would end in that and therefore have the status of an inflected word; and consequently in hariṣu, karoṣi, and so forth, retroflexion of s would not occur, due to the prohibition "sātpadādyoḥ". Since [the meaning] could be achieved by adhātupratyayau, the use of two negatives is for the sake of clarity.
The use of a great technical name is out of conformity with Vedic revelation. And thus it is read in the Atharvaveda: "What is a verbal root? What is a nominal stem? What is an affix?"
Unanalyzed words such as ḍittha are the examples here. And older grammarians say that this very [rule] is an indicator for the view of unanalyzed forms.