Maybe a little analogy will help:
Pa:li is a Middle Indo-Aryan language, the language of the Buddhist "canon"
(Tipitaka). It has the following vowels: /a a: i i: u u: e o/. The
phonemes /e/ and /o/ have two allophones: [e], [o] in closed syllables,
[e:], [o:] in open syllables.
The analogue of the traditional explanation of Brugmann's law in
Indo-Iranian would be to posit that "proto-Pa:li" had short /e/ and /o/,
which became lengthened in open syllables in Pa:li (but for some reason the
same did not apply to /a/, /i/ and /u/). In the other languages descended
from "proto-Pa:li", such as modern Hindi, this lengthening did not take
place, and we have only short /e/ and /o/.
The analogue of my explanation would be that the ancestor of Pa:li had only
_long_ /e:/ and /o:/, which became shortened in closed syllables in Pa:li,
but were shortened everywhere in other languages, such as Hindi.
We know of course that for Pa:li, the second explanation is correct:
Sanskrit had long /e:/ and /o:/ (from earlier */ai/ and */au/), which, due
to the fact that there were no short /e/ and /o/ to oppose with, were free
to become shortened, first in closed syllables (the situation in Pa:li),
eventually everywhere (the situation in Hindi).
=======================
Miguel Carrasquer Vidal
mcv@...