Slavic <voda> "water" is the most notorious exception to Winter's Law
(the law that causes vowels to be lengthened in Balto-Slavic before
PIE *d, *g(W), but not before *bh, *dh, *g(W)h).
As I stated earlier, I believe the word "water" had two related but
separate paradigms in PIE: the normal form (**wá:d-an-) which
developed into NA *wódr, G *wédnos, AbI *wédnot, LD *wédni, and the
collective form (*wad-á:n-), which developed into NA *udó:r, G *udéns,
*IAb *udént, DL *udén(i). Now in Slavic, the first set should have
further developed to *vódU(r) or *vódI(r) [I suppose], Obl. *vén- (?),
without Winter's Law, which is blocked by an immediately following
resonant (-r, -n-).
The second set would have normally developed to *vydá, *vydé~ (> *vydy
by simple analogy, this not being a soft stem), with Winter's Law
transforminmg *(v)Ud- to *(v)yd-. Cf. Latv. u:dens "water".
To get at the attested a:-stem <voda>, <vody>, we only have to assume
that *<vódU(r)> and *<vydá> interacted, producing the "mixed" form
<vodá>.
Forms derived from "water": Slav. ve^dro "bucket", because of Lith.
vé:daras, Latv. ve:ders probably from *ve:der- < *wed-er-o-. Russ.
vëdro "nice weather" < *vedr-U < *wed-r-o-. I have trouble explaining
the length in Slav. <vydra> "otter" (Lith. ú:dra, Latv. u:drs). Why
didn't -r- block Winter's law here?
=======================
Miguel Carrasquer Vidal
mcv@...