We point out some subtleties with gauge fixings (which sometimes include th
e so-called "brane bending" effects) typically used to compute the graviton
propagator on the Randall-Sundrum brane. In particular, the brane, which h
as nonvanishing tension, explicitly breaks some part of the diffeomorphisms
, so that there are subtleties arising in going to, say, the axial gauge or
the harmonic gauge in the presence of (nonconformal) matter localized on t
he brane. We therefore compute the graviton propagator in the gauge where o
nly the graviphoton fluctuations are set to zero (the diffeomorphisms neces
sary for this gauge fixing are intact), but the graviscalar component is un
touched. We point out that in the Gaussian normal coordinates (where the gr
aviscalar component vanishes on the brane) the graviton propagator blows up
in the ultraviolet near the brane. In fact, the allowed gauge transformati
ons, which do not lead to such ultraviolet behavior of the graviton propaga
tor, are such that the coupling of the graviscalar to the brane matter cann
ot be gauged away in the ultraviolet. Because of this, at the quantum level
, where we expect various additional terms to be generated in the brane wor
ld-volume action including those involving the graviscalar, fine-tuning (wh
ich is independent of that for the brane cosmological constant) is generica
lly required to preserve consistent coupling between bulk gravity and brane
matter. We also reiterate that in such warped backgrounds higher curvature
terms in the bulk are generically expected to delocalize gravity. (C) 2001
Elsevier Science B.V. All rights reserved.