Eddington in "Space, Time, and Gravitation" has an excellent discussion
of this if my memory serves. It is true that AT A POINT any
gravitational field can be transformed away by getting in an accelerated
reference frame, but over an extended region that is not the case. The
centrifugal field becomes extremely difficult to explain away using
Mach's principle (that is, by postulating gravitating masses that cause
the effect). Especially swince one sees no extremely dense sources
nearby when rotating the bucket. Thus one must postulate on the basis of
the gravitational effect of the distant stars and galaxies, but the debt
becomes larger (i.e. more and more distant mass is needed) the farther
out you go looking for the gravitating mass that causes the curvature of
the water.
For this reason, many would agree that Einstein's strict adherence to
Mach was useful but misleading, and in fact acceleration with respect to
space itself is meaningful and detectable. Makes perhaps even more sense
when one sees that the microwave background seems to provide an absolute
reference frame even for velocity. Motion relative to the background is
detectable, due to Doppler effects, and the Earth's absolute velocity
with respect to the background radiation is known.