The get_points_dist method in PySAL operates only in euclidean space, so you would need to project your input data to use it.
However you could use the arcdist method.
Note: Point tuples should be provided in (Long, Lat) order.
For example:
>>> import pysal
>>> pt0 = (-74.0, 40.7)
>>> pt1 = (-77.0, 38.8)
>>> pysal.cg.arcdist(pt0, pt1)
arcdist does have some limitations, it assumes a perfect sphere and will not take into account the geoid (WGS84). It defaults to KiloMeters, to change this you must specify the radius of the sphere (earth) in whatever units you need. Constants for KM and Miles are provides,
pysal.cg.sphere.RADIUS_EARTH_KM
pysal.cg.sphere.RADIUS_EARTH_MILES
Example:
>>> pysal.cg.arcdist(pt0, pt1, pysal.cg.sphere.RADIUS_EARTH_KM)
The KDTree in PySAL will also work with spherical coordinates and will provide accurate nearest neighbor results, even at the anit-meridian (dateline/180 lng/-180 lng) and the polls.
>>> pts = [(-179.0,0.0), (170,0.0), (179.0,0.0)]
>>> kd = pysal.cg.KDTree(pts, distance_metric='Arc', radius = pysal.cg.sphere.RADIUS_EARTH_KM)
>>> d, i = kd.query((-180.0,0.0), k=3)
>>> d
array([ 111.19492664, 111.19492664, 1111.94926645])
>>> i
array([0, 2, 1])