I just got email from the great Dave Monet, who commented:
Just in passing I should note that I download useful stuff from the JPL Horizons site. So far as I know, they have the best ephemerides. I would certainly never download these things from USNO. For my parallax factor code, I keep a daily table of (X,Y,Z) and interpolate.
I don’t have an opinion on this (although @KSK and @ljones probably do), but I thought I’d be remiss not to pass it on
For planetary ephemerides - where the Earth is and where other planets are - the ephemerides from JPL are indeed the standard.
There are various ‘flavors’ of the JPL ephemeris files, which is something to be aware of. Some cover different date ranges and some include more or less perturbers. The currently most-used ones are DE405 and DE430. (DE430 is the newest, and I’m not entirely sure why everyone’s not just using it, but it may be just a lag in picking up a new release kind of thing, although maybe it’s also because of the nutation model??). Here’s a description of the various versions available: http://ssd.jpl.nasa.gov/?planet_eph_export
Every prediction of the observed location of a moving object will need to use and reference a high-quality ephemeris database like this. If we end up writing our own software, we’ll have to write something to ingest and handle these files, and if we use someone else’s software, we will be using whatever they’ve written to ingest and interpret these files (everyone does it, but no one uses a standard separate package as far as I know actually, I think most people use the JPL provided routines to package the downloadable ascii files into a binary, but then the use may diverge - JPL does provide a toolkit ‘SPICE’).