Kaplan 1998,AJ 115:361-372:High precision astrometry, VLBI & optical:
Abstract: This paper discusses the correspondence between two approaches to astrometric observational reductions: the approach based on angular observables used for optical observations, and the approach based on the interferometric delay observable used for very long baseline radio interferometry (VLBI) observations. A procedure is presented by which VLBI algorithms can be used for optical observations. This scheme can help to guarantee consistent treatment of observational results in the two regimes. Differences between angle- and delay-based algorithms in current use are shown to be less than 1 as. However, the physical models used as the bases for the algorithms must be improved to reach external accuracies at such levels.
Kaplan et al 1989, AJ97:1197, New IAU system, Paper III:
Abstract: A set of algorithms is presented for computing the apparent directions of planets and stars on any date to milliarcsecond precision. The expressions are consistent with the new IAU astronomical reference system for epoch J2000.0. The algorithms define the transformation between fundamental reference data, such as star and radio-source catalogs and planetary ephemerides, and astrometric observables.
Percival 1994, ADASS III, State machine for computing astron. coords:
Abstract: We consider the common problem of computing apparent and topocentric places of stars for the purpose of pointing a telescope. Detailed algorithmic descriptions exist (see, for example, Kaplan et al. 1989). In addition, several software packages such as NOVAS and Starlink's SLALIB by Patrick Wallace considerably ease the burden in building specific application programs. A few problems remain, however. Portability can be a problem, in that some real-time platforms have grudging or non-existent support for Fortran, which is the language of implementation for NOVAS and SLALIB (SLALIB in C is now available). Also, efficiency can be a problem if the subroutines try to do too much, not allowing the programmer to fragment the calculation as needed. SLALIB offers many convenient entry points, which avoids this problem, but the programmer is still left to weave the subroutines together to achieve a desired result.
About the ICRS: http://www.iers.org/iers/earth/icrs/icrs.html
A very nice background page describing "celestial coordinate issues from the AstroGrid wiki": http://wiki.astrogrid.org/bin/view/Astrogrid/CelestialCoordinates
Virtual Observatory Metadata specification for Space Time Coordinates:
A proposed implementation of "crossmatching catalogues": http://wiki.astrogrid.org/pub/Astrogrid/DataDocs/crossmatch.html from AstroGrid