To create FITS arrays representing the images of objects in the sky requires projection of the celestial sphere onto the array plane. The original FITS papers did not specify how this process was to be carried out. The first approach to this problem, implemented in the Astronomical Image Processing System (AIPS) and described in AIPS Memo 27 (Greisen 1983), has been widely used, but no conventions have been formally endorsed by the IAU FITS Working Group. E. Greisen and M. Calabretta (1996; hereafter GC) have expanded the AIPS approach into a comprehensive proposal for conventions to be used in projecting spherical coordinates onto a two-dimensional plane. This proposal has evolved in response to considerable community discussion and is intended for formal submission to the IAUFWG. Their proposal discusses a number of other World Coordinates issues as well.
The remainder of this section (4) will describe the GC proposal, some other proposed conventions that are gaining acceptance, and additional FITS community practices. Following these conventions in creating a data set will make it easier for many others to read and understand the data. In particular, designers of FITS files that will be distributed outside the originating site should not adopt conventions that conflict with those described in this Guide without consulting with other members of the FITS community, particularly the anticipated readers of the file. In particular, to prevent confusion, avoid using the keywords and notation of the conventions described here with different meanings. Because these conventions have not been formally approved as part of standard FITS, data sets that use other conventions are not out of conformance. However, these conventions are widely used and have the best software support; creators of FITS data that use different conventions will probably need to supply supporting software to their users.