Scheme is defined in Revised^(2,3,4) Reports on the Scheme Programming Language, it also has an IEEE/ANSI standard (IEEE Standard 1178-1990, "IEEE Standard for the Scheme Programming Language", published by IEEE in 1991. ISBN 1-55937-125-0.) that mostly coincides with R4RS.
Up to and including R3RS, there was no formalized definition of a macro facility, and every implementation had a different one. R4RS introduced hygienic macros, a humungous definition of "clean" macros that avoided the well-known unique-name-creation and name-clashing problems. Much gnashing of teeth ensued, since many people need macros only for the most primitive tasks, and think the hygienic macro facility to be massive overkill.
Another bone of contention is the non-presence of an eval in the standard, but the presence of load and various file-access functions.
Finally, some implementors think that a Scheme without call/cc could be implemented more efficiently and would fill a real-world ecological niche. (There are already enough Schemes without call/cc out there, but they have to live without blessing by the standards.)
R5RS is supposed to include (different) macros, eval, dynamic-wind and multiple return values.
(22.03.95: still there)
(22.03.95: address seemed ok, but transfer timed out on me)
(22.03.95: astonishingly good connection from Berlin)
(22.03.95: still there.)
``is a new effort starting between LCS and the AI Lab to develop useful software packages in Scheme for use by research projects and for distribution on the net.''(22.03.95: still there.)
This html version of its documentation (rev. 0.95, for SLIB version 1d2, May 1993) (a sterling example of ugly and inefficient markup) was helpfully generated by the texi2html converter. Thank the godess that there aren't any mathematical formula in the texinfo source.
(22.03.95: this is a local copy, and therefore not likely to change)